1. Developmental Biology and Stem Cells
Download icon

Inhibition of DYRK1A disrupts neural lineage specificationin human pluripotent stem cells

  1. Stephanie F Bellmaine
  2. Dmitry A Ovchinnikov
  3. David T Manallack
  4. Claire E Cuddy
  5. Andrew G Elefanty
  6. Edouard G Stanley
  7. Ernst J Wolvetang
  8. Spencer J Williams
  9. Martin Pera Is a corresponding author
  1. University of Melbourne, Australia
  2. University of Queensland, Australia
  3. Monash University, Australia
  4. The Royal Children’s Hospital, Australia
  5. Florey Neuroscience and Mental Health Institute, Australia
Research Article
Cited
0
Views
828
Comments
0
Cite as: eLife 2017;6:e24502 doi: 10.7554/eLife.24502

Abstract

Genetic analysis has revealed that the dual specificity protein kinase DYRK1A has multiple roles in the development of the central nervous system. Increased DYRK1A gene dosage, such as occurs in Down syndrome, is known to affect neural progenitor cell differentiation, while haploinsufficiency of DYRK1A is associated with severe microcephaly. Using a set of known and newly synthesized DYRK1A inhibitors, along with CRISPR-mediated gene activation and shRNA knockdown of DYRK1A, we show here that chemical inhibition or genetic knockdown of DYRK1A interferes with neural specification of human pluripotent stem cells, a process equating to the earliest stage of human brain development. Specifically, DYRK1A inhibition insulates the self-renewing subpopulation of human pluripotent stem cells from powerful signals that drive neural induction. Our results suggest a novel mechanism for the disruptive effects of the absence or haploinsufficiency of DYRK1A on early mammalian development, and reveal a requirement for DYRK1A in the acquisition of competence for differentiation in human pluripotent stem cells.

https://doi.org/10.7554/eLife.24502.001

Introduction

The dual-specificity tyrosine-phosphorylation-regulated (Dyrk) kinases belong to a family collectively referred to as CMGC kinases (Aranda et al., 2011) that includes cyclin dependent kinases, mitogen associated protein kinases, glycogen synthase kinases, and cyclin dependent-like kinases The distinguishing biochemical features of Dyrk kinases are their ability to phosphorylate serine, threonine and tyrosine residues, and the autoregulation of their kinase activity through tyrosine phosphorylation. The five mammalian Dyrk kinases are divided into two classes. The Class 1 kinase DYRK1A has been implicated in a diverse variety of biological processes, including central nervous system development, Down syndrome, cancer, beta cell proliferation and diabetes, and Alzheimer’s disease, and the discovery of novel DYRK1A inhibitors has been a goal of many recent studies (Abbassi et al., 2015; Aranda et al., 2011; Shen et al., 2015; Smith et al., 2012; Stotani et al., 2016).

DYRK1A has multiple roles in central nervous system development (Tejedor and Hämmerle, 2011). Genetic studies in mice (Fotaki et al., 2002) and man (Bronicki et al., 2015; Courcet et al., 2012; Dang et al., 2017; DDD Study et al., 2017; Ji et al., 2015; Møller et al., 2008; van Bon et al., 2016; Yamamoto et al., 2011) have revealed that haploinsufficiency of DYRK1A can lead to severe disorders of brain development, including microcephaly, as well as a generalized developmental delay. DYRK1A lies within the Down syndrome critical region on chromosome 21, and an excessive gene dosage of DYRK1A is thought to account for some of the central nervous system phenotypes of this disorder (Duchon and Herault, 2016). Studies of DYRK1A overexpression have elucidated some of its functions during neurogenesis. In embryonic neuroepithelium, a transient increase in DYRK1A expression results in the cessation of the proliferative divisions that expand the progenitor compartment, and premature entry of these cells into a pro-differentiation neurogenic pathway (Hämmerle et al., 2011). In several model systems, DYRK1A overexpression led to exit of neural stem cells from the cell cycle, through mechanisms involving cyclin D1 and p53 (Najas et al., 2015; Park et al., 2010; Soppa et al., 2014; Yabut et al., 2010). DYRK1A gene dosage also affects later stages of neurogenesis, including neuronal dendritogenesis (Benavides-Piccione et al., 2005; Göckler et al., 2009). DYRK1A has also been implicated in tau protein phosphorylation in the pathogenesis of Alzheimer’s disease (Coutadeur et al., 2015).

Previously we showed that the indole derivative ID-8, in combination with WNT3A, could maintain human embryonic stem cells (hESC) in long-term culture under defined conditions in the absence of exogenous activators of the nodal or FGF signalling pathways, both of which are generally considered to be essential for human pluripotent stem cell (hPSC) maintenance (Hasegawa et al., 2012). In the presence of WNT3A, ID-8 modestly enhanced hESC plating efficiency, and strongly inhibited the induction of lineage specific differentiation genes normally observed following WNT treatment of undifferentiated stem cells. Using affinity chromatography, we found that ID-8 bound to Dyrk family members DYRK2 and DYRK4 in extracts of human pluripotent stem cells. We further showed that stable knockdown of DYRK1A and DYRK2 caused a modest increase in the plating efficiency of hESC, but we did not establish whether this effect was related to enhancement of attachment and survival, or to inhibition of differentiation. Thus although these studies suggested an important action of ID-8 on hESC through modulation of Dyrk kinase activity, the actual molecular target of the compound related to its specific biologic activities remained unclear.

In this study we examine the biological activity of ID-8 and a related series of novel indole compounds to determine the role of Dyrk kinase inhibition in stem cell regulation. Human kinome screening, structure activity relationships and targeted gene activation and inactivation studies implicate DYRK1A as the biologically significant target of ID-8. We show that DYRK1A inhibition results in a block to neural specification of human embryonic stem cells. This block is not a uniform response across the entire hPSC population, but instead reflects the ability of DYRK1A inhibitors to insulate the self-renewing subpopulation of hESC from powerful differentiation induction signals. We consider these results in the context of stem cell fate determination, and the deleterious effects of DYRK1A loss on central nervous system development.

Results

Specificity of a series of indole kinase inhibitors for DYRK1A

We examined the specificity of kinase inhibition by ID-8 (Figure 1a) and a related series of novel indole compounds using a biochemical in vitro assay. ID-8 was screened against a panel of 339 human protein kinases by measuring incorporation of radioactive ATP into appropriate substrates (Reaction Biology). Activity (based on percentage inhibition at a 10 μM dose of ID-8) against the top ten protein kinase targets, and several members of the CMGC family, are listed in Figure 1b. The specificity of the compound is displayed in a kinome inhibition map in Figure 1c (for the complete results of the screen, see Supplementary file 1). ID-8 indeed showed selectivity against the CMGC kinase family, with DYRK1B, GSK3B and DYRK1A being the top three kinase targets. Although a biotinylated derivative of ID-8 bound DYRK2 and DYRK4 in affinity chromatography pull down assays (Hasegawa et al., 2012), ID-8 itself showed little activity against these kinases, or against DYRK3. Next we determined IC50 values for a subset of these kinase targets, using the same 33P incorporation assay (Table 1). This study revealed the high specificity of ID-8 for DYRK1A and DYRK1B, and confirmed a lack of activity against DYRK2 and DYRK4. Inhibition of PIM1, GSK3A, GSK3B, and CLK4 at IC50 values between 280–450 nM was also observed.

Figure 1 with 1 supplement see all
Protein kinase inhibition by ID-8.

(A) ID-8 structure. (B) top kinase targets of ID-8 by degree of inhibition at 10 μm, and inhibition of other CMGC group members. (C) kinome tree illustrating the extent of protein kinase inhibition in six families of protein kinases. Illustration reproduced courtesy of Cell Signalling Technology, Inc. (www.cellsignal.com). (D) Structures of indole compounds related to ID-8.

https://doi.org/10.7554/eLife.24502.002
Table 1
IC50 values for ID-8 against a selection of kinases.

Ranks are given based on the original screening data presented in Figure 1B and Supplementary file 1. NC = an IC50 value could not be calculated.

https://doi.org/10.7554/eLife.24502.004
RankKinaseIC50 (nM)
1DYRK1B54
3DYRK1A78
5PIM1280
7GSK3a380
8CLK4440
2GSK3b450
9CLK14200
47CLK2NC
65DYRK3NC
78DYRK4NC
121CLK3NC

We next synthesised a series of ID-8 derivatives modified at the 1, 3 and 6 positions (Figure 1d). The 1 and 6 positions were targeted because modelling studies (discussed below) implicated the –OH and –OCH3 groups as critical to binding at the active site of DYRK1A. The NO2 group was modified since this group might lead to unwanted pharmacological properties in vivo, such as metabolic conversion to toxic nitroso or hydroxylamine derivatives. These compounds were prepared using: (1) the Nenitzescu reaction used for the preparation of ID-8 itself (for compounds 48 and 49); (2) by derivatization of ID-8 (for compounds 28, 29, 30, 34, 35); or (3) by Fischer indole synthesis (compounds 38 and 51). The synthetic pathways used to form these analogues are shown in Figure 1—figure supplement 1 and described in detail in Supplementary file 2.

Biochemical screening of the analogues against the top ten kinase hits for the parent ID-8 are shown in Table 2. Compounds 28, 30 and 48 showed excellent selectivity for DYRK1A and DYRK1B. In particular, compound 28 showed much less activity against DYRK1B, CLK1, GSK3B or LRRK2, relative to ID-8. Compound 30 showed better selectivity against DYRK1A relative to CLK1, CLK4 or GSK3A. Compound 48 showed a similar spectrum of kinase inhibition to ID-8 but was less potent against DYRK1A and DYRK1B. Compound 45 showed activity against all of the kinases in the selected panel and maintained high potency of inhibition of DYRK1A and DYRK1B. Considering the modifications responsible for DYRK1A inhibition, these structure-activity results indicate that that inhibition is very sensitive to modifications at the 6-OH group. Installation of nitro group at the 5-position (compound 51) ablated activity against all kinases tested, except for PIM1 and PIM2. Finally, substitution of the NO2 group at position 3 by an acetylamino group (Compound 34) eliminated activity against all kinases tested.

Table 2
Activity of novel indole compounds against DYRK1A and related kinases relative to ID-8.
https://doi.org/10.7554/eLife.24502.005
CompoundR1R2R3IC50 (μM)
DYRK1ADYRK1BDYRK2CLK1CLK4GSK3αGSK3βLRRK2PIM1PIM3
ID-86-OHNO2p-OCH30.1040.040NC1.371.050.4280.1530.1000.3760.176
346-OCOCH3NHCOCH3p-OCH3NCNCNCNCNCNCNCNCNCNC
286-OCH3NO2p-OCH30.3460.695NCNC2.232.363.440.8540.1760.259
296-OCH2C6H5NO2p-OCH36.523.57NCNCNC18.014.38.1618.58.28
306-OCOCH3NO2p-OCH30.6800.299NCNC11.9522.151.710.8914.151.21
456-OHNO2p-OH0.1220.0602.300.3640.1630.3750.2330.0460.2430.180
486-OHNO2m-OCH30.5940.257NC13.04.772.320.6922.170.6540.593
496-OHNO2o-OCH30.8290.34141.910.78.596.123.821.981.440.892
515-NO2NO2p-OCH3NCNCNCNCNCNCNCNC7.624.57
38HNO2p-OCH3NCNCNCNCNCNCNCNCNC20.74

DYRK1A inhibition blocks neural specification of hESC

In our previous study (Hasegawa et al., 2012), we established that ID-8 treatment modestly increased the replating efficiency of hESC. This effect was further enhanced by addition of WNT3A. WNT3A alone promoted propagation of hESC, but also strongly induced expression of differentiation markers. The addition of ID-8 to WNT3A- treated cells partially suppressed the induction of differentiation markers seen in the presence of WNT3A alone. We therefore first investigated the effects of ID-8 alone on hESC cultured under conditions that promote self-renewal. Propagation of hESC in the presence of ID-8 did not affect the cell cycle distribution of hESC (Figure 2—figure supplement 1a–b), or the overall distribution of the proportion of cells in the culture bearing stem cell surface markers (Figure 2—figure supplement 1c). There was no consistent inhibition of apoptosis by ID-8 (data not shown).

Since we observed previously that the most striking effect of ID-8 on hESC was to inhibit the expression of early lineage specification markers induced by WNT3A, we decided to investigate the effects of the compound on lineage specification in more depth, and to assess the role of DYRK1A inhibition in this effect. Given the importance of DYRK1A in the development of the nervous system, we chose to examine the effects of ID-8 and its homologues on early neural specification in hESC. We used the combined inhibition of BMP and nodal/activin signalling (Chambers et al., 2009) to induce neural specification, and assayed the conversion of hESC to neural progenitors by flow cytometry using a PAX6 reporter cell line developed in our laboratories (Figure 2a, Figure 2—figure supplement 2, and Materials and methods).

Figure 2 with 2 supplements see all
Inhibition of neural specification of hESC by ID-8.

(A) design of time course experiments. Dual SMAD inhibitor induction was carried out from Day 0–7 and cells were assayed for PAX6 expression on Day 16; ID-8 was added from Day 0–16, Day 0–7, Day 0–9, or Day 9–16. (B) phase contrast micrographs showing morphology of control hESC, neural progenitors induced by dual SMAD inhibition, and neural progenitors incubated with 0.5 µM ID-8 throughout the neural induction protocol, on Day 16. (C) dose response study of the inhibition of induction of PAX 6 positive cells by ID-8 showing flow cytometry profile (left panel) and percentage of PAX6 positive cells (right panel). Error bars, SD; *p<0.05, **p<0.0.01. (D) effect of timing of ID-8 exposure at 0.5 µM on inhibition of PAX6 induction. Error bars, SD; **p<0.05, *p<0.0.01. (E) indirect immunofluorescence analysis of NESTIN (green) and POU5F1 expression (red) in control hESC, neural progenitors, and cultures subjected to neural induction in the presence of 5.0 µM ID-8. Nuclear counterstain, dark blue. (F) flow cytometry profiles showing expression of stem cell surface molecules GCMT-2 and CD9 in control cells, neural progenitors and cultures subjected to neural induction in the presence of 5.0 µM ID-8. A-D, studies carried out with HES3 (PAX6mCherry) cell line; E-F, WA09 hESC.

https://doi.org/10.7554/eLife.24502.006

Dual SMAD inhibition led to the appearance of neural rosettes in control cultures and the induction of PAX6 as expected (Figure 2b–c). However, we observed a clear morphological difference between control cells that had been induced with dual SMAD only versus cells that also received ID-8. The former had formed obvious rosette structures characteristic of early neural differentiation by Day 16, whereas the latter more nearly resembled pluripotent stem cells (Figure 2b). Flow cytometry analysis showed that the addition of dual SMAD inhibitors led to the induction of the early human neural progenitor marker PAX6 (Zhang et al., 2010) in about 75% of cells after 16 days. By contrast, in the presence of ID-8, there was a dose dependent reduction in PAX 6 induction to around 5% of control levels at 5 μM (Figure 2c). We studied the time dependency of the ID-8 effect, to determine whether the compound was required during neural induction or later, after transfer to neural progenitor media. ID-8 had to be present during the period of neural induction and neural progenitor formation to achieve full inhibition of PAX6 expression (Figure 2d); it had no effect if added after the neural induction period. The decrease in PAX6 expression caused by ID-8 was accompanied by a decrease in NESTIN staining and an increase in POU5F1 staining in treated colonies (Figure 2e). We also evaluated the expression of stem cell surface markers in cultures subjected to neural induction alone or in the presence of ID-8. We have previously shown that flow cytometry using the cell surface markers GCTM-2 and CD9 can identify a continuum of pluripotent stem cell states within the population (Hough et al., 2014). Induction with dual SMAD inhibition alone resulted in a rapid depletion of the entire pluripotent stem cell population (Figure 2f). The addition of ID-8 had an unusual effect. As in control cells, dual SMAD inhibition resulted in loss of stem cell marker expression across much of the population, but the fraction expressing the highest level of stem cell markers persisted (Figure 2f). The proportion of cells remaining in this fraction relative to the rest of the population was markedly increased (140 fold) when ID-8 was present during dual SMAD inhibition.

Effects of known DYRK1A and GSK3β inhibitors and ID-8 analogues on neural specification of hESC

We then examined the effect of a series of known inhibitors of DYRK1A (Figure 3a) for effects in the PAX6 induction assay. In the same experiments, we also tested a known inhibitor of GSK3ß, to determine if ID-8 activity against this kinase (noted above) could account for its effects on neural induction. CHIR99021, a highly specific GSK3ß inhibitor, is used in some protocols for maintenance of hESC and therefore might be expected to mimic the inhibition of differentiation observed with ID-8 if the latter were acting through the GSK3B pathway. Of the compounds tested, only harmine and CHIR99021 produced statistically significant inhibition of neural induction in the dose tested (Figure 3b). Although some of these previously studied compounds are equipotent inhibitors of DYRK1A compared to ID-8, it is notable that ID-8 and harmine show relative selectivity for DYRK1A and B over DYRK2 and DYRK4 (Supplementary file 3).

Effect of known DYRK1A and GSK3ß inhibitors and ID-8 analogues on induction of PAX6 positive cells by dual SMAD inhibition.

(A) structures of known DYRK1A inhibitors studied. (B) percentage of PAX6 positive cells present at Day 16 in cultures of pluripotent cells, cultures of neural progenitors induced by dual SMAD inhibition (DMSO), and cultures subjected to neural induction in the presence of DYRK1A inhibitors at a dose of 0.5 µM. Error bars, SD; **p<0.05, *p<0.0.01. (C) percentage of PAX6 positive cells present at Day 16 in cultures of pluripotent cells, cultures of neural progenitors induced by dual SMAD inhibition (DMSO), or cultures subjected to neural induction in the presence of ID-8 analogues. All compounds were tested at 0.5 µM. Error bars, SD; **p<0.05, *p<0.0.01. (D) indirect immunofluorescence micrographs showing stem cell cultures subjected to dual SMAD inhibition alone or in combination with 5.0 µM ID-8, 5.0 µM compound 28, or 3.0 µM CHIR 99021 and stained with antibodies to POU5F1, PAX6 or NESTIN (all red) and DAPI nuclear counterstain (dark blue). (E) flow cytometry analyses of stem cell surface marker (GCTM-2 antigen and TG30 anti-CD9 antibody) expression in cultures subjected to dual SMAD inhibition alone or in combination with 5.0 µM ID-8, 5.0 µM compound 28, or 3.0 µM CHIR 99021. B and C, studies carried out with HES3 (PAX6mCherry) cell line; D-E, WA09.

https://doi.org/10.7554/eLife.24502.009

We next tested our series of novel indole derivatives of ID-8 for their activity in inhibiting PAX-6 induction. Only compounds 28, 30 and 45 displayed activity at 0.5 μM in this assay (Figure 3c). Thus, activity in this assay was quite sensitive to any alteration of the parent compound structure. As observed with ID-8, treatment of cells with compound 28 during dual SMAD inhibition resulted in colonies that contained few NESTIN positive cells and an increase in POU5F1 positive cells (Figure 3d). Considering the overall spectrum of kinase inhibition represented in the panel, the results argue that inhibition of DYRK1A accounts for compound activity in the assay. Compounds 34, 51 and 38 were inactive against DYRK1A and showed no activity in the differentiation assay. Compound 51 retained activity against PIM1 and PIM3, suggesting that off-target effects on these enzymes were unrelated to blockade of neural induction. Compound 28 showed much reduced activity against DYRK1B and GSK3B, and no activity against CLK1, but it was still active in the assay. Compound 30 likewise showed negligible activity against CLK1 and much reduced activity against GSK3A. Compounds 48 and 49, inactive in the neural induction assay, were active against CLK4 and PIM1 and PIM2.

Although CHIR99021 treatment did reduce the extent of PAX-6 induction by dual SMAD inhibition, unlike ID-8 CHIR99021 co-administration did not result in retention of a pluripotent phenotype. Following neural induction, colonies formed in the presence of CHIR contained fewer POU5F1 positive cells compared to colonies induced in the presence of ID-8 or compound 28 (Figure 3d). These results were corroborated by studies of stem cell surface maker expression (Figure 3e). Expression of the cell surface antigens identified by antibodies GCTM-2 and TG30 (anti-CD9) was abolished by dual SMAD inhibition. Cells subjected to dual SMAD inhibition in the presence of ID-8 or compound 28 retained stem cell surface markers, but cells treated with CHIR99021 did not. These findings suggest that CHIR99021 was acting in a manner different to ID-8 to block neural induction, most probably through activation of the Wnt signaling pathway to induce alternate neural states (more caudal multipotent progenitors or neural crest) in conjunction with inhibition of stem cell renewal by SMAD2/3 inhibition, as shown previously by several groups (Chambers et al., 2016; Denham et al., 2012; Denham et al., 2015; Menendez et al., 2011).

ID-8 offsets effects of DYRK1A overexpression on neural specification of hESC and DYRK1A knockdown phenocopies effects of inhibitors

Off-target effects of ID-8 on DYRK1B, DYRK2, GSK3B, PIM1 or PIM3 or CLK1 or CLK4 might account for the inhibition of induction of neural differentiation by this compound. GSK3A, DYRK3, DYRK4, or LRRK2 are not expressed in hESC (www.stemformatics.org). The relationships between activity in kinase assays and activity of our panel of compounds in the neural induction bioassay described above would argue against a role for GSK3B, CLK1, CLK4, PIM1 or PIM3. To examine more directly the role of DYRK1A inhibition in the action of ID-8 on neural differentiation, we used a different model, namely a paradigm in which DYRK1A is specifically overexpressed in stem cells (Figure 4a). We used RNA guided gene activation mediated by the CRISPR system to switch on DYRK1A during induction of neural specification. gRNAs were designed to deliver a dCas9-VP64 activator construct to the DYRK1A regulatory region to enhance DYRK1A expression during induction of neural differentiation (Figure 4a). In this system, activation of DYRK1A (Figure 4b) during neural induction led to an increase in PAX6 expression that was offset by ID-8 or the other three inhibitors (compounds 28, 30 and 45, Figure 4c) active in the dual SMAD PAX6 induction assay described above. The DYRK1A inhibitors did not reduce the extent of induction of DYRK1 by CRISPR activation (Figure 4b).

Figure 4 with 1 supplement see all
Effects of CRISPR activation and inducible knockdown of DYRK1A on PAX6 expression during neural specification by dual SMAD induction.

(A) design of experiment to induce DYRK1A; (B) DYRK1A transcript levels following CRISPR activation in control cells or in presence of ID-8, or compounds 28, 30, or 45. Error bars, SD; **p<0.05. (C) PAX6 transcripts following DYRK1A activation in control cells or cells treated with inhibitors at 50 or 100 nM. Error bars, SD; **p<0.05; all compounds produced significant reductions at all doses at p<0.05. (D) inducible shRNA knockdown of DYRK1A in WA09 hPSC. (E) effect of DYRK1A knockdown on PAX6 transcript levels following dual SMAD induction of neural specification. Result of triplicate experiments shown, mean ±SEM, **p<0.01, ***p<0.001. Studies in A and B were carried out with derivatives of hESC line WA09, those in C and D utilised derivatives of Genea022.

https://doi.org/10.7554/eLife.24502.010

To further confirm the role of DYRK1A inhibition in blockade of neural specification, we used inducible shRNA to knockdown DYRK1A during neural specification by dual SMAD inhibition. WA09 hESC line was transduced with three separate doxycycline inducible DYRK1A shRNA constructs. Two out of three clones displayed strong knockdown of DYRK1A in the presence of doxycycline (Figure 4d). During neural specification, these two clones showed doxycycline inducible suppression of PAX6 expression (Figure 4e). Thus, DYRK1A knockdown phenocopies the effects of the inhibitors on PAX6 expression. We then confirmed the effects of inducible shRNA knockdown of DYRK1A in two more pluripotent stem cell lines, GENEA022 (the hESC line used in the DYRK1A activation study above) and C11(an hiPSC cell line). First we demonstrated that our active small molecule DYRK1A inhibitors could block the induction of PAX6 transcripts during neural induction by dual SMAD inhibition in WA09, GENEA022 and C11 cell lines (Figure 4—figure supplement 1A). Then we showed that shRNA knockdown of DYRK1A could phenocopy the effects of the inhibitors in GENEA022 and C11 cells (Figure 4—figure supplement 1B,C). Immunoblotting of detergent extracts of WA09 cells treated with scrambled control shRNA or DYRK1A shRNA in the presence of doxycycline confirmed significant reduction of DYRK1A protein levels by two of the constructs (Figure 4—figure supplement 1D).

Modelling of ID-8/DYRK1A interaction

Eleven X-ray crystal structures are available for DYRK1A in complex with a range of ligands (Figure 5—figure supplement 1). In order to rationalize the structure-activity relationships of the ID-8 analogues, we sought to understand how these compounds bind to DYRK1A by molecular modelling. Seven reported structures deposited in the protein databank were superimposed using Maestro and the ligand binding sites were closely examined (Figure 5a). In the majority of cases the amino group of K188 and the amide nitrogen atom of L241 were involved in hydrogen bonding to the ligands. In these crystal structures containing diverse ligands, the ligands bind to both of these amino acids through polar interactions with either nitrogen or oxygen atoms. The average distance between both interacting atoms on the ligands was 8.3 ± 0.2 Å.

Figure 5 with 1 supplement see all
In silico modelling of binding of ID-8 and compound 45 to DYRK1A.

(A) Overlay of seven ligand-bound X-ray structures of ligand-bound complexes of DYRK1A showing similar positioning of active site residues K188, E203, E239 and L241 (PDB codes: 5AIK, 3ANQ, 3ANR, 4YU2, 4MQ1, MQ2 and 4AZE). (B) Overlay of the docked pose of ID-8 (grey) with the experimentally-determined pose of LDN-211898 (green) with DYRK1A derived from the 5AIK structure. (C) Overlay of docked poses of ID-8 (grey) and compound 45 (orange). Maestro (version 2015–3) was used to dock ligands into 5AIK using the XP algorithm and the default settings were applied.

https://doi.org/10.7554/eLife.24502.012

Using the crystal structure complexed with LDN-211898 (PDB code 5AIK) a set of docking experiments using ID-8 and compound 45 were undertaken. ID-8 docked in three alternative modes, one of which was very similar to the bound states of other DYRK1 ligands, as illustrated by overlay with the experimentally-determined LDN-211898 bound into 5AIK (Figure 5b). In this pose, the plane of the indole ring of ID-8 coincides with the aromatic sections of LDN-211898. ID-8 docks such that there is a hydrogen bond from the 6-hydroxyl to the amide group of L241, which is achieved from a slightly different direction to LDN-211898 by binding deeper into the active site, while the para-methoxy group of ID-8 is bound to E203 (adjacent to the amino group of K188). The distance between the two interacting atoms of ID-8 is 8.22 Å, similar to that of other DYRK1A inhibitors. Similarly, the amino group of K188 hydrogen bonds to the para-hydroxyl group of ID-8 by placing this hydroxyl group deep into the binding site. Compound 45 was able to dock into the binding site of 5AIK (Figure 5c) forming hydrogen bonds between the 6-hydroxyl and the amide nitrogen of L241 (2.97 Å) as well as the backbone carbonyl group of E239 (3.09 Å). The nitro group formed an interaction with the primary amino group of K188 (3.02 Å). The distance between the two interacting oxygen atoms of compound 45 is 7.78 Å. While compound 45 did not dock in the same mode as ID-8 it is easily be envisaged that this mode of binding is possible, as this compound possesses the same key polar interaction oxygens present in ID-8. A confounding factor encountered in docking both ID-8 and 45 is that the nitro group is more polar than the other two hydrogen bonding groups, and the docking algorithm is biased towards placing this group within the site to achieve polar interactions.

Discussion

Here we present evidence that the indole compound ID-8 and a series of related molecules act to inhibit the neural specification of hESC through inhibition of DYRK1A. Our biochemical studies revealed that ID-8 is a relatively specific inhibitor of DYRK1A, and that the activity of a series of related compounds on neural specification most closely correlated to their ability to inhibit this enzyme. Further, in pluripotent stem cells specifically engineered to overexpress DYRK1A, neural specification induced by dual SMAD inhibition is enhanced, and this effect is offset by addition of ID-8 or other DYRK1A inhibitors. Finally, DYRK1A knockdown phenocopied the effect of the inhibitors on PAX6 expression during neural specification.

Our docking analysis provides a set of possible binding modes for ID-8 and compound 45, which allow a rationalization of the potency of ID-8 analogues (Table 2) as inhibitors of DYRK1A. For example, the lack of inhibition for compounds 38 and 51 likely arise from the absence of a hydroxyl group at position 6 which therefore cannot interact with L241. Substitution at this position with a methyl group in compound 28 led to a small reduction in potency, while larger acetyl or benzyl in compounds 29, 30 and 34 led to a marked reduction in potency. Compounds 48 and 49, bearing methoxy groups oriented ortho or meta, respectively, cannot make ideal interactions with E203. Studies in man have implicated missense mutations affecting residues around the proposed binding site of ID-8 and compound 45 in the microcephaly syndrome associated with DYRK1A deficiency (Ji et al., 2015).

hPSC provide powerful models for understanding early human development. The biological actions of the DYRK1A inhibitors described here provide new insight into the effect of the absence of this kinase on neural development in the embryo. In human or mouse, hemizygous loss of DYRK1A leads to microcephaly, a severe disorder of brain development. Some of the phenotypes of Dyrk1a knockout mouse, and haploinsufficient humans (Bronicki et al., 2015; Courcet et al., 2012; Fotaki et al., 2002; Ji et al., 2015; Møller et al., 2008), could be explained by an action at a very early stage of lineage specification in the embryo. Our previous work showed that ID-8 treatment of hESC blocked the ability of WNT3A to induce expression of mesoderm or endoderm genes (Hasegawa et al., 2012). This study reveals that DYRK1A inhibition specifically interferes with the early specification of hESC to neural fates. Together, these findings suggest that DYRK1A loss might exert a profound effect on endowment of the three germ layer lineages in homozygous or heterozygous null embryos. Dyrk1a homozygous null mice die at E10.5, with growth delay in multiple organ systems, including heart, brain, liver and branchial arches. Humans heterozygous for loss or mutation of DYRK1A display severe microcephaly, but also commonly show developmental delay and intrauterine growth retardation. It is possible that failure to allocate sufficient numbers of pluripotent progenitors into the three germ layer lineages could result in generalised developmental deficiencies and delay, along with severe central nervous defects. The neural specification assay here is suitable for rapid screening of the effects of DYRK1A inhibition on neural development in human cells, and could aid in development of novel inhibitors, as could the CRISPR mediated DYRK1A induction assay, which mimics the gene dosage effect of DYRK1A overexpression in Down syndrome. Indeed, the results of this assay suggest that DYRK1A inhibitors could be used to ameliorate neural phenotypes in Down syndrome, though clearly the dose of the agents would need to be controlled rather precisely to avoid excessive reduction in enzyme activity.

Our observations are in contrast to previously reported actions of DYRK1A inhibition in mouse embryonic stem cells, where treatment with epigallocatechin-gallate caused reduced expression of pluripotency genes and enhanced expression of mesodermal and endodermal lineage specific genes (Canzonetta et al., 2008). However, epigallocatechin-gallate is a broad spectrum kinase inhibitor that lacks specificity for DYRK1A, and its effects on mouse ES cells could be attributable to interference with other pathways, or to species differences in stem cell regulation.

The actions of DYRK1A inhibitors on human pluripotent stem cells provide important insight into the role of this kinase in stem cell regulation. ID-8 treatment did not affect expression of stem cell markers in cultures maintained under conditions that support self-renewal. Neither did the compound affect proliferation of stem cells maintained under these conditions. We showed previously that ID-8 could promote colony formation of hESC, but this effect was modest, except in combination with WNT3A. The strongest effect of treatment of hESC with ID-8 observed in our previous work was the inhibition of induced differentiation. In cells treated with WNT3A (Hasegawa et al., 2012), WNT3A induction of mesodermal and endodermal genes was strikingly attenuated by ID-8. Here we showed that the powerful neural induction using dual SMAD inhibition could be offset by treatment of stem cells with ID-8.

Careful inspection of flow cytometry profiles after ID-8 treatment revealed that the response of the cell population to DYRK1A inhibition was not homogenous. In fact, a large proportion of cells treated with dual SMAD inhibition with or without ID-8 lost stem cell surface markers. However, in ID-8 treated cultures, the cellular subset bearing the highest levels of stem cell markers appeared to remain intact, and indeed was proportionally expanded 140 fold in differentiating cultures relative to other subpopulations under the condition of DYRK1A inhibition. We have previously shown that this cellular subset represents the self-renewing component of hESC cultures, and that it has molecular properties distinct to the bulk of the stem cell population (Hough et al., 2014). Recently we have observed that this subpopulation of hESC has an unusual cell cycle, with a short or non-existent G1 phase compared to the remainder of the population (unpublished). It is known that hPSC become competent to respond to differentiation signals in G1 phase of the cycle (Boward et al., 2016). Our results suggest that DYRK1A inhibition by ID-8 prevents the self-renewing subpopulation from acquiring competence to respond to induction of differentiation, rendering it refractory to strong extrinsic inductive signals. Thus, transition to a differentiation permissive state may require activation of DYRK1A, an effect that may be actuated through the lengthening of G1 in the self-renewing population. Inhibition of DYRK1A could prevent the lengthening of the G1 phase and thereby block progression of the self-renewing hESC subpopulation towards neural specification. Further study of DYRK1A will provide additional insight into its role in early development, and the mechanism whereby its activity modulates the competence of hESC to respond to inducers of lineage specification.

Materials and methods

DYRK1A inhibitors and other compounds

LDN193189 (Stemgent, Lexington, MA, #130-096-226), SB431542 (Stemgent, #130-097-448), CHIR99021 (Stemgent, #130-095-555) and harmine hydrochloride (Sigma-Aldrich, St. Louis, MO, #SMB00461) were purchased from commercial suppliers. All other inhibitors were synthesised as described in the Supplemental Methods. All compounds were solid powders, stable at room temperature. Compounds were dissolved in DMSO to achieve desired stock concentrations, and were aliquoted for long-term storage at −20 or −80°C. Stock concentrations were diluted further in DMSO, to desired concentration for adding to cell culture, such that DMSO concentration in cell culture media did not exceed 0.15%.

Kinase inhibition assays

Kinase inhibition assays were performed by Reaction Biology (www.reactionbiology.com). Each kinase was produced as a recombinant protein and used a matched peptide substrate. The reaction buffer was 20 mM Hepes (pH 7.5), 10 mM MgCl2, 1 mM EGTA, 0.02% Brij35, 0.02 mg/ml BSA, 0.1 mM Na3VO4, 2 mM DTT, 1% DMSO. Radioassays were performed using 10 μM ATP and 10 μCi/μl [γ-33P]ATP. Enzyme concentrations were 0.2–50 nM, and peptide substrate concentration was 20 μM. Reactions were incubated for 2 hr at room temperature, then spotted onto P81 ion exchange paper, and product detected by radiometry. For initial screening, candidate inhibitors were included at 10 μM, and activity was calculated as percentage inhibition relative to the control with no inhibitor. For determination of IC50 values, inhibitors were included into kinase reactions at 10 concentrations with 3-fold serial dilutions starting at 30 μM.

Cell lines

WA09 and HES3 (PAX6mCherry)cell lines were maintained by and provided by the Stem Cell Core Facility of Stem Cells Australia. The knock-in reporter HES3 (PAX6mCherry) cells were developed as described (below). hESC lines WA09 (obtained from WiCell (WA09 MEF platform), GENEA022 (provided by Genea Biocells), and hiPSC line C11 (Briggs et al., 2013) were maintained at the University of Queensland laboratories on fibroblast feeder cell support in medium supplemented with Knockout Serum Replacer and FGF-2 (Invitrogen/Thermofisher Scientific, Waltham, MA), with daily media changes. For propagation in defined media, WA9 cells were adapted from culture on mouse embryo fibroblasts (mechanically dissociated organ cultures) at passage 74 or 87, and maintained thereafter on hES-qualified Matrigel in mTesR1 (STEMCELL Technologies, Vancouver, BC, Canada) medium using dispase to disaggregate the cells. HES3(PAX6mCherry) were cultured on hES-qualified Matrigel in mTesR1 medium. Medium was changed daily, and cells were passaged with dispase. Cell line identity was confirmed by STR profiling.

Generation of the PAX6 targeting vector

DNA fragments representing the 5’ and 3’ homology arms were derived from bacterial artificial chromosome clones (BAC) RP11-26B16 and RP11-307I15 (Roswell Park Cancer Institute), which included sequences spanning the PAX6 stop codon. The PAX6 targeting vector was assembled using a combination of recombineering and standard cloning procedures. Sequences representing the 3.6 Kb 5’ homology arm were amplified by PCR and subsequently cloned into a vector containing a PGKNeo selection cassette and sequences corresponding to the 7.8 Kb 3’ homology arm, the latter having been previously introduced into this plasmid using recombineering. This vector was then linearised at an Mlu1 site at the junction of the 5’ homology arm and the PGKNeo selection cassette, enabling the sequential introduction of a T2A sequence that replaced the PAX6 stop codon and, an mCherry-IRES-puromycin cassette. Prior to electroporation, the final plasmid was digested with Pac1 and AsiS1 to release the targeting vector as depicted in Figure 2—figure supplement 2a.

Generation of the PAX6-mCherry reporter line

The HES3 cell line (Richards et al., 2002) was electroporated with the PAX6 targeting vector using standard conditions (Costa et al., 2007). Single G418 resistant colonies were screened for correct integration of the targeting vector using PCR primers that corresponded to sequences within the PGKneo cassette and genomic sequences 3’ of the 3’ end of the 3’ PAX6 homology arm: neo4 (cgatgcctgcttgccgaatatcatg) & 39783R (gccgattcctgagcctttcatacc). Two positive clones identified in this initial screen were then validated by repeating the PCR analysis performed in the original screen and by performing a second PCR analysis that confirmed correct recombination of the vector at the 5’ end. This second PCR, which utilised the primer pairs cherryrev2 (ccatgttatcctcctcgcccttgc) and 27749F (gctaacctgtcccacctgatttcc), generated the predicted product of 8.6 Kb. The result of this confirmatory PCR analysis is shown in Figure 2—figure supplement 2b. These positive clones were expanded and subsequently transfected with a cre recombinase expression vector as described previously (Davis et al., 2008), resulting in excision of the PGKNeo cassette. Successful removal of this cassette was confirmed using PCR analysis. Subclones of each primary clone were then submitted for karyotype analysis, performed by the Cytogenetics Department at Southern Cross Pathology, Monash Medical Centre. A single subclone, PAX6 # 85.2 with a normal female karyotype was then chosen for further analysis.

Neural induction assay

The neural induction assay involved the directed differentiation of hESC to neural progenitor cells, and is based on the dual-SMAD inhibition methods by Chambers et al. (2009). HES3(PAX6mCherry) or WA09 cells in mTeSR1 were passaged using 1 mg/ml dispase onto Matrigel coated 3 cm dishes and cultured in mTesR1. The cells were grown to minimum 30% confluency prior to supplementation with 5 µM SB431542 and 100 nM LDN193189, in the presence or absence of kinase inhibitors diluted in DMSO, or DMSO vehicle only for the controls, with a total DMSO concentration of 0.15% in each dish. Media was changed every 2 days, with continual inhibitor supplementation. At day 7, the cells were transferred into Neurobasal media (Invitrogen/Thermofisher) with FGF2 and EGF, continuing kinase inhibitor supplementation, without LDN/SB inhibitors. At day 9, cells were mechanically lifted/scraped from the bottom of the dish, dissociated to clumps by pipetting, and passaged onto fresh Matrigel coated 3 cm dishes in Neurobasal medium supplemented with 5 μM Rho kinase inhibitor Y-27632, achieving approximately 30% confluency immediately following passage. At day 11, Rho kinase inhibitor was withdrawn, and cells were maintained in Neurobasal media with continual supplementation of DYRK inhibitors with media changes every 2 days, until the cells were fixed for immunofluorescence or harvested for flow cytometry. For statistical analysis, P values were calculated using a paired two-tailed Student’s t-test with correction for multiple comparisons using the Benjamin-Hochberg procedure with a false discovery rate of 0.075. All experiments were repeated 2–3 times with three technical replicates in each data point.

Measurement of pluripotency marker expression, proliferation and apoptosis

Flow cytometry measurement of the expression of the pluripotent stem cell surface markers GCTM2 and CD9, and indirect immunofluorescence microscopy for POU5F1 and nestin on paraformaldehyde fixed adherent cell cultures, were performed as previously described (Hough et al., 2014). The cell cycle distribution of control and ID-8 treated pluripotent WA09 cells was analysed using the Click-IT EdU Flow Cytometry Kit AF488 (Invitrogen/Thermofisher, #C-10425). Apoptosis of control and ID-8 treated pluripotent WA09 cells was analysed using the Abcam In situ BrdU-Red DNA Fragmentation (TUNEL) Assay Kit (Abcam, Cambridge, MA). Both kits were used according to the manufacturers’ instructions. All experiments were repeated two to three times.

Effect of indole compounds on induction of neural specification driven by DYRK1a overexpression

To identify gRNAs capable of driving CRISPRa activation of the DYRK1A gene, a number of unique 20mers in region between −250 to −50 bps relative to the main transcription start site were cloned in the pX462 gRNA and dCas9-VP64-expression vector and tested by transfection into the 293FT human cells, with subsequent qPCR analysis of DYRK1A expression. Generation of stable cell lines was achieved by transduction of the 2 selected best-performing gRNA (gRNA #1 CCGGCAAATACCGCAGTCCC; gRNA#2, CGCTGGAACCGCGAGCCGAG)-expressing lentiviruses (pLKO.1-sgRNA-Neo) into the recipient pre-validated hESC Genea022 line expressing HA-dCas9-VP64 in a doxycycline-inducible fashion. Neural differentiation, used as a backdrop for the inhibitor assay, was performed for 6 days in KSR medium supplemented with 10 µM SB431542 and 0.5 µM LDN193189 and devoid of pluripotency-maintenance growth factors. Control cultures grown ±doxycyline, and cultures treated with inhibitor compounds, contained 0.1% DMSO for the duration of the assay. For statistical analysis, P values were calculated using a paired two-tailed Student’s t-test with correction for multiple comparisons using the Benjamin-Hochberg procedure with a false discovery rate of 0.075. All experiments were repeated 2–3 times with three technical replicates per data point.

DYRK1A knockdown studies

For an inducible knockdown of the DYRK1A expression, we transduced human pluripotent stem cell cultures (WA09, GENEA022, and C11) with VSV-G-pseudotyped lentiviral particles: one control (non-targeting) and three DYRK1A ORF-targeting shRNAs in doxycycline-inducible SMARTvectors (GE Healthcare/Dharmacon, product # DHA-VSC11653 and V3SH11255). The virus was packaged using the 293FT cell line (LifeTechnologies/Invitrogen) transfected with a second-generation lentiviral packaging plasmid psPAX2, a gift from Didier Trono (Addgene plasmid # 12260). Selection for cells harboring expressable proviral integrants was started 5 days after transduction with puromycin at 2 μg/mL. DYRK1A knockdown was induced by treatment with doxycycline at 0.5 μg/mL for 3 days prior to, and during the 6 days of, the dual SMAD inhibition-driven neural differentiation. Effects of small molecule DYRK1A inhibitors and shRNA on DYRK1A and PAX6 transcript induction during neural specification were assessed in the cell lines used for knockdown studies at Day 16 in the protocol described above. PAX6 transcript levels were measured by qPCR on total RNA (0.5 µg)-derived cDNA (using iScript polyA and random-priming cDNA synthesis system, Bio-Rad) extracted from pluripotent cells subjected to a standard (dual SMAD inhibition) neuroepithelial differentiation in presence of the indicated compounds for 16 days. Detergent extracts of shRNA treated cells were separated on denaturing and reducing 12% tris-glycine-sodium dodecyl sulphate acrylamide gels. The samples were prepared using a protease and phosphatase-supplemented mix of RIPA and 2x Laemmli buffers (Sigma-Aldrich). Following gel transfer to polyvinylidene difluoride membranes, the blots were probed using primary antibodies against DYRK1A (D30C10 Rabbit mAb #8765, Cell Signalling Technologies,U.S.A.) and GAPDH (14C10 Rabbit mAb #2118, CST), and a secondary anti-rabbit IgG, HRP-linked Antibody (#7074, CST). Visualisation was performed using Clarity ECL substrate, and densitometric quantification performed using ImageLab4.1 software (Bio-Rad, U.S.A.).

Structural modelling

Eleven crystal structures are available of DYRK1A with bound ligands. These were superimposed using Maestro and an analysis of the binding sites was carried out. Maestro (version 2015–3) was used to dock ligands into 5AIK using the XP algorithm and the default settings were applied. The protein was prepared using the protein preparation wizard.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
    Autism-associated Dyrk1a truncation mutants impair neuronal dendritic and spine growth and interfere with postnatal cortical development
    1. T Dang
    2. WY Duan
    3. B Yu
    4. DL Tong
    5. C Cheng
    6. YF Zhang
    7. W Wu
    8. K Ye
    9. WX Zhang
    10. M Wu
    11. BB Wu
    12. Y An
    13. ZL Qiu
    14. BL Wu
    (2017)
    Molecular Psychiatry, 10.1038/mp.2016.253, 28167836.
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38

Decision letter

  1. Lee L Rubin
    Reviewing Editor; Harvard Stem Cell Institute, Harvard University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Inhibition of DYRK1A disrupts neural lineage specification in human pluripotent stem cells" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Marianne Bronner as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

We have now received two reviews on your manuscript "Inhibition of DYRK1A Disrupts Neural Lineage Specification in Human Pluripotent Stem Cells". While the second reviewer is more enthusiastic than the first, there seems to be general agreement that the manuscript is potentially publishable in eLife after additional experimental work is done.

Essential Revisions:

Since the paper focuses heavily on ID-8 and its analogues, it is very important to show definitively that the action of the compounds is truly via inhibiting DYRK1a. It is not sufficient to show that particular compounds have activities in cell-free systems that appear to translate into their effects on cells. The authors themselves state that some of the enzymes that can be inhibited by the compounds are not even expressed in huES cells. Moreover, it may be that the potency differences observed across a series of compounds can be ascribed to variability in cellular permeability. Thus, as requested by both reviewers, it is key that the authors provide more evidence that the compounds act on target cells at their effective concentrations by inhibiting DYRK1A and not, for example, GSK3β. Furthermore, as requested by both reviewers, more information should be provided as to the details of compound testing in cell-free assays and on the ES cells. These items are addressed in detail by Reviewer 1 (points 1-3); Reviewer 2 (point 3b). In addition, please make clear what conclusions you are drawing from Figure 3. Along those lines, both reviewers also request that you show convincingly using genome modification methods that DYRK1a reduction or deletion phenocopies the effects of your compounds.

In addition, both reviewers would like you to show that the compounds have similar effects on more than 1 huES line. Reviewer 2 has requested a confirmatory study (point 2) that seems important to carry out since it would provide strong support for the mechanisms that you propose.

Reviewer #1:

This paper described the optimization of a small set of ID-8 like compounds as specific inhibitors for DYRK1A and DYRK1B that bind to the ATP binding pocket. Using a computational model, the authors rationalized the SAR within the series on DYRK1A. Using these chemical tools, together with the cell line reporting the level of PAX6 expression, as well as CRISPR-mediated gene activation, the authors argue that the inhibition of DYRK1A prevents the neural specification of huES cells. The ID-8 is previously proposed to be the potential DYRK inhibitor, although not as characterized, and the molecular mechanisms underlying DYRK1A's effect is not characterized in this report.

1) The methods for the kinase assays in Table 1 and 2 should be detailed, including the source of kinase protein preparation, substrate peptides, Km for each peptide substrate, and ATP? Is each assay performed at Km for substrate and Km for ATP? As the IC50s of the chemical compounds depend on the kinase assay condition, such information must be provided, and critical in evaluating the results for specificity. Why the discrepancy between Table 1, and Table 2 in terms of the IC50s of ID8 against DYRK1A (0.054 vs. 0.104μM), GSK3b (0.45μM vs. 0.15μM)?

2) For the evaluation of cellular activities of ID8 series compounds, dose responses of key compounds in this assay, should be included in the Table 2. Cellular toxicity needs to be addressed. At 5 μm, does ID-8 cause any reduction in the number of cultured cells (Figure 2C, right panel)? CRISPR-mediated conditional deletion of DYRK1A would be a direct and conclusive genetic evidence to demonstrate that DYRK1A is necessary to support the neural specification, and should be included in the argument.

3) It is not conclusive whether GSK3b underlies the effect of ID-8 on neural specification, as CHIR99021 has an effect on PAX reporter assay. CHIR99021 has a different profile across kinome, and the noticed difference on Oct-4 expression may be derived from off-target effects of CHIR99021. Providing a GSK3b dependent cellular readout with ID-8 compounds in these cells, such as p-tau, or p-CRMP, could help answer wither ID-8 compounds hit GSK3b in these cells.

4) The author showed that ID-8 compounds blocked the neural specification, most likely through inhibiting DYRK1A. Previously, it is shown that ID-8 promotes the stem-ness of huES cells by promoting self-renewal (Hasegawa et al., 2012). Are the observations in this report merely reflecting the fact that the huES cells in the presence of ID8 have more stem-ness, hence resistance to be induced for any lineage specialization, or that DYRK1A is a key switch for neural lineage specification? Can huES cells be induced into neurons, in the presence of ID8, by overexpressing neurogenin2 that bypass the neuroprogenitor stage (Sudhof lab, 2013 Neuron)? In either case, the molecular targets need to be identified/discussed to provide a mechanism for such observation, are EGFR, FGF, p53 etc. involved in such a functional role (PMID 20237271; PMID 18771760; Ferron et al., 2010).

Reviewer #2:

In their manuscript entitled "Inhibition of DYRK1A Disrupts Neural Lineage Specification in Human Pluripotent Stem Cells," Bellmaine et al. demonstrate that the addition of ID-8, and similar chemical inhibitors of DYRK1A, prevents hESCs from differentiating to NPCs. This is a very clear and straightforward report. With what I hope are some simple additional experiments, this manuscript could be suitable for publication in eLife.

1) Inter-hESC line variability – in the Materials and methods, the authors describe using each of a HES3-knockin-PAX6 reporter, H9 and WA9 hESC lines. It would be helpful if they clarified which lines were used in which figures. Moreover, the manuscript would be improved if the effect of ID-8 was directly compared across all three hESC lines in a single figure panel (i.e. Pax6 qPCR or AB-based FACS).

2) NPCs: The authors suggest that ID-8 treatment is preventing hESCs from specifying to an NPC fate, supported by data showing ID8 doesn't change hESC replication or pluripotency markers (Figure 2—figure supplement 1). To confirm that this is a patterning defect, rather than a replication, death or differentiation phenotype, would it be possible to FACS purify remaining Pax6-GFP positive cells following ID8 treatment and assess their capacity to replicate and undergo neuronal differentiation (either with or without ID8 present)? Specifically, I am curious whether the remaining NPCs generate both neurons and astrocytes, and if neurons, both excitatory and inhibitory.

3) Disease relevance: The authors make repeated note of the presumed role of DYRK1A in neurodevelopmental disease, both in their Introduction and the Discussion, but fail to discuss any similarities between DYRK1A+\- hiPSC phenotypes and ID8 effects. Can they phenocopy DYRK1A+\- phenotypes across an ID8 dose-response curve? Even more important, can they rescue developmental effects in DS hiPSCs across an ID8 dose-response curve? (I understand that this would be the most difficult experiment of all the ones I have suggested, and that DS hiPSCs may not be available to this group – it is just the most interesting question I had in reading this manuscript.)

a) Given that the authors demonstrated an ability to prevent neural differentiation with ID8, even when up-regulating DYRK1A, can they similarly impair neural differentiation by knocking down DYRK1A with dCas9-KRAB? This would confirm the specificity of the ID8 activity.

b) Similarly, DYRK1A is a member of the dual-specificity tyrosine phosphorylation-regulated kinase family. It must have some known downstream targets that the authors can query by western blot, etc., to confirm the specificity of ID-8. If this was done in previous publications, it should be noted and cited.

4) Technology relevance: Does the addition of ID8 improve the efficacy/yield of non-neuronal (i.e. cardiac, muscle, endoderm) differentiation protocols? Are there any known activators of DYRK1A that could be used to improve neuronal differentiation protocols?

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Inhibition of DYRK1A disrupts neural lineage specification in human pluripotent stem cells" for further consideration at eLife. Your revised article has been favorably evaluated by Marianne Bronner (Senior Editor), a Reviewing Editor, and two reviewers.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

After additional discussion with the reviewers, we reached the conclusion that you should still carry out the experiment highlighted as reviewer 2's point #1 (repeating your main assay using 2 additional lines). We believe that this experiment is straightforward to do and will be important in convincing readers of the robustness of your observations.

Reviewer #2:

In their revised manuscript, Bellmaine et al. now better demonstrate that inhibition of DYRK1A prevents hESCs from differentiating to NPCs. The authors were reasonably responsive to my concerns, particularly in adding a genetic inhibition of DYRK1A to complement the pharmacological studies. Nonetheless, two simple experiments remain.

1) Inter-hESC line variability – the authors did a great job clarifying which hESC lines were used in each figure panel. Nonetheless, I still think it would be helpful to compare all three hESC lines across at least a single assay (i.e. Pax6 qPCR or FACS), in order to assess how robust the effects are across hESC lines.

2) NPCs: The authors suggest that ID-8 treatment is preventing hESCs from specifying to an NPC fate. It would be interesting to collect the remaining Pax6-GFP positive cells following ID8 treatment (or shRNA knockdown) and assess either their gene expression and/or their capacity to replicate and undergo neuronal differentiation (either with or without ID8 present)? Specifically, I am curious whether the remaining NPCs generate both neurons and astrocytes, and if neurons, both excitatory and inhibitory.

https://doi.org/10.7554/eLife.24502.018

Author response

Essential Revisions:

Since the paper focuses heavily on ID-8 and its analogues, it is very important to show definitively that the action of the compounds is truly via inhibiting DYRK1a. It is not sufficient to show that particular compounds have activities in cell-free systems that appear to translate into their effects on cells. The authors themselves state that some of the enzymes that can be inhibited by the compounds are not even expressed in huES cells. Moreover, it may be that the potency differences observed across a series of compounds can be ascribed to variability in cellular permeability. Thus, as requested by both reviewers, it is key that the authors provide more evidence that the compounds act on target cells at their effective concentrations by inhibiting DYRK1A and not, for example, GSK3β.

We agree with this point and provide new data and some additional background information to address these concerns. First, with respect to inhibition of GSK3b as a potential mechanism (the active compounds in our study showed some inhibitory activity towards this kinase), we showed in the first submission that the potent and selective GSK3b inhibitor CHIR 99021 also reduced PAX6 induction following dual SMAD inhibition, similar to DYRK1A inhibitors. However, we argued that the action of CHIR 99021 was fundamentally different to that of DYRK1A inhibitors, because treatment with the latter agent but not the former resulted in retention of a pluripotent phenotype. We based this claim on indirect immunofluorescence staining for nestin and POU5F1 following CHIR 99021 treatment shown in Figure 3C (compare to Figure 2E with ID-8). We grant that this evidence on its own is fairly weak, so we now provide further support for this interpretation. First we cite previous studies showing that inhibition of GSK3b in combination with SMAD inhibition in hPSC indeed blocks PAX6 induction but directs cells towards other neural fates-neural crest or more caudal neural precursors (1-4) subsection “Effects of known DYRK1A and GSK3β inhibitors and ID-8 analogues on neural specification of hESC”, last paragraph. These studies (which we should have noted in the first submission) are consistent with our findings on CHIR 99021 but fundamentally different to our results with ID-8. We also provide new immunocytochemistry and flow cytometry data which confirm results in Figure 2F and show that while co-treatment of cells with ID-8 (or another DYRK1A inhibitor, compound 28) and dual SMAD inhibition results in retention of pluripotent phenotype by a significant fraction of cells, co-treatment with CHIR 99021 plus dual SMAD inhibitors results in complete loss of stem cell marker expression across the population, similar to dual SMAD inhibition alone subsection “Effects of known DYRK1A and GSK3β inhibitors and ID-8 analogues on neural specification of Hesc”, last paragraph, and new Figure 3D, E. See also the response to point 4 below.

Furthermore, as requested by both reviewers, more information should be provided as to the details of compound testing in cell-free assays and on the ES cells. These items are addressed in detail by Reviewer 1 (points 1-3); Reviewer 2 (point 3b).

We now include the details of the assays in the Materials and methods subsection “Kinase inhibition assays”. See also responses to reviewers.

In addition, please make clear what conclusions you are drawing from Figure 3.

We apologise that the point of this figure was not explained more clearly. This figure shows that some known (albeit less specific) inhibitors of DYRK1A also block neural lineage specification, that treatment with a GSK3B inhibitor CHRI 99021 has this effect but does not result in retention of pluripotency, and that other potent and specific DYRK1A inhibitors that we synthesised block neural lineage specification. We now include a subheading to introduce the presentation of the data in Figure 3 and a new figure title which we hope will clarify the objective of the studies therein, subsection “Effects of known DYRK1A and GSK3β inhibitors and ID-8 analogues on neural specification of hESC” and Figure 3 legend “Effect of known DYRK1A and GSK3ß inhibitors and ID-8 analogues on induction 588 of PAX6 positive cells by dual SMAD inhibition”.

Along those lines, both reviewers also request that you show convincingly using genome modification methods that DYRK1a reduction or deletion phenocopies the effects of your compounds.

We include new data showing that inducible shRNA knockdown of DYRK1A indeed blocks PAX6 induction by dual SMAD inhibition (Summary; Introduction, last paragraph; subsection “ID-8 offsets effects of DYRK1A overexpression on neural specification of hESC and 256 DYRK1A knockdown phenocopies effects of inhibitors”; Discussion, first paragraph; subsection “DYRK1A knockdown studies” Figure 4 legend and new Figure 4D-E.

In addition, both reviewers would like you to show that the compounds have similar effects on more than 1 huES line.

We have now specified more clearly what hESC lines were used in which studies. Three lines were used in the study: HES-3 PAX6mCherry as the reporter cell line for induction assays, Genea022 for the CRISPR DYRK1A activation studies, and WA09 for most additional work. Although we did not carry out a specific experiment with all three lines, there is sufficient overlap in the endpoints to demonstrate that the effect is not cell lines specific. We have identified the cell lines used in particular experiments in the Figure legends of Figures 2, 3 and 4.

Reviewer 2 has requested a confirmatory study (point 2) that seems important to carry out since it would provide strong support for the mechanisms that you propose.

Please see the response to the reviewer below.

Finally, we have added citations to three new clinical genetics studies (5-7) published since our original report that provide further strong evidence that DYRK1A haploinsufficiency results in severe intrauterine growth retardation and microcephaly in the human (Introduction, second paragraph). These studies further highlight the relevance of our work to human neurodevelopmental genetics.

Reviewer #1:

[…] 1) The methods for the kinase assays in Table 1 and 2 should be detailed, including the source of kinase protein preparation, substrate peptides, Km for each peptide substrate, and ATP? Is each assay performed at Km for substrate and Km for ATP? As the IC50s of the chemical compounds depend on the kinase assay condition, such information must be provided, and critical in evaluating the results for specificity. Why the discrepancy between Table 1, and Table 2 in terms of the IC50s of ID8 against DYRK1A (0.054 vs. 0.104μM), GSK3b (0.45μM vs. 0.15μM)?

We have now included this information: see response to Editor point 2 above. The results in Tables 1 and 2 report independent experiments run separately. These experiments were performed by a commercial research organisation on two occasions approximately one year apart. The differences in absolute IC50 values reflect assay variation relating to preparations of different batches of enzyme and substrate, but the variation does not affect interpretations regarding specificity or relative potency.

2) For the evaluation of cellular activities of ID8 series compounds, dose responses of key compounds in this assay, should be included in the Table 2. Cellular toxicity needs to be addressed. At 5 μm, does ID-8 cause any reduction in the number of cultured cells (Figure 2C, right panel)?

In our previous study (Hasegawa et al., (8)) we observed no toxicity of ID-8 up to 10 µM, and our experience in this study was similar.

CRISPR-mediated conditional deletion of DYRK1A would be a direct and conclusive genetic evidence to demonstrate that DYRK1A is necessary to support the neural specification, and should be included in the argument.

A key point. Please see response 4 to the Editors above.

3) It is not conclusive whether GSK3b underlies the effect of ID-8 on neural specification, as CHIR99021 has an effect on PAX reporter assay. CHIR99021 has a different profile across kinome, and the noticed difference on Oct-4 expression may be derived from off-target effects of CHIR99021. Providing a GSK3b dependent cellular readout with ID-8 compounds in these cells, such as p-tau, or p-CRMP, could help answer wither ID-8 compounds hit GSK3b in these cells.

Another critical issue. Please see the response to the Editors above point 1.

4) The author showed that ID-8 compounds blocked the neural specification, most likely through inhibiting DYRK1A. Previously, it is shown that ID-8 promotes the stem-ness of huES cells by promoting self-renewal (Hasegawa et al., 2012). Are the observations in this report merely reflecting the fact that the huES cells in the presence of ID8 have more stem-ness, hence resistance to be induced for any lineage specialization, or that DYRK1A is a key switch for neural lineage specification?

This is a key point. In Hasegawa et al. (8) we showed ID-8 did indeed promote self-renewal but this effect quite modest except when the compound was used in combination with WNT3A (please see Figure 1 in that manuscript). In fact, the most striking effect of ID-8 observed in that study was the inhibition of WNT3A induction of mesendodermal genes (Figure 1 in Hasegawa et al. (8)), though in retrospect we did not emphasize this enough in that paper. ID-8 in fact has minor effects under conditions that promote hESC self-renewal (in this study see Figure 2—figure supplement 1) but, as in Hasegawa et al. (8) the most prominent effect of the compound is to insulate of subset of hESC from powerful differentiation stimuli. We have added a brief note in the Discussion to emphasize that the effect of ID-8 in our previous study was modest except in the presence of WNT3A (Discussion, fifth paragraph).

Can huES cells be induced into neurons, in the presence of ID8, by overexpressing neurogenin2 that bypass the neuroprogenitor stage (Sudhof lab, 2013 Neuron)? In either case, the molecular targets need to be identified/discussed to provide a mechanism for such observation, are EGFR, FGF, p53 etc. involved in such a functional role (PMID 20237271; PMID 18771760; Ferron et al., 2010).

Though this question of direct reprogramming of ID-8 treated cells is interesting, the interaction of reprogramming factors with pathways altered by the drug is liable to be complex and not easily interpretable.

Reviewer #2:

[…] 1) Inter-hESC line variability – in the Materials and methods, the authors describe using each of a HES3-knockin-PAX6 reporter, H9 and WA9 hESC lines. It would be helpful if they clarified which lines were used in which figures. Moreover, the manuscript would be improved if the effect of ID-8 was directly compared across all three hESC lines in a single figure panel (i.e. Pax6 qPCR or AB-based FACS).

We have clarified which cell lines were used in particular experiments, please see response 5 to Editors above.

2) NPCs: The authors suggest that ID-8 treatment is preventing hESCs from specifying to an NPC fate, supported by data showing ID8 doesn't change hESC replication or pluripotency markers (Figure 2—figure supplement 1). To confirm that this is a patterning defect, rather than a replication, death or differentiation phenotype, would it be possible to FACS purify remaining Pax6-GFP positive cells following ID8 treatment and assess their capacity to replicate and undergo neuronal differentiation (either with or without ID8 present)? Specifically, I am curious whether the remaining NPCs generate both neurons and astrocytes, and if neurons, both excitatory and inhibitory.

We did not look at this specific question in this study because we have previously shown that cells maintained in ID8 can be grown long term without toxicity or loss of pluripotency (8). It would be challenging to recover the PAX6 cells following flow cytometry without loss of viability. A key point of the present study is that DYRK1A insulates cells from potent signals that drive lineage specification, and our work here and in our previous study shows that this is achieved without toxicity, growth inhibition, or irreversible loss of pluripotency.

3) Disease relevance: The authors make repeated note of the presumed role of DYRK1A in neurodevelopmental disease, both in their Introduction and the Discussion, but fail to discuss any similarities between DYRK1A+\- hiPSC phenotypes and ID8 effects. Can they phenocopy DYRK1A+\- phenotypes across an ID8 dose-response curve? Even more important, can they rescue developmental effects in DS hiPSCs across an ID8 dose-response curve? (I understand that this would be the most difficult experiment of all the ones I have suggested, and that DS hiPSCs may not be available to this group – it is just the most interesting question I had in reading this manuscript.)

We speculate in the Discussion that the effects of ID-8 on pluripotent stem cells might account for much of the DYRK1A haploinsufficiency syndrome-namely that failure to specify sufficient neural progenitor cells might result in poor endowment of the CNS and severe effects on brain development observed in patients. Though we do not assert this directly, the experiment in Figure 4 suggests that the drugs could indeed reverse the effects of DYRK1A overexpression (the DS condition). We add a note to this effect in the Discussion (third paragraph).

a) Given that the authors demonstrated an ability to prevent neural differentiation with ID8, even when up-regulating DYRK1A, can they similarly impair neural differentiation by knocking down DYRK1A with dCas9-KRAB? This would confirm the specificity of the ID8 activity.

We now provide data showing that this is infact the case, see response to the Editors, point 4.

b) Similarly, DYRK1A is a member of the dual-specificity tyrosine phosphorylation-regulated kinase family. It must have some known downstream targets that the authors can query by western blot, etc., to confirm the specificity of ID-8. If this was done in previous publications, it should be noted and cited.

This is a solid suggestion. One of the challenges of studying this particular protein kinase is that it interacts with a myriad of signaling pathways. Future work will address downstream targets in hPSC but this is beyond the scope of the current study.

4) Technology relevance: Does the addition of ID8 improve the efficacy/yield of non-neuronal (i.e. cardiac, muscle, endoderm) differentiation protocols? Are there any known activators of DYRK1A that could be used to improve neuronal differentiation protocols?

It is unlikely that ID-8 would enhance other differentiation protocols. As shown in Hasegawa et al. (8) the compound profoundly inhibits the induction of mesendodermal differentiation by WNT3A. The results in Figure 4 indicate that indeed DYRK1A activation could enhance neural differentiation, but we are unaware of specific compounds with this effect.

1) Chambers SM, Mica Y, Lee G, Studer L, Tomishima MJ. Dual-SMAD

Inhibition/WNT Activation-Based Methods to Induce Neural Crest and Derivatives from Human Pluripotent Stem Cells. Methods Mol Biol 2016;1307:329-43.

2) Denham M, Bye C, Leung J, Conley BJ, Thompson LH, Dottori M. Glycogen synthase kinase 3beta and activin/nodal inhibition in human embryonic stem cells induces a pre-neuroepithelial state that is required for specification to a floor plate cell lineage. Stem Cells 2012;30(11):2400-11.

3) Denham M, Hasegawa K, Menheniott T, Rollo B, Zhang D, Hough S, et al. Multipotent caudal neural progenitors derived from human pluripotent stem cells that give rise to lineages of the central and peripheral nervous system. Stem Cells 2015;33(6):1759-70.

4) Menendez L, Yatskievych TA, Antin PB, Dalton S. Wnt signaling and a Smad pathway blockade direct the differentiation of human pluripotent stem cells to multipotent neural crest cells. Proc Natl Acad Sci U S A 2011;108(48):19240-5.

5) Dang T, Duan WY, Yu B, Tong DL, Cheng C, Zhang YF, et al. Autismassociated Dyrk1a truncation mutants impair neuronal dendritic and spine growth and interfere with postnatal cortical development. Mol Psychiatry 2017.

6) Evers JM, Laskowski RA, Bertolli M, Clayton-Smith J, Deshpande C, Eason J, et al. Structural analysis of pathogenic mutations in the DYRK1A gene in patients with developmental disorders. Hum Mol Genet 2017;26(3):519-26.

7) van Bon BW, Coe BP, Bernier R, Green C, Gerdts J, Witherspoon K, et al. Disruptive de novo mutations of DYRK1A lead to a syndromic form of autism and ID. Mol Psychiatry 2016;21(1):126-32.

8) Hasegawa K, Yasuda SY, Teo JL, Nguyen C, McMillan M, Hsieh CL, et al. Wnt signaling orchestration with a small molecule DYRK inhibitor provides long-term xeno-free human pluripotent cell expansion. Stem Cells Transl Med 2012;1(1):1828.

9) van de Leemput J, Boles NC, Kiehl TR, Corneo B, Lederman P, Menon V, et al. CORTECON: a temporal transcriptome analysis of in vitro human cerebral cortex development from human embryonic stem cells. Neuron 2014;83(1):51-68.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

After additional discussion with the reviewers, we reached the conclusion that you should still carry out the experiment highlighted as reviewer 2's point #1 (repeating your main assay using 2 additional lines). We believe that this experiment is straightforward to do and will be important in convincing readers of the robustness of your observations.

Reviewer #2:

In their revised manuscript, Bellmaine et al. now better demonstrate that inhibition of DYRK1A prevents hESCs from differentiating to NPCs. The authors were reasonably responsive to my concerns, particularly in adding a genetic inhibition of DYRK1A to complement the pharmacological studies. Nonetheless, two simple experiments remain.

We now include new data showing that during dual SMAD induction of neural specification, DYRK1A inhibition by four of our active inhibitors reduces the level of PAX6 mRNA transcript induction in three of the four cell lines used in this study (WA09, GENEA022, and C11, an iPSC line used in new confirmatory experiments (below). The Wolvetang laboratory, which carried out this additional work, does not currently have access to the fourth cell line used in the study (HES-3 PAX6mCherry). These data are found in new Figure 4 figure supplement 1A, reported in the Results and Material and methods sections The C11 cell line is referenced (Briggs et al., 2013). We also present new DYRK1A shRNA knockdown data on additional cell lines, see response to Reviewer 2 comment 1 below.

1) Inter-hESC line variability – the authors did a great job clarifying which hESC lines were used in each figure panel. Nonetheless, I still think it would be helpful to compare all three hESC lines across at least a single assay (i.e. Pax6 qPCR or FACS), in order to assess how robust the effects are across hESC lines.

Please see Essential revisions above. We now also present DYRK1A shRNA knockdown data on two additional cell lines, Genea22 (previously used for the DYRK1A induction study) and another iPSC line, C11. These data confirm the previous shRNA data obtained with WA09. These data are included in new Figure 4—figure supplement 1B, C., reported in the last paragraph of the Results subsection “ID-8 offsets effects of DYRK1A overexpression on neural specification of hESC and DYRK1A knockdown phenocopies effects of inhibitors”, the shRNA methods are the same as those described for WA09, and the C11 cell line is referenced Briggs et al., 2013.

2) NPCs: The authors suggest that ID-8 treatment is preventing hESCs from specifying to an NPC fate. It would be interesting to collect the remaining Pax6-GFP positive cells following ID8 treatment (or shRNA knockdown) and assess either their gene expression and/or their capacity to replicate and undergo neuronal differentiation (either with or without ID8 present)? Specifically, I am curious whether the remaining NPCs generate both neurons and astrocytes, and if neurons, both excitatory and inhibitory.

We accept that this would be an interesting experiment to carry out, however, the isolation of these cells presents technical challenges as noted previously, and is beyond the scope of this study which specifically addresses the inhibition of neural specification.

https://doi.org/10.7554/eLife.24502.019

Article and author information

Author details

  1. Stephanie F Bellmaine

    1. School of Chemistry, University of Melbourne, Victoria, Australia
    2. Department of Anatomy and Neuroscience, University of Melbourne, Victoria, Australia
    3. Bio21 Molecular Science and Biotechnology Institute, University of Melbourne, Victoria, Australia
    Contribution
    Conceptualization, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  2. Dmitry A Ovchinnikov

    Australian Institute for Bioengineering and Nanotechnology, University of Queensland, Brisbane, Australia
    Contribution
    Conceptualization, Investigation, Visualization, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  3. David T Manallack

    Monash Institute of Pharmaceutical Sciences, Faculty of Pharmacy and Pharmaceutical Sciences, Monash University, Victoria, Australia
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Visualization, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Claire E Cuddy

    Department of Anatomy and Neuroscience, University of Melbourne, Victoria, Australia
    Contribution
    Formal analysis, Validation, Investigation
    Competing interests
    No competing interests declared
  5. Andrew G Elefanty

    1. Murdoch Childrens Research Institute, The Royal Children’s Hospital, Victoria, Australia
    2. Department of Anatomy and Neuroscience, Faculty of Medicine, Dentistry and Health Sciences, University of Melbourne, Victoria, Australia
    3. Department of Anatomy and Developmental Biology, Faculty of Medicine, Nursing and Health Sciences, Monash University, Victoria, Australia
    Contribution
    Resources, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  6. Edouard G Stanley

    1. Murdoch Childrens Research Institute, The Royal Children’s Hospital, Victoria, Australia
    2. Department of Anatomy and Neuroscience, Faculty of Medicine, Dentistry and Health Sciences, University of Melbourne, Victoria, Australia
    3. Department of Anatomy and Developmental Biology, Faculty of Medicine, Nursing and Health Sciences, Monash University, Victoria, Australia
    Contribution
    Resources, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  7. Ernst J Wolvetang

    Australian Institute for Bioengineering and Nanotechnology, University of Queensland, Brisbane, Australia
    Contribution
    Conceptualization, Resources, Methodology, Writing—review and editing
    Contributed equally with
    Spencer J Williams and Martin Pera
    Competing interests
    No competing interests declared
    ORCID icon 0000-0002-2146-6614
  8. Spencer J Williams

    1. School of Chemistry, University of Melbourne, Victoria, Australia
    2. Bio21 Molecular Science and Biotechnology Institute, University of Melbourne, Victoria, Australia
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Contributed equally with
    Ernst J Wolvetang and Martin Pera
    Competing interests
    No competing interests declared
  9. Martin Pera

    1. Department of Anatomy and Neuroscience, University of Melbourne, Victoria, Australia
    2. Walter and Eliza Hall Institute of Medical Research, Florey Neuroscience and Mental Health Institute, Victoria, Australia
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Writing—original draft, Project administration, Writing—review and editing
    Contributed equally with
    Ernst J Wolvetang and Spencer J Williams
    For correspondence
    mpera@unimelb.edu.au
    Competing interests
    Reviewing editor, eLife
    ORCID icon 0000-0001-6239-0428

Funding

Australian Research Council (Special Research Initiative SR1101002)

  • Dmitry A Ovchinnikov
  • Andrew G Elefanty
  • Edouard G Stanley
  • Ernst J Wolvetang
  • Martin Pera

National Health and Medical Research Council (Senior Research Fellowship)

  • Andrew G Elefanty
  • Edouard G Stanley

University of Melbourne (Strategic APA)

  • Stephanie F Bellmaine
  • Spencer J Williams
  • Martin Pera

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by the University of Melbourne, the Australian Research Council (SR1101002). Work in the laboratories of EGS and AGE was supported by grants from the Australian Stem Cell Centre, Stem Cells Australia, and the National Health and Medical Research Council (NHMRC) of Australia. AGE and EGS are Senior Research Fellows of the NHMRC.

Reviewing Editor

  1. Lee L Rubin, Reviewing Editor, Harvard Stem Cell Institute, Harvard University, United States

Publication history

  1. Received: December 22, 2016
  2. Accepted: September 1, 2017
  3. Accepted Manuscript published: September 8, 2017 (version 1)
  4. Version of Record published: October 25, 2017 (version 2)

Copyright

© 2017, Bellmaine et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 828
    Page views
  • 248
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Scopus, Crossref.

Comments

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cancer Biology
    2. Neuroscience
    Peter M LoCoco et al.
    Research Article Updated