FK506‐binding protein 5 regulates cell quiescence‐proliferation decision in zebrafish epithelium

Using a zebrafish ionocyte model, transcriptomics and genetic analyses were performed to identify pathways and genes involved in cell quiescence‐proliferation regulation. Gene ontology and Kyoto encyclopedia of genes and genomes pathway analyses revealed that genes involved in transcription regulation, cell cycle, Foxo signalling and Wnt signalling pathway are enriched among the up‐regulated genes while those involved in ion transport, cell adhesion and oxidation–reduction are enriched among the down‐regulated genes. Among the top up‐regulated genes is FK506‐binding protein 5 (Fkbp5). Genetic deletion and pharmacological inhibition of Fkbp5 abolished ionocyte reactivation and impaired Akt signalling. Forced expression of a constitutively active form of Akt rescued the defects caused by Fkbp5 inhibition. These results uncover a key role of Fbkp5 in regulating the quiescence‐proliferation decision via Akt signalling.

The active cell cycle consists of four phases: G1, S, G2 and M. In addition, the G0 phase or quiescence phase exists outside of the proliferative cycle. Cells in the quiescence phase, while non-dividing and dormant, can re-enter the cell cycle upon injuries or receiving appropriate signals and stimuli [1]. Quiescence prevents long-lived cells (such as adult stem cells) from accumulation of genomic aberrations and stress. Dysregulation of the proliferationquiescence balance can lead to hyper-and hypoproliferative human diseases, including cancer and fibrosis [1].
The molecular mechanisms underlying the proliferationquiescence decision have been studied extensively using cultured cells [1]. Based on data collected from cultured mammalian cells kept in mitogen-free media for different time periods, Pardee [2] proposed that a 'restriction' point exists in late G1 phase [2]. Cells that have crossed this check point prior to the mitogen removal usually complete the next cell cycle. In contrast, cells that have not reached this check point at the time of mitogen removal become quiescent and enter the G0 phase. Yao et al. [3] showed that Rb and E2F play a critical role in regulating the proliferation-quiescence decision in cultured mammalian cells [3]. Subsequent studies have led to the proposal of a bifurcation mechanism controlled by CDK2 and its inhibitor p21 [4]. Cells that cross this bifurcation point at the end of mitosis with high CDK2 activity enter the next active cell cycle. Those with low CDK2 activity enter the quiescence phase [4]. These in vitro findings have been integrated into a two-step model recently [1]. According to this model, the proliferation-quiescence decision is regulated by two bistable switches. The Rb-E2F acts as the first switch. This bistable switch integrates extracellular mitogenic signals via Cyclin D:CDK4/6 activity. The Rb-E2F switch activity is also influenced by DNA damage and metabolic signals. A second bistable switch is regulated by CDK2 activity and p21 degradation. The time between these two points represents a window for cell cycle reversibility [1]. While these findings have provided important insights into how cells exit the active cell cycle and enter the quiescence, key questions remain. To date, the molecular nature and precise position of the commitment point in the cell cycle are unknown [1]. We have a limited understanding of how quiescent cells are reactivated and re-enter the active cell cycle. Moreover, these models are developed primarily based on cell culture systems, mostly in immortalised cell lines. We now understand that there are tremendous complexities in vivo, ranging from crosstalk with various hormones and growth factors, the availability of local nutrients and oxygen, to various feedback loops. These complexities cannot be captured by cell culturebased assays.
We have recently discovered that a population of epithelial cells in zebrafish, known as Na + -K + -ATPase-rich ionocytes or NaR cells, are reactivated and undergo robust cell division in response to low calcium stress or genetic manipulations [5][6][7][8][9]. NaR cells are structurally and functionally similar to human intestinal and renal epithelial cells and contain all molecular components for transcellular Ca 2+ transport, including the epithelial Ca 2+ channel Trpv6. Zebrafish live in hypoosmotic aquatic habitats and use NaR cells to take up Ca 2+ from the surrounding water to maintain their body calcium homeostasis [10]. Under normal medium, the constitutively open Trpv6 mediates Ca 2+ influx continuously and maintains a high level of cytoplasmic free Ca 2+ ([Ca 2+ ] c ) in these cells. The high [Ca 2+ ] c suppresses Akt-Tor signalling via the conserved protein phosphatase PP2A and promotes cell quiescence [9]. When fish are transferred to an embryo medium containing low [Ca 2+ ] or when Trpv6 is genetically deleted, the [Ca 2+ ] c drops and NaR cells reenter the cell cycle due to the re-activation of Igfbp5a-Igf1r-PI3 kinase-Akt-Tor signalling pathway [5][6][7][8][9]11]. These findings are consistent with the known role of the AKT-mTOR signalling in reactivating adult stem cells and T cells in mammals and Drosophila [12][13][14], suggesting this is an evolutionarily conserved mechanism(s).
To identify pathways and molecules regulating the quiescence-proliferation transition in vivo, we engineered a stable zebrafish transgenic line Tg(igfbp5a:GFP) in which NaR cells are genetically labelled with GFP. These transgenic fish faithfully report NaR cell reactivation [7][8][9]. Zebrafish are oviparous vertebrates. Their freeliving, tiny and transparent embryos make it a popular model organism in developmental biology. NaR cells in particular are located on the surface of yolk sac skin larval fish, making Tg(igfbp5a:GFP) fish well suited for in vivo analysis of the quiescence-proliferation regulation. RNA-seq analysis results showed that genes involved in cell cycle, gene transcription, Foxo signalling etc. are upregulated in reactivated NaR cells while genes involved in ion transport, cell adhesion and oxidation-reduction are down-regulated. Further data mining and genetic experiments uncovered Fkbp5 as a key regulator of NaR cell quiescence-proliferation decision in vivo.

Results
The transcriptomic profile of reactivated NaR cells The RNA-seq experiment design is shown in Fig. 1A. NaR cells were isolated by fluorescence-activated cell  sorting (FACS) sorting from Tg(igfbp5a:GFP) fish treated with the induction medium or control medium. RNA was isolated from these cells and subjected to RNA-seq analysis. As reported previously [7], NaR cells in the induction medium-treated fish underwent proliferation and formed clusters of newly divided cells in the yolk sac region while those in the control medium remained undivided (Fig. 1B). A total of 27 107 genes were detected. The genes' mapping ratios were 83.5%~88.1% (Fig. S1A,B). Differential gene expression (DEG) analysis detected a total of 2164 DEGs. Among them, 1168 genes were up-regulated and 996 were down-regulated in reactivated NaR cells ( Fig. 1C; Table S1). Hierarchical clustering of DEGs showed that the reactivated NaR cell datasets clustered distinctly from the control NaR cell dataset (Fig. 1D).
In agreement, principal component analysis showed that the reactivated NaR cell datasets clustered distinctly from those of the quiescent NaR cell dataset (Fig. S1C,D). To confirm the RNA-seq results, qRT-PCR assays were performed using a different set of FACS-sorted NaR cells and similar changes were detected in 17 out of the 17 genes examined ( Fig. 1E; Fig. S2).

Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis
The identified DEGs were annotated with GO terms and sorted based on functional categories. Genes involved in transcription regulation, oxidation reduction, multicellular organism development, DNA repair and cell proliferation are enriched (Fig. 1E). When the up-regulated and down-regulated DEGs were analysed separately, genes involved in transcription/regulation, multicellular organism development, cell cycle/cell division, DNA repair and Wnt signalling are enriched in the up-regulated DEGs (Fig. 1F). Among the downregulated genes, the most enriched GO term is transport, followed by oxidation reduction, cell adhesion and vesicle-mediated transport (Fig. 1G). KEGG pathway enrichment analysis was performed next. The DEGs were significantly enriched in cell cycle, Foxo signalling, p53 signalling, PPAR signalling and Fanconi anaemia pathway (Fig. 1H). These pathways were also enriched in the up-regulated DEGs (Fig. 1I). Additionally, Wnt signalling, cellular senescence and melanogenesis genes were also enriched (Fig. 1I). The enrichment of up-regulated 'cell cycle' genes is in good agreement with the fact that these are rapidly dividing cells. Although KEGG analysis suggested an enrichment of the p53 signalling and cell senescence pathways (Fig. 1H,I), many of the genes are overlapping with those in the cell-cycle pathway (Fig. S3). Genes in this category include cdk21, gad-d45aa, ccne2, cdkn1a,ccnb2, ccnd1, chek1 and e2f2 (Fig. S3). There are also significant overlaps in the Wnt and melanogenesis pathways (Fig. S5), including wnt7ba, wnt8b, wnt 11, wnt6b, wnt7bb and wnt7aa, consistent with the role of Wnt signalling in regulating melanocyte development [15].

Fkbp5 is up-regulated and is indispensable in NaR cell reactivation
We mined the RNA-seq dataset further by ranking the up-regulated DEGs based on their relative mRNA abundance in NaR cells ( Table 1). As expected, the calcium channel gene trpv6, which has been shown to regulate NaR cell proliferation-quiescence decision by suppressing the Igf1r-PI3 kinase-Akt-Tor signalling pathway [9], is among this group. Others are col2a1a, caspb, her9, ankhb, pck1, fkbp5, si:dkey-22i16.3, ponzr3 and wu:fj16a03 (Table 1). We were particularly intrigued by Fkbp5, a member of the conserved FK506 binding protein (FKBP) family. The FKBP family consists of a large number of structurally related proteins known for their capability of binding to and modulating the immunosuppressive actions of FK506 as well as rapamycin, a well-known mTOR inhibitor [16]. Rapamycin treatment inhibited NaR cell reactivation and proliferation [7]. Among the 19 zebrafish fkbp gene family members ( Fig. 2A), the expression of fkbp5 and fkbp-like, but not other fkbp genes, was increased in reactivated NaR cells (Fig. 2B). The up-regulation of fkbp5 mRNA expression was confirmed by qRT-PCR (Fig. 2C). In comparison, the mRNA levels of fkbp7, a related member of the Fkbp gene family, showed no change (Fig. 2C).
To determine whether Fkbp5 plays a role in reactivating NaR cells, Tg(igfbp5a:GFP) fish were treated with FK506, a pan FKBP inhibitor [17]. FK506 treatment inhibited the induction medium-induced NaR cell proliferation, whereas it did not change the basal NaR cell number in fish kept in the control medium ( Fig. 3A,B). Next, a CRISPR-Cas9-based F0 gene deletion approach was used to genetically delete Fkbp5. This method can convert over 90% of injected embryos directly into biallelic knockout [18,19] and is well suited for functional analysis of essential genes. Injection of fkbp5 targeting gRNAs abolished NaR cell reactivation while scrambled gRNAs had not such an effect (Fig. 3C). Deletion of Fkbp5 did not alter the number of NaR cells under the control medium (Fig. 3C). In comparison, CRISPR/Cas9-mediated knockdown of Fkbp7 had little effect (Fig. 3D). Together, these data suggest that elevated Fkbp5 expression promotes NaR cell quiescence-proliferation transition.

Fkbp5 promotes NaR cell proliferation via Akt signalling
If Fkbp5 acts in the Igf1r-PI3 kinase-Akt pathway to stimulate NaR cell reactivation and proliferation, then deletion of Fkbp5 should alter Akt signalling activity in these cells. Likewise, constitutive activation of Akt should reverse the effects of Fkbp5 inhibition. Indeed, while the induction medium treatment resulted in a robust increase in the number of phosphor-Akt positive cells in un-injected or scrambled gRNA injected fish (Fig. 4A), this effect was impaired by CRISPR/ Cas9-mediated fkbp5 deletion (Fig. 4B). Next, myrAkt, a constitutively active form of Akt [20], was expressed in a subset of NaR cells using a Tol2 transposonmediated genetic mosaic assay [21]. While FK506 treatment impaired NaR cell reactivation and proliferation in normal NaR cells, this effect was reversed by myrAkt expressing (Fig. 4C). These data suggest that Fkbp5 promotes NaR cell reactivation and proliferation via Akt signalling.

Discussion
In this study, we used a zebrafish model and mapped transcriptomic changes during epithelial cell quiescence to proliferation transition in vivo. Our transcriptomic analysis showed that genes involved in transcription/ regulation, cell cycle, Foxo, Wnt and PPAR signalling are increased while those involved in ion transport, cell adhesion and oxidation-reduction are downregulated in the reactivated cells. These findings support the notion that low calcium stress induces differentiated NaR cells to re-enter the active cell cycle via reactivating the PI3 kinase-Akt pathway [8,9,11]. Further data mining, genetic and biochemical analyses showed that Fkbp5 plays a critical role in stimulating NaR cell proliferation-quiescence transition via Akt signalling. As expected, a significant enrichment of the cell cycle genes was observed in reactivated and dividing NaR cells. Although p53 signalling and cellular senescence pathway are also enriched, a close look suggested that many of the enriched genes are overlapping with those in the cell cycle genes (Fig. S3).  or BAC (igfbp5a:myr-Akt-mCherry) were raised. At 3 dpf, they were transferred to the induction medium containing DMSO or 10 mM FK506. At 5 dpf, NaR cells labelled by both GFP and mCherry were scored following a previously established scoring system [21]. NaR cells that that did not divide, divided one time or two times are scored as À, + and ++, respectively. *P < 0.05; ****P < 0.0001; ns, not significant. Total cell number is shown above each column. GADD45 and CDK1 have been implicated in the modulation of cell cycle and apoptosis in mammalian cells [22]. Likewise, ccne2, ccnd1 and ccnd2, which encode Cyclin E2, D1 and B2, are key regulators of cell cycle and frequently overexpressed in cancer cells [23][24][25]. In zebrafish, Cdk21 has been shown to promote male germ cell proliferation and meiosis [26]. In addition to cell cycle genes, several DNA repair pathway genes are enriched in the reactivated and proliferation NaR cells. Cell division requires the faithful copying of the genome once per cell cycle. Robust cell replication can lead to replication stress and activates DNA damage repair pathways such as homologous recombination [27]. It is possible that the rapid dividing NaR cells may experience replication stress. This is consistent with the enrichment of Fanconi anaemia genes in these cells. The Fanconi anaemia pathway functions as a DNA damage repair system [28].
In previous studies, we have shown that the reactivation of Akt is required and sufficient in promoting NaR cell reactivation and proliferation [5,[7][8][9]. In this study, many genes in the Foxo signalling pathway are enriched (Fig. S4). A key mechanism in the regulation of FOXO proteins is the phosphorylation by AKT, in response to insulin or IGFs stimulation. This modification promotes FOXO proteins to exit from the nucleus and decreases the expression of the FOXO target genes [29]. Among the DEGs in this pathway are Sgk1 and Sgk3, two Akt-like protein kinases. SGK1 has been reported to be activated by insulin and IGFs via the PDK1 and TORC2 [30,31]. SGK1 increases the protein abundance and/or activity of several ion channels, solute carriers and Na + /K + -ATPases [32]. Overexpression of SGK3 in hepatocellular carcinoma has been shown to increase cell cycle progression through G1 by inactivating glycogen synthase kinase-b and stabilising CCND1 [33]. Future genetic studies are needed to elucidate the possible roles of Sgk1 and/or Sgk3 in NaR cell quiescence-proliferation regulation.
Multiple Wnt ligands are enriched in the up-regulated genes in proliferating NaR cells. Wnt signalling plays a crucial role in melanocyte development [15]. In Xenopus epidermis, Wnt signalling is required for cilia formation and differentiation of multiciliate epithelial cells. In basal cells, Wnt signalling prevents specification of epithelial cell types via DN-TP63, a master transcription factor [34]. NaR cells are located on the yolk sac epidermis in zebrafish embryos and larvae. It will be interesting to determine whether these Wnt ligand genes play a direct role in regulating NaR cell development and proliferation. Another pathway of interest is PPAR signalling. Several genes, including pck1 and pck2, are enriched in the reactivated NaR cells. While PCK1 is well known as a gluconeogenic enzyme, recent studies show that PCK1 and its mitochondrial counterpart PCK2 also increase TCA cycle flux to stimulate cell proliferation in colon cancer-derived cell lines [35][36][37]. It is unclear whether any of these PPAR genes plays a role in regulating NaR cell fate.
NaR cells are one of the several types of ionocytes derived from epidermal stem cells [9,10]. NaR cells are polarised cells with an apical opening facing the external environments and basolateral sides facing the internal body fluids. These cells are responsible for transcellular Ca 2+ transport, which also involves Na + and K + trafficking [10]. Our transcriptomic analysis showed that the GO term of 'transport' was most highly enriched among the down-regulated genes. Out of the 996 down-regulated genes, 57 are in this group. Within this category, 11 genes are known to be involved in ion transport. These include several calcium ion channels, potassium channels and Na + , Cl À and H + channels and transporters. The downregulation of these genes, together with the reduction of cell adhesion genes, supports the notation that NaR cells undergo dedifferentiation and become reactivated. We have previously found that the epithelial calcium channel trpv6 plays a critical role in stimulating NaR cell quiescence. This study identified three additional calcium channels, including cacng1a (calcium channel, voltage-dependent, gamma subunit 1a), cacng3a and trpv1 down-regulated in reactivated NaR cells. Trpv1 is critical for heat sensing [38]. Currently, little is known about the function of cacng1a and cacng3a. The roles of these channels in NaR cell quiescenceproliferation balance, if any, are unknown at present and deserves further investigation.
A key finding made in this study is that Fkbp5 played a critical role in promoting NaR cell quiescenceproliferation transition. This conclusion is supported by several lines of evidence. First, fkbp5 mRNA expression is up-regulated in reactivated NaR cells. In comparison, the expression of fkbp7 and many other fkbp genes was not changed. Second, pharmacological inhibition of Fkbp prevented NaR cell quiescence-proliferation transition. Furthermore, genetic deletion of Fkbp5 impaired NaR cell reactivation and proliferation. This effect is specific because deletion of Fkbp7 had little effect. Although a previous study reported that FKBP51 (encoded by the human FKBP5 gene) promotes the expression of epithelial-to-mesenchymal transition genes [39], a role of Fkbp5 in cell quiescenceproliferation decision has not been reported. FKBPs have a variety of molecular masses that are involved in several signal transduction pathways [16,40,41]. Human FKBP51 was originally discovered as a co-chaperone of Hsp90 and a component of the Hsp90-steroid receptor heterocomplex [42]. FKBP51 contains 2 FKBP domains and 3 tetratricopeptide-repeats domain. While the first FKBP domain (FK1) has peptidylprolyl isomerase activity and provides a binding pocket for FK506 and rapamycin, FK2 lacks the PPIase activity. Among the FKBP family members, FKBP51 is most closely related to FKBP52 (encoded by the human FKBP-4 gene) [43]. Despite the high sequence similarity, these two proteins had different and often opposite roles in several biological processes. While one was found to facilitate the dephosphorylation of Akt via the phosphatase PHLPP [44,45], the other had a capability to promote Akt phosphorylation through PI3K/PDK1 and mTORC2 [46]. In this study, we provided genetic evidence that Fkbp5 acts upstream of Akt to promote NaR cell reactivation and proliferation. This conclusion is supported by the fact that CRISPR/Cas9mediated deletion and pharmacological inhibition of Fkbp5 attenuated Akt signalling in NaR cells while targeted expression of myrAkt 'rescued' FK506 effect on NaR cell reactivation. As a central node for signal transduction and proliferation, Akt integrated multiple intracellular signalling including Ca 2+ flux. Intracellular Ca 2+ /calmodulin signalling has been implicated in Akt activation in cultured mammalian cells [47]. Our recent study showed that ER Ca 2+ efflux activated Ca 2+ /calmodulindependent protein kinase kinase (CaM-KK), and this in turn activated Akt to induce NaR cell reactivation [11]. Human FKBP 12 and FKBP12.6 are both known to regulate ER Ca 2+ flux via the association with ER Ca 2+ channels like ryanodine receptors (RyRs) and inositol 1,4,5-trisphosphate receptors (IP3Rs) [48,49]. Our transcriptomic analysis showed the mRNA levels of fkbp1aa, fkbp1ab and fkbp1b, which encode zebrafish Fkbp12 and Fkbp12.6a and -b, were not significantly changed in reactivated NaR cells. Treatment of zebrafish larvae with RyRs inhibitors ryanodine and dantrolene blocked NaR cell reactivation while the IP3R inhibitor xestospongin C did not have such an effect [11]. Future studies are needed to determine whether Fkbp5 regulates Akt signalling by altering RyRs-mediated ER Ca 2+ flux in NaR cells.

Chemicals and reagents
All chemical reagents were purchased from Fisher Scientific (Pittsburgh, PA, USA) unless stated otherwise. Liberase TM was purchased from Sigma-Aldrich (St Louis, MO, USA).

Zebrafish
Zebrafish were raised following standard zebrafish husbandry guidelines [50]. Embryos were obtained by natural cross and staged following Kimmel et al. [51]. All embryos were raised in the standard E3 embryo medium [50] until 3 dpf. The control embryo medium (containing 0.2 mM [Ca 2+ ]) and the induction embryo medium (containing 0.001 mM [Ca 2+ ]) were prepared following a previously reported formula [5]. In some experiments, 0.003% (w/v) N-phenylthiourea was added to the media to inhibit pigmentation. The Tg(igfbp5a:GFP) fish line was generated in a previous study [7]. All experiments were conducted in accordance with the guidelines approved by the University of Michigan Institutional Committee on the Use and Care of Animals.

FACS, RNA-seq and quantitative real-time PCR (qPCR) analysis
Isolation and sorting of NaR cells by FACS were performed as previously described [7]. For RNA-seq, RNA was isolated from FACS-sorted NaR cells using a RNeasy Micro Kit (Qiagen, Valencia, CA, USA). The integrity of the RNA samples was confirmed using the Agilent Bioanalyzer 2100 with an RNA 6000 Pco Kit (Agilent, Santa Clara, CA, USA). All samples had an RNA Integrity Number of 8 or greater. cDNA libraries were prepared using a SMART-Seq v4 Ultra Low Input RNA Kit (Singulomics, Palo Alto, CA, USA). Briefly, the first-strand cDNA was synthesised, full-length ds cDNA was amplified by LD-PCR. The amplified cDNA was purified using an Agencourt AMPure XP Kit (Beckman Coulter Life Sciences, Indianapolis, IN, USA) and validated using the Agilent 2100 Bioanalyzer (Agilent). An Illumina Library was prepared. Single-end, 50 bp sequencing was performed on the Illumina Hi-Seq 4000 platform. The quality control of raw reads was tested using FastQC. The raw reads were trimmed using the Trimmomatic [52] and filtered. The preprocessed reads were mapped to the genome sequence of zebrafish GRCz10 (GCA_000002035.3) using TopHat and the aligned reads were assembled into transcripts using Cufflinks. The assembled transcripts were merged with the reference annotation (Danio_rerio.Zv9.68) using cuffmerge. The SVA package in R was used for removing batch effects. The htseq-count function was used to detect the number of counts in BAM files. Differential expression analysis was performed using WILCOX.TEST in R package. Heatmap of hierarchical clustering was created using the PHEATMAP R package. The CLUSTERPROFILTER package in R was used to perform GO and pathway analysis. To validate the RNAseq results, NaR cells were FACS sorted and RNA isolation was carried out. cDNA was synthesised using Superscript III Reverse Transcriptase. qPCR was performed on an ABI 7500 fast Real-Time PCR system (Applied Biosystems, Foster City, CA, USA) using SYBR Green (Bio-Rad, Hercules, CA, USA). PCR primers are listed in Table S2.

Phylogenetic analysis of the Fkbp protein family
Phylogenetic tree of the FKBP family in humans (hs), zebrafish (dr) and Drosophila (dm) was constructed using the Maximum Likelihood method based on the JTT matrix-based model [53]. The tree was constructed using MEGA7 [54].

Drug treatment
Drugs used in this study were dissolved in DMSO and further diluted in water. Drug treatments began at 3 dpf. The fish were subjected to immunostaining after 1 day treatment or NaR cell quantification after 2 days of treatment.

CRISPR/Cas9-mediated F0 gene deletion
Four sgRNAs were used to perform transient deletion of fkbp5 and fkbp7 following [19]. The sgRNAs were synthesised by in vitro transcription based on a published method [55]. Cas9 mRNA was synthesised by in vitro transcription using the pT3.Cas9-UTRglobin plasmid as template. sgRNAs (40 ngÁlL À1 ) were mixed with Cas9 mRNA (400 ngÁlL À1 ) and co-injected into Tg(igfbp5a:GFP) embryos at one cell stage. The injected embryos were raised in E3 embryo medium. To induce NaR cell reactivation, fish were transferred to the control and induction embryo medium at 3 dpf as previously reported [7]. Knockout efficacy was validated using headloop PCR following ( [18,19]).

Tol2 transposon-mediated transgenesis assay
A Tol2 transposon-mediated genetic mosaic assay was used to target the expression of a transgene in a subset of NaR cells [21]. BAC(igfbp5a:myrAkt-mCherry) DNA and Tol2 mRNA were mixed and injected into 1-cell stage Tg (igfbp5a:GFP) embryos. The embryos were raised and subjected to various treatments. Cells co-expressing mCherry and GFP were identified, and the cell proliferation index was determined as previously reported [21].

Whole-mount immunostaining
Zebrafish larvae were fixed in 4% paraformaldehyde, permeabilised in methanol and subjected to whole mount immunostaining of phosphor-Akt as described previously [5].

Statistical analysis
Statistical analyses were performed in consultation with the University of Michigan's Consulting for Statistics, Computing and Analytics Research (CSCAR) team. Cell proliferation index data were analysed using pairwise Chi-square tests. Data are shown as mean AE standard deviation (SD). Statistical analyses between two groups were performed using unpaired two-tailed t test. When comparing the means of multiple groups to a control group, Dunnett's test was used to correct for multiple comparisons after one-way ANOVA. Statistical significances from all tests were accepted at P < 0.05 or higher.

Funding
This work was supported by NSF IOS-1755268 to CD. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author contributions
CD, CL, and YL performed the conceptualization. YL, CL, and XB involved in investigation. YL and CL performed the visualization. CD and ML involved in supervision. CD, YL, and CL wrote the original draft. YL, CL, XB, M, and CD involved in review and editing.

Data availability
The data that support the findings of this study are openly available in Table S1 and Data S1-S4.  were used for analysis. Fig. S2. Changes in gene expression. NaR cells were isolated by FACS sorting as described in Fig. 1A. The mRNA levels of the indicated genes were measured by qPCR (blue) or RNA-seq (green) and shown as the ratio between the induction groups and the control groups.   S4. Heatmap of enriched genes in the Foxo signalling pathway in the four control (C1-C4) and four induction (R1-R4) groups of NaR cells. Fig. S5. Heatmaps of enriched genes in Wnt signalling (A) and melanogenesis (B) pathway in the four control (C1-C4) and four induction (R1-R4) groups of NaR cells. Note the significant overlap in EDGs in these 2 categories. Fig. S6. Heatmaps of enriched genes in the PAAR signalling (A) and Fanconi anaemia pathway (B) in the four control (C1-C4) and four induction (R1-R4) groups of NaR cells. Table S1. Significant regulated genes. Table S2. Genes and primers used in qPCR. Data S1. Supplemental data - Figure 1E.