ESRG maintains human pluripotent stem cell self-renewal by binding NPM1 to regulate BMP4 protein and TGF-β signaling pathway

Abstract

Human pluripotent stem cells (hPSCs) present considerable potential for regenerative medicine; however, the standardization and large-scale production of these cells are hindered by an incomplete understanding of the molecular mechanisms governing self-renewal. The long non-coding RNA ESRG is integral to maintaining hPSC self-renewal, with its depletion leading to reduced levels of nucleophosmin 1 (NPM1) protein, thereby compromising the self-renewal capacity of hPSCs. Mechanistically, ESRG physically interacts with NPM1, and a reduction in ESRG/NPM1 expression results in elevated bone morphogenetic protein 4 (BMP4) levels and decreased polypyrimidine tract-binding protein 1 (PTBP1) levels, which in turn destabilize TGF-β1 mRNA and attenuate TGF-β signaling activity. Notably, treatment with the TGF-β agonist SRI-011381 partially rescues the self-renewal defects caused by ESRG or NPM1 knockdown. Collectively, our findings elucidate that the ESRG–NPM1–BMP4–PTBP1 axis governs hPSC self-renewal by post-transcriptionally regulating TGF-β1 mRNA stability and sustaining TGF-β signaling. This study enhances the understanding of the molecular regulatory network underlying hPSC self-renewal maintenance and validates ESRG as a promising target for improving the in vitro expansion of hPSCs.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13578-026-01580-5.

Background

Human pluripotent stem cells (hPSCs) possess robust self-renewal capacity, endowing them with broad applicability in regenerative medicine and personalized therapy [1] [2]. The self-renewal of hPSCs is regulated by an intricate network of transcriptional control mechanisms and epigenetic modifications, among other regulatory factors. Within this framework, lncRNAs play a critical role in maintaining self-renewal and determining cell fate through mechanisms including the modulation of protein expression levels and the mediation of key signal transduction pathways [3–5]. Nonetheless, the primary regulatory lncRNA that stabilizes long-term self-renewal and facilitates in vitro propagation has yet to be identified, and the corresponding theoretical framework remains incomplete.

ESRG, a lncRNA originally identified and characterized by our group, is predominantly nuclear-localized and plays a critical role in sustaining the self-renewal and pluripotency of hPSCs [6, 7]; its expression level gradually decreases during hPSC differentiation. In previous work, we demonstrated that ESRG interacts with minichromosome maintenance protein 2 (MCM2), facilitating its nuclear retention and suppressing both differentiation and apoptosis in hPSCs [6]. Additionally, ESRG binds heterogeneous nuclear ribonucleoprotein A1 (HNRNPA1) to regulate transcription factor 3 (TCF3) alternative splicing, thereby supporting hPSC self-renewal [7]. Nonetheless, overexpressing these downstream effectors only partially rescued the phenotypic defects induced by ESRG knockdown. To elucidate the mechanistic underpinnings of ESRG function, we conducted RNA pull-down assays coupled with mass spectrometry analysis. The findings indicated that, in addition to MCM2 and HNRNPA1, ESRG may also interact with NPM1.

NPM1 is a multifunctional nucleolar protein integral to essential cellular processes such as centrosome duplication, ribosome biogenesis, and cell cycle regulation [8]. Mutations or deficiencies in NPM1 are closely linked to dysregulated DNA repair, cell cycle arrest, and impaired differentiation [9]. These phenotypic outcomes are relevant to the pathogenesis of stem cell dysfunction-associated diseases. For example, targeted inhibition of NPM1 suppresses the proliferation of hepatic progenitor cells and induces apoptosis, exerting an anti-tumor effect [10]. A mutation in the C-terminal domain of NPM1 alters leukemia-associated transcriptional programs, leading to the malignant transformation of hematopoietic stem cells (HSCs) and onset of acute myeloid leukemia [11]. Furthermore, NPM1 knockdown induces the premature aging of HSCs, supporting its function in regulating HSC senescence and inflammation [12]. NPM1 has also been shown to regulate differentiation processes in hPSCs; for example, it suppresses endogenous retrovirus activity, thereby influencing germ layer lineage specification in embryonic stem cells (ESCs) [13]. However, the molecular mechanisms by which NPM1 maintains self-renewal in hPSCs remain poorly understood.

Bone Morphogenetic Protein 4 (BMP4), a member of the TGF-β superfamily, is an essential growth factor involved in diverse biological processes, including vascular development, angiogenesis, and osteogenesis [14]. Empirical studies have demonstrated that BMP4 directs hPSC differentiation toward diverse germ layer lineages [15, 16] and facilitates the formation of distinct tissue structures [17]. This property establishes BMP4 as a key regulator in the directed differentiation and organoid formation systems of hPSCs. In practice, specific cell types are induced by applying defined concentrations of recombinant BMP4 in the culture medium.

The TGF-β superfamily signaling pathway relies on the antagonistic dynamic equilibrium between its two principal branches: the Activin/Nodal-Smad2/3 pathway and the BMP-Smad1/5/8 pathway [18]. This balance is important for the precise regulation of self-renewal and differentiation in hPSCs. The Activin/Nodal-Smad2/3 signaling pathway helps maintain the undifferentiated state of hPSCs by suppressing neuroectodermal differentiation [19], which supports the long‑term stable culture of these cells. Accordingly, sufficient levels of Activin A or TGF-β are required to maintain stemness in chemically defined, feeder‑free culture systems [20, 21]. In contrast, without exogenous inhibition, the BMP-Smad1/5/8 signaling pathway generally directs hPSCs towards trophoblast or endodermal lineages [22, 23]. Thus, the naive pluripotency characteristics of hPSCs are generally linked to a steady state with high Activin/Nodal and FGF signaling, and low or inhibited BMP signaling. However, recent studies show that BMP4 interacts with TGF‑β signaling in both synergistic and antagonistic ways. For example, upon binding to its receptor, BMP4 activates the TGF-β signaling cascade, including both Smad-dependent and Smad-independent branches, and promotes in vitro differentiation of adult stem cells (ADSCs) [24]. Conversely, BMP4 can also counteract TGF-β1-induced myofibroblast differentiation [25]. Therefore, precise modulation of the signaling balance within the TGF-β superfamily is a key strategy for the long-term stable in vitro culture of hPSCs, and understanding how BMP4 and TGF-β signaling regulate hPSC self-renewal provides a useful framework for advancing basic research on hPSCs.

This study aims to investigate the molecular mechanism by which lncRNA ESRG regulates hPSC self-renewal through its interaction with the NPM1 protein. Our results show that ESRG binds directly to NPM1 and increases its protein stability. This in turn regulates the protein levels of both BMP4 and its interacting partner PTBP1, modulates TGF‑β1 mRNA stability to influence TGF‑β signaling, and thus maintains the self‑renewal capacity of hPSCs.

Materials and methods

Cell culture

Undifferentiated H9 and H1 human embryonic stem cells (hESCs) and induced pluripotent stem cells (iPSCs) (Nuwacell, RC01001) were used at passages 20–30. Cells were cultured on six-well plates coated with primary mouse embryonic fibroblasts (MEFs, passages 3–6), Matrigel (Corning, 354230), or Laminin‑521 (Thermo Fisher Scientific, A29249). Three culture media were employed: one consisting of DMEM/F12 (Gibco, 11330032) supplemented with 20% KnockOut™ Serum Replacement (Gibco, 10828028), 1% MEM non-essential amino acids (Gibco, 11140050), 1% GlutaMAX™-I (Gibco, A1286001), 4 ng/mL basic fibroblast growth factor (PeproTech, AF-100-18B-50UG), and 0.1 mM β‑mercaptoethanol (Sigma, M3148); the other were commercially available hPSC medium (Shownin, RP01001; Gibco, A2656101). 10 μM Y-27632 (Selleck, S1049) was added only on the first day, and the medium was changed daily. When colonies reached appropriate size, cells were passaged using 1 mL hPSC Dissociation buffer (Nuwacell, RP01007) for 3–5 min followed by centrifugation. For all experiments, cells were pre-adapted to culture on Matrigel (Corning, 354230) or Laminin‑521 (Thermo Fisher Scientific, A29249) and serially passaged for at least 3 generations prior to use.

Cell transfection

The transfection of siRNA (Supplementary Table 1) was conducted using Lipofectamine™ RNAiMAX (Thermo Fisher Scientific, 13778150). Expression plasmids pcDNA3.1(+)-NPM1 and pcDNA3.1(+)-C-DYK-PTBP1, both purchased from GenScript and verified by Sanger sequencing, were transfected using FuGENE® HD Transfection Reagent (Promega, E2311). For each well of a six-well plate, transfection complexes were prepared in 200 μL Opti‑MEM™ (Gibco, 31985070): either 5 μL RNAiMAX mixed with 5 μL siRNA, or 5 μL FuGENE mixed with 3 μg expression plasmid. After mixing and incubation at room temperature for 20 min, the mixture was added to the culture system. Cells were collected for RNA extraction at 48 h post-transfection, and for protein extraction at 72 h post-transfection.

Western blot analysis

Cells were lysed in 100 μL of protein lysis buffer (Beyotime, P0013) supplemented with 1% protease inhibitor cocktail (MCE, HY-K0010). After centrifugation at 12,000 rpm for 15–20 min at 4°C, the protein supernatant was collected, mixed with SDS-PAGE sample buffer (Beyotime, P0286), and boiled for 5–10 min. Proteins were separated by SDS-PAGE and transferred onto a PVDF membrane (Millipore, IPVH00010). Membranes were blocked with 5% non-fat milk for 2 h at room temperature, then incubated with primary antibodies (Supplementary Table 2) diluted at 1:1000 at 4°C overnight. After washing with TBST three times for 5 min each, membranes were incubated with HRP-conjugated secondary antibodies (Supplementary Table 2) diluted at 1:3000 for 1 h at 37°C. Protein bands were visualized using enhanced chemiluminescence (Millipore, WBULS0100) and detected with a chemiluminescence imaging system.

Alkaline phosphatase (ALP) staining

Cells were fixed with in situ hybridization fixative (Servicebio, G1113-500ML) for 20–30 min at room temperature. Subsequently, a fresh ALP staining working solution was prepared according to the manufacturer’s instruction (Beyotime, C3206). The cells were then incubated with the ALP staining solution at 37°C in the dark for 20–30 min. Post-staining, cells were gently washed 3–5 times with distilled water (5 min each) to remove residual stain. Images were acquired using an inverted light microscope.

Flow cytometry detection

For flow cytometric analysis, cells were collected at 48 h (for cell cycle and proliferation) and 72 h (for apoptosis) post-siRNA transfection. After trypsinization and centrifugation, cells were stained following the protocols of the cell proliferation kit (Beyotime, C0071S), cell cycle kit, and apoptosis kit (BD Biosciences, 559763). The cells were incubated with the staining solution at 37°C for 30 min, with gentle agitation every 5 to 10 min to ensure uniform distribution of the dye. After incubation, the cells were washed three times with Dulbecco’s phosphate-buffered saline (DPBS) (Servicebio, G4200-100ML) to remove any unbound dye. Subsequently, the samples were subjected to flow cytometric analysis, with channel settings optimized for the specific fluorophores utilized.

RNA pull-down assay

Plasmids were extracted from E. coli DH5α competent cells in accordance with the manufacturer’s instructions provided with the plasmid extraction kit (Thermo Fisher Scientific, K0503). Purified plasmids were linearized by restriction enzyme digestion, and the resulting DNA fragments were purified using a DNA purification kit (Beyotime, D0033). For in vitro transcription, the linearized DNA template (3 μg) was combined with 2 μL biotin RNA labeling mix (Roche, 11685597910), 2 μL T7 RNA polymerase (Thermo Fisher Scientific, EP0112), 2 μL transcription buffer (Thermo Fisher Scientific, F8156AF), and 1 μL RNase inhibitor (Thermo Fisher Scientific, EO0382) in a total volume of 20 μL. This mixture was then incubated at 37°C for 6 h to synthesize biotin-labeled RNA. After transcription, 8 μg of biotinylated RNA was denatured at 90°C for 2 min, then incubated with 2 volumes of RNA structure buffer (10 mM Tris pH 7.0, 0.1 M KCl, 10 mM MgCl₂) at room temperature for 20 min to allow proper secondary structure formation. Subsequently, the refolded RNA was incubated with the prepared protein supernatant at room temperature for 2 h with gentle rotation to allow RNA–protein binding. After the binding reaction, streptavidin-coated magnetic beads (MCE, HY-K0208) were added to the mixture and incubated at room temperature for 1–2 h to capture the biotinylated RNA–protein complexes. Following a series of washes, bound proteins were eluted by boiling in SDS‑PAGE loading buffer and analyzed by silver staining and immunoblotting.

RNA immunoprecipitation (RIP) assay

The RIP assay was performed with the EZ-Magna RIP Kit (Millipore, 17-701) according to the manufacturer’s instructions. RNA coprecipitated with NPM1 was isolated, and qPCR was subsequently employed to determine the enrichment level of ESRG, with non-specific binding assessed via parallel experiments using normal rabbit IgG (Supplementary Table 2).

Co-IP experiment

Cells were collected and lysed to obtain protein extracts. The lysates were then incubated with 5 μg primary antibody at 4°C for 12 h. Then 35 μL protein A/G magnetic beads (Thermo Fisher Scientific, 88802) were added, and incubation continued for another 5 h at 4°C with rotation. After three washes with NP-40 lysis buffer (Beyotime, P0013F) buffer, the immunoprecipitated complexes were resuspended in 40 μL of NP-40 lysis buffer, mixed with 10 μL of SDS-PAGE loading buffer, and then boiled at 95°C for 5–10 min prior to immunoblot analysis.

RNA fluorescence in situ hybridization (FISH)

The cells were initially fixed using an in situ hybridization fixative for at least 20 min at room temperature, followed by three washes with DPBS. Subsequently, the cells underwent permeabilization with 0.3% Triton X-100 (Sigma, T9284) for 10 min and were then blocked with 1% bovine serum albumin (Thermo Fisher Scientific, 30066575) for 30 min at room temperature. After blocking, the cells were incubated with ESRG FISH probes (Supplementary Table 3) overnight at 37°C. Unbound probes were removed by sequential washing, and nuclei were counterstained with DAPI (Beyotime, C1006) in the dark. Target RNA signals were visualized by confocal microscopy.

Immunofluorescence

The initial procedures (fixation, permeabilization, and blocking) followed the established FISH protocol. Subsequent to the blocking step, the cells underwent incubation with the primary antibody (diluted in blocking buffer containing 1% BSA at a ratio of 1:200) at 4°C for overnight. On the following day, the cells were subjected to three washes with DPBS to remove unbound primary antibody. Subsequently, the cells were incubated with the appropriate fluorophore-conjugated secondary antibody (diluted 1:200 in blocking buffer) (Supplementary Table 3) at 3°C for a duration of 2 h. After incubation, the cells were washed three times with DPBS for 5 min each at RT to eliminate excess secondary antibody. Ultimately, the nuclei were stained with 1 mL of DAPI for 10 min.

CCK8 analysis

On the initial day (day 0), cells were plated in a 96-well format at a density of 8000 cells per well, with each well containing 100 μL of culture medium. Commencing from day 0, 10 μL of CCK-8 solution (Beyotime, C0042) was introduced into each well at 24 h intervals, followed by incubation at 37 °C with 5% CO₂ for 1.5 h to allow color development. The optical density (OD) value at 450 nm was then measured using a microplate reader. At least three technical replicates were included per group.

mRNA stability assay

The mRNA stability assay was performed as previously described [26] with minor modifications. Briefly, cells transfected with siRNA or siControl were treated with 5 μg/mL Actinomycin D (S8964, Selleck) to block transcription, then harvested at 0, 2, 4 and 6 h, followed by total RNA extraction and qRT-PCR to detect target mRNA expression (GAPDH as internal reference) for half-life calculation. All operations were performed under RNase-free conditions.

Real-time quantitative PCR (qPCR)

Total RNA extracted with TRIzol (Thermo Fisher Scientific, 15596018CN) was reverse transcribed using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, K1621). For qPCR analysis, the synthesized cDNA was mixed with qPCR Master Mix (Vazyme, Q411-02) and gene-specific primers (Supplementary Table 4), with GAPDH as the endogenous control. The qPCR reactions were executed following the thermal cycling conditions outlined in the qPCR Master Mix protocol. Relative expression was calculated using the 2⁻^ΔΔCt method [27]. All reactions were performed in triplicate experiments.

RNA-seq and bioinformatic analysis

Total RNA was extracted from control and NPM1-knockdown cells at 48 h using TRIzol reagent (Thermo Fisher Scientific, 15596018CN). RNA samples were stored at -80°C and shipped to MajorBio Co., Ltd. (Shanghai, China) for library preparation and 150 bp paired-end sequencing on an Illumina NovaSeq 6000 platform. Clean reads were mapped to the human reference genome GRCh38.p13 (Ensembl). Mapping rates ranged from 96.58 to 97.22%. A total of 33001 genes and 151738 transcripts were detected (defined as having at least one read in any sample). Differential expression analysis was performed using DESeq2, with thresholds of |log₂ fold change| ≥ 1 and adjusted p value (padj)  ≤ 0.05. KEGG pathway enrichment analysis was conducted using the clusterProfiler R package, with padj  < 0.05 considered statistically significant. RNA‑seq data were deposited in NCBI GEO under accession GSE317438.

Extraction of nuclear, cytoplasmic and extracellular proteins

After collection, cells were washed three times with PBS. Nuclear and cytoplasmic proteins were separately extracted using the PARIS™ Kit Protein and RNA Isolation System (Thermo Fisher Scientific, AM1921), following the manufacturer’s instructions with differential centrifugation.

For extracellular proteins, the culture medium was replaced 16–24 h before harvesting. The supernatant was collected and centrifuged at 4℃, 3000 rpm for 5 min to remove cell debris, followed by filtration through a 0.22 μm filter to eliminate residual contaminants. Ultrafiltration devices with a molecular weight cutoff (MWCO) no larger than one‑third the size of the target protein were used. For the mature, secreted form of BMP4 (approximately 20–30 kDa as a dimer), a 10 kDa MWCO Amicon Ultra‑15 centrifugal filter unit (Millipore, UFC9010) was selected. The supernatant was then transferred to the filter unit and concentrated by centrifugation at 4℃, 4000 × g until the desired volume was reached.

EU RNA synthesis assay

Two days after siRNA transfection, nascent RNA synthesis was measured using the BeyoClick™ EU‑488 RNA Synthesis Detection Kit (Beyotime, R0301S). A 2 × EU working solution (2 mM) was prepared by diluting the 100 mM EU stock solution 1:50 in culture medium. An equal volume of prewarmed 2 × EU solution was added to cells to achieve a final 1 × concentration. Cells were labeled for approximately 10% of the cell cycle, then fixed, permeabilized, and subjected to a 30 min Click reaction at room temperature in the dark. After washing, nuclei were counterstained with Hoechst 33342.

Statistical analyses

All data were analyzed using GraphPad Prism 9.5.1. Details of technical and biological replicates, as well as independent experiments, are provided in the figure legends. Comparisons between two groups were performed using unpaired Student’s t‑test, and multiple group comparisons were conducted using one‑way or two‑way ANOVA test. Data are presented as mean ± standard deviation (SD) from three independent experiments. Statistical significance was defined as *P < 0.05, **P < 0.01 and ***P < 0.001 indicating increased significance. Normality and variance homogeneity were formally assessed before all statistical analyses to ensure valid and reproducible results.

Results

lncRNA ESRG interacts with NPM1 in hPSCs

In this study, we verified the direct interaction between ESRG and NPM1 in hPSCs using RNA pull-down and RIP assays (Figs. 1A, B and S1A, B). Furthermore, the nuclear co-localization of ESRG and NPM1 was observed in iPSCs, as well as in the H9 and H1 cell lines by RNA-FISH combined with immunofluorescence assays (Figs. 1C and S1C). NPM1 contains multiple functional domains, including a nuclear export signal (NES), acidic regions, a nuclear localization signal (NLS), and a nucleolar localization signal (NoLS) [28]. To identify the specific region of NPM1 responsible for binding to ESRG, we generated two truncated constructs based on its amino acid sequence: NPM1 D1–2 (amino acids 1–150) and NPM1 D3–4 (amino acids 151–294) (Fig. 1D). The NPM1 D1–2 truncation contains the N-terminal core domain, which is important for the nuclear-cytoplasmic transport of NPM1 and for mediating interactions with its binding partners via a positively charged surface [29]. The NPM1 D3–4 domain harbors a NoLS sequence enriched with aromatic residues that are unique to NPM1. Subsequently, we constructed Flag-tagged expression vectors for NPM1 D1–2 and NPM1 D3–4, and performed transfection for 48 h before subsequent RNA pull-down and RIP experiments. The results showed that ESRG specifically binds to the NPM1 D1–2 domain (Figs. 1E, F and S1D). To further map the region responsible for the interaction between ESRG and NPM1, we performed RNA pull-down assays using a series of truncated ESRG constructs (1–500 nt, 501–1500 nt, 1501–2000 nt, 2000–3153 nt). These assays revealed that NPM1 associates with the 1–500 nt segment of ESRG (Fig. 1G). Together, these data indicate that the 1–500 nt region of ESRG interacts with the N-terminal 1–150 amino acid region of NPM1.

Fig. 1.

ESRG binds to NPM1 in hPSCs. A RNA pull-down assay detected the binding between ESRG and NPM1. B RIP assay detected the binding between ESRG and NPM1, n = 3, Data are presented as mean ± SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, ***P < 0.001. C ESRG was detected by RNA-FISH, while NPM1 protein was visualized by immunofluorescence in H9 cells. Scale bar =  20 μm D. Protein Structure of NPM1. E The binding of NPM1 D1–2 and D3–4 domains to ESRG was detected by RIP assay with anti-Flag antibody, n = 3. Data are presented as mean ± SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, ns, p < 0.001. F The binding of NPM1 D1–2 and D3–4 domains to ESRG was detected by RNA pull-down assay. G The binding of ESRG-1 (1–500 nt), ESRG-2 (501–1500 nt), ESRG-3 (1501–2000 nt), and ESRG-4 (2000–3153 nt) to NPM1 was detected by RNA pull-down assay

ESRG regulates NPM1 protein stability by affecting the interaction between UBQLN2 and NPM1

To investigate the role of ESRG in regulating NPM1 protein stability, we performed ESRG knockdown in hPSCs. Western blot analysis showed that NPM1 protein levels were decreased after ESRG depletion (Figs. 2A and S2A), but qPCR detected no significant alteration in NPM1 mRNA levels (Fig. S2B), indicating that ESRG regulates NPM1 at the protein level. Given the importance of the ubiquitin–proteasome system (UPS) in protein degradation, we hypothesized that ESRG may regulate NPM1 stability through this pathway. To test this hypothesis, hPSCs were treated with cycloheximide (CHX) to inhibit protein synthesis. NPM1 degradation was accelerated in ESRG-knockdown cells compared with siControl cells (Figs. 2B and S2C), indicating a reduced protein half-life. This increased degradation was mitigated by the proteasome inhibitor MG132 (Figs. 2C and S2D). Furthermore, IP assays showed enhanced ubiquitination of NPM1 following ESRG knockdown (Figs. 2D and S2E). Collectively, these results suggest that ESRG stabilizes NPM1 by inhibiting UPS-mediated protein degradation.

Fig. 2.

ESRG regulates NPM1 protein stability by affecting the interaction between UBQLN2 and NPM1. A Western blot analysis was performed to measure NPM1 protein levels after ESRG knockdown for 72 h. B NPM1 protein levels were examined by Western blot at 0, 3, 6, and 9 h after CHX (20 μg/mL) treatment in siControl and siESRG hPSCs. C NPM1 protein levels were examined by Western blot in siESRG-transfected cells treated with MG132 (20 μM) for 4 h. D NPM1 ubiquitination was detected by IP assay after ESRG knockdown. Cells were treated with MG132 (20 μM) for 4 h before harvest. NPM1 was immunoprecipitated (IP) with anti-NPM1, and ubiquitin was detected by immunoblotting (IB) with anti-Ub. E The interaction between NPM1 and UBQLN2 was examined by Co-IP assay using anti-NPM1 for IP followed by IB with anti-UBQLN2. F NPM1 protein levels were examined by Western blot in siControl, siUBQLN2, and siUBQLN2 + siESRG groups. G NPM1 ubiquitination was detected by IP assay following UBQLN2 knockdown. H The interaction between NPM1 and UBQLN2 was examined by IP assay after ESRG knockdown followed by MG132 (20 μM) treatment for 4 h. NPM1 was IP with anti-NPM1, and UBQLN2 was detected by IB with anti-UBQLN2

To explore how ESRG regulates NPM1 ubiquitination, we queried public protein–protein interaction databases. This analysis suggested that NPM1 potentially interacts with Ubiquilin 2 (UBQLN2), which contains a ubiquitin-like domain and a ubiquitin-associated domain that are involved in targeting ubiquitinated proteins for degradation [30, 31]. Co-IP experiments confirmed the interaction between NPM1 and UBQLN2 in hPSCs (Fig. 2E), and this interaction was enhanced following ESRG knockdown (Figs. 2H, S2F). Further investigation revealed that silencing UBQLN2 expression partially mitigated the reduction in NPM1 protein levels in ESRG-knockdown cells (Fig. 2F), and reduced the ubiquitination level of NPM1 (Fig. 2G).

These findings demonstrate that ESRG antagonizes NPM1 ubiquitination and degradation by impairing the UBQLN2-NPM1 interaction, thereby sustaining NPM1 protein levels in hPSCs.

NPM1 maintains the self-renewal of hPSCs

To investigate the biological function of NPM1 in hPSCs, we performed knockdown using two specific siRNAs, siNPM1-1 and siNPM1-2. qPCR, Western blotting (Fig. 3A), and immunofluorescence analyses (Fig. 3B) confirmed efficient knockdown. Morphological observations showed that after NPM1 depletion, cells lost their typical round, compact colony morphology and became irregular and fusiform, with loosened cell–cell contacts and increased floating dead cells (Fig. 3C and S3A). Concurrently, cell proliferation rates decreased (Fig. 3D and S3C) and apoptotic cell numbers increased (Fig. 3E). Western blotting showed downregulation of the stemness markers OCT4, SOX2, and NANOG (Fig. 3F and S3B). Cell cycle analysis indicated an increased proportion of cells in G1 and G2/M phases and a decreased proportion in S phase, indicating that cells underwent both G1 and G2/M arrest (Fig. 3G). In addition, ALP staining intensity was reduced in NPM1-knockdown cells, indicating decreased ALP activity (Fig. 3H and S3D). Given that NPM1 is a nucleolar protein, we examined nascent RNA synthesis in hPSCs after ESRG or NPM1 knockdown. EU incorporation assays revealed that depletion of either ESRG or NPM1 significantly suppressed global RNA synthesis (Fig. S3E).

Fig. 3.

NPM1 maintains the self-renewal of hPSCs. A qPCR was used to detect the knockdown efficiency of NPM1 by siRNA in hPSCs, n = 3. Data are presented as mean ± SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, ***p < 0.001. Western blot was performed to examine NPM1 protein levels after NPM1 knockdown. B The knockdown efficiency of NPM1 by siRNA in hPSCs was measured by IF, scale bar = 50 μm. C Phenotypic changes of cells were recorded by microscope ( siControl and siNPM1 -1), scale bar = 100 μm. D Cell proliferation ability after NPM1 knockdown was analyzed by flow cytometry. Data are presented as mean ± SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, *p < 0.05. E Cell apoptosis rate after NPM1 knockdown was detected by flow cytometry, n = 3. Data are presented as mean  ±  SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, ***p < 0.001. F Western blot was performed to detect the expression levels of stemness markers (NANOG, SOX2, and OCT4) after NPM1 overexpression. G Cell cycle distribution after NPM1 knockdown was detected by flow cytometry, n = 3. The percentages of cells in G1, S, and G2/M phases are presented as mean ± SD from three independent experiments. Statistical comparisons between two groups were performed using unpaired two-tailed Student’s t-test,***p < 0.001. H Alkaline phosphatase staining of hPSCs after knockdown of NPM1, scale bar = 100 μm

In conclusion, our results support that NPM1 acts as a critical regulator in maintaining the self-renewal capacity of hPSCs. The phenotypes observed after NPM1 knockdown, including morphological alterations, downregulation of stemness markers, and impaired proliferation, were highly similar to those observed upon ESRG knockdown [6]. Combined with our previous finding that ESRG modulates NPM1 protein stability, we propose that ESRG may maintain NPM1 protein stability by interacting with NPM1, thereby sustaining the self-renewal of hPSCs.

ESRG maintains hPSC self-renewal by targeting its downstream effector NPM1

To test our hypothesis that ESRG regulates hPSC self-renewal through NPM1, we performed rescue experiments by knocking down ESRG while overexpressing NPM1 in hPSCs. siControl and ESRG-knockdown cells were used as controls to evaluate cellular phenotypic and functional alterations. We found that compared with the ESRG knockdown group, cells overexpressing NPM1 showed partial phenotypic rescue, with attenuated differentiation-associated morphology (Fig. 4A and S4A). Furthermore, protein levels of SOX2, OCT4, and NANOG were markedly restored (Fig. 4B). ALP staining intensity and activity were enhanced in the NPM1 overexpression group, indicating restored pluripotency (Fig. 4C and Fig. S4B).

Fig. 4.

ESRG maintains hPSC self-renewal by targeting its downstream effector NPM1. A Cell morphology was recorded by microscopy (siControl, siESRG and siESRG + oeNPM1), scale bar = 100 μm. B Western blot was performed to detect the expression levels of stemness markers (NANOG, SOX2, and OCT4) after ESRG knockdown and NPM1 overexpression. C Alkaline phosphatase staining of hPSCs after ESRG knockdown and NPM1 overexpression, scale bar = 100 μm. D Cell proliferation ability was detected by flow cytometry (siESRG vs siESRG + oeNPM1), n = 3. Data are presented as mean ± SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, ***p < 0.001. E Cell apoptosis was detected by flow cytometry (siESRG vs siESRG + oeNPM1), n = 3. Data are presented as mean ± SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, ***p < 0.001.

Flow cytometry showed enhanced cell proliferation (Fig. 4D, S4C) and diminished apoptosis (Fig. 4E, S4D), indicating that NPM1 overexpression partially reverses the proliferation defects and elevated apoptosis caused by ESRG knockdown. These results indicate that NPM1 overexpression rescues multiple phenotypic defects in ESRG-depleted hPSCs, including compromised pluripotency, impaired proliferation, and elevated apoptosis. These data demonstrate that NPM1 is essential for maintaining the self-renewal of hPSCs as a key downstream effector of ESRG.

ESRG maintains the self-renewal of hPSCs by regulating the TGF-β signaling pathway via NPM1

To elucidate the downstream molecular mechanisms by which ESRG maintains the self-renewal of hPSCs through its interaction with NPM1, hPSCs were treated with NPM1 siRNA for 48 h, and we then performed comparative gene expression profiling between NPM1-knockdown and control groups by transcriptome sequencing. KEGG pathway enrichment analysis showed that differentially expressed genes were enriched in the TGF-β signaling pathway, stem cell pluripotency maintenance, and RNA degradation (Fig. 5A). The TGF-β signaling pathway exerts a bidirectional role in regulating hPSC self-renewal [32, 33]. To examine whether NPM1 modulates the TGF-β pathway, we knocked down NPM1 and measured protein levels of TGF-β1, Smad2, and Smad3 by Western blot. Protein levels of TGF-β1, Smad2, and Smad3 were reduced after NPM1 knockdown (Fig. 5B). ESRG knockdown and NPM1 overexpression partially restored the expression of TGF-β1, Smad2, and Smad3 (Fig. 5C). To further explore the regulatory roles of ESRG and NPM1 in the TGF-β pathway, we treated ESRG- or NPM1-knockdown cells with the TGF-β agonist SRI-011381. This treatment rescued the reduced protein levels of TGF-β1, Smad2, and Smad3 (Fig. 5D, E). Furthermore, it restored ALP activity (Fig. 5 G) and attenuated the reduction in proliferation caused by ESRG or NPM1 knockdown (Fig. S5A, B). These results indicate that the ESRG–NPM1 axis maintains hPSC self-renewal by positively regulating the TGF-β signaling pathway.

Fig. 5.

ESRG maintains the self-renewal of hPSCs by regulating the TGF-β signaling pathway via NPM1. A KEGG pathway enrichment analysis of differentially expressed genes between control and NPM1-knockdown cells identified by RNA-seq, n = 3. B Western blot analysis of TGF-β1, Smad2, and Smad3 after NPM1 knockdown. C Western blot analysis of TGF‑β1, Smad2, and Smad3 expression following ESRG knockdown and NPM1 overexpression. D Western blot analysis of TGF-β1, Smad2, and Smad3 in ESRG-knockdown cells treated with 500 µM SRI-011381 for 12 h. E Western blot analysis of TGF-β1, Smad2, and Smad3 in NPM1-knockdown cells treated with 500 µM SRI-011381 for 12 h. F Western blot analysis of OCT4, SOX2, and NANOG protein levels in ESRG‑ or NPM1‑knockdown cells treated with 500 µM SRI-011381 for 12 h. G Alkaline phosphatase staining of hPSCs after knockdown of ESRG or NPM1, scale bar = 100 µm

ESRG regulates BMP4 expression via NPM1 to modulate the TGF-β signaling pathway

To explore how the ESRG-NPM1 axis regulates the TGF-β pathway to maintain the self-renewal capacity of hPSCs, we performed the following analyses. First, using sequence alignment from the NCBI database and protein interaction prediction from BioGRID, we identified a potential direct interaction between NPM1 and BMP4, which was verified by Co-IP assays (Fig. 6A). BMP4 is a key member of the TGF-β superfamily. It shares similar signal transduction mechanisms with the TGF-β signaling pathway [34]. Considering that BMP4 is a secreted signaling molecule with specific subcellular localization, we examined the interaction between NPM1 and BMP4 in different cellular fractions. Immunofluorescence co-localization and fraction-specific Co-IP assays showed that the binding between NPM1 and BMP4 mainly occurred in the nucleus, weakly in the cytoplasm, and no interaction was detected in the extracellular fraction (Fig. 6B, C). To clarify the regulatory effect of NPM1 on BMP4, we knocked down NPM1 in hPSCs using siRNA. Western blot analysis showed that the intracellular BMP4 protein level was significantly increased in NPM1-knockdown cells (Fig. 6D); CHX chase assays demonstrated that NPM1 knockdown slowed the degradation of BMP4 protein and prolonged its half-life (Fig. 6E). qPCR analysis showed that BMP4 mRNA levels were significantly upregulated after NPM1 knockdown (Fig. S6A); mRNA stability assays with actinomycin D (to block new transcription) showed no significant difference in BMP4 mRNA half-life between NPM1-knockdown and control groups (Fig. S6B), indicating that NPM1 does not affect BMP4 mRNA stability. In addition, Western blot detection of cell culture supernatants showed that the levels of secreted BMP4 were significantly increased after NPM1 knockdown, consistent with the upregulation of intracellular BMP4 protein (Fig. S6C). In summary, NPM1 regulates BMP4 at multiple levels, including its protein stability, mRNA abundance, and extracellular secretion. Our data showed that NPM1 affects BMP4 mRNA levels and extracellular secretion, and modulates BMP4 protein levels by regulating its half-life, without significantly altering BMP4 mRNA stability.

Fig. 6.

ESRG regulates BMP4 expression via NPM1 to modulate the TGF-β signaling pathway. A The interaction of NPM1 with BMP4 in hPSCs was detected by Co-IP. B NPM1 and BMP4 protein were visualized by immunofluorescence in H9 cells. Scale bar = 10 μm. C Co-IP analysis of the interaction between BMP4 and NPM1 in nuclear, cytoplasmic and extracellular fractions. D Western blot analysis of BMP4 protein levels after NPM1 knockdown. E BMP4 protein levels were examined by Western blot at 0, 3, 6, and 9 h after CHX (20 μg/mL) treatment in siControl and siNPM1 hPSCs.

These results suggest that NPM1 negatively regulates BMP4 expression through direct binding, implying a negative feedback mechanism that contributes to the maintenance of hPSC self-renewal.

To test this hypothesis, we knocked down BMP4 following NPM1 knockdown in hPSCs. Compared with NPM1 siRNA alone, this sequential knockdown restored the expression of pluripotency markers OCT4, NANOG, and SOX2 (Fig. 7A), with increased cell proliferation (Fig. 7B) and elevated ALP activity (Fig. 7C).

Fig. 7.

The ESRG-NPM1 axis maintains the self-renewal of hPSCs by regulating BMP4. A Western blot analysis of stemness markers (NANOG, SOX2, and OCT4) was performed after simultaneous knockdown of NPM1 and BMP4. B Cell proliferation after simultaneous knockdown of NPM1 and BMP4 was detected by CCK-8 assay, n = 3. Data represent mean ± SD from three independent biological replicates. Statistical significance was determined by two‑sided Student’s t‑test for pairwise comparisons between siNPM1–1 and siNPM1–1 + siBMP4, and between siNPM1–2 and siNPM1–2 + siBMP4. *p < 0.05, **p < 0.01, ***p < 0.001. C Alkaline phosphatase staining of hPSCs after simultaneous knockdown of NPM1 and BMP4, scale bar = 100 µm

These findings indicate that the ESRG-NPM1 axis sustains hPSC self-renewal by negatively regulating BMP4 protein levels. Given that the ESRG-NPM1 axis also influences the TGF-β pathway, we propose that BMP4 and TGF-β signaling may act antagonistically in regulating hPSC self-renewal.

Currently, two perspectives exist regarding the interaction between BMP4 and the TGF-β signaling pathway. To examine the regulatory relationship between BMP4 and the TGF-β pathway, we knocked down NPM1 and BMP4 simultaneously in hPSCs. Our results showed that BMP4 knockdown partially alleviated the inhibition of the TGF-β pathway caused by NPM1 knockdown (Fig. 8A). However, the mechanism by which the ESRG-NPM1-BMP4 axis modulates the TGF-β pathway requires further investigation. Subsequently, BioGRID database analysis and experimental validation revealed that BMP4 binds to PTBP1 (Fig. 8B). We further found that PTBP1 protein levels were increased after BMP4 knockdown (Fig. 8D). Previous studies have demonstrated that PTBP1 recognizes and binds to TGF-β1 mRNA, thereby stabilizing TGF-β1 mRNA [35]. Our results indicated that PTBP1 binds to mature TGF-β1 mRNA (Fig. 8E) without affecting its pre-mRNA levels (Fig. S7).

Fig. 8.

The ESRG/NPM1/BMP4 axis regulates TGF-β signaling by modulating the stability of TGF-β1 mRNA. A Western blot analysis of TGF-β1, Smad2 and Smad3 proteins after simultaneously knockdown NPM1 and BMP4 expression. B Co-IP assay for detecting the interaction between BMP4 and PTBP1. C Western blot analysis of PTBP1 protein after knockdown of ESRG or NPM1 expression. D Western blot analysis of PTBP1 protein after knockdown of BMP4 expression. E PTBP1 protein levels were examined by Western blot at 0, 3, 6, and 9 h after CHX (20 μg/mL) treatment in siControl and siBMP4 hPSCs. F The binding of PTBP1 and TGF-β1 mRNA was detected by RIP assay after knockdown of ESRG or NPM1 expression, n = 3, Data are presented as mean ± SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, *** p < 0.001. G qRT-PCR was used to examine the half-life of TGF-β1 mRNA after knockdown of PTBP1 expression, n = 3. Data are presented as mean ± SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, **P < 0.01. H qRT-PCR was used to examine the half-life of TGF-β1 mRNA after overexpression of PTBP1, n = 3. Data are presented as mean ± SD from three independent experiments. Statistical significance was determined by unpaired two-tailed Student’s t-test, **p < 0.01. I qRT-PCR was used to examine the half-life of TGF-β1 mRNA after knockdown of ESRG or NPM1 expression. n = 3, Data represent mean ± SD from three independent biological replicates.Statistical analysis was performed by one-way ANOVA with Tukey’s post-hoc test. **P < 0.01

Functional assays showed that PTBP1 knockdown reduced TGF-β1 mRNA stability (Fig. 8G), whereas PTBP1 overexpression exerted the opposite effect (Fig. 8H). In addition, knockdown of either ESRG or NPM1 decreased PTBP1 expression (Fig. 8C), attenuated the binding between PTBP1 and TGF-β1 mRNA and shortened the half-life of TGF-β1 mRNA (Fig. 8I).

These findings suggest that the ESRG-NPM1-BMP4 axis regulates the TGF-β pathway and maintains hPSC self-renewal by modulating TGF-β1 mRNA stability.

To further clarify whether the ESRG/NPM1/BMP4/PTBP1 regulatory axis modulates TGF‑β signaling by regulating TGF-β1 mRNA stability, thereby sustaining the self‑renewal of hPSCs, we knocked down PTBP1 in hPSCs. Depletion of PTBP1 decreased the protein levels of Smad2, p‑Smad2, Smad3, and p‑Smad3 (Fig. 9A), accompanied by reduced ALP activity (Fig. 9C). Conversely, overexpression of PTBP1 increased the levels of total and phosphorylated Smad2/3 proteins (Fig. 9B). These results functionally validate the ESRG/NPM1/BMP4/PTBP1/TGF-β1 regulatory cascade in maintaining the self‑renewal of hPSCs.

Fig. 9.

The ESRG‑NPM1‑BMP4‑PTBP1 cascade modulates TGF-β1 signaling to maintain hPSC self‑renewal. A Western blot analysis of Smad2, p‑Smad2, Smad3, and p‑Smad3 expression after knockdown of PTBP1. B Western blot analysis of Smad2, p‑Smad2, Smad3, and p‑Smad3 expression after overexpress of PTBP1. C Alkaline phosphatase staining of hPSCs after knockdown of PTBP1, scale bar = 100 µm

Discussion

Long non-coding RNAs (lncRNAs), once considered transcriptional noise, are now recognized as key regulators of gene expression, with research rapidly advancing functional characterization to mechanistic dissection. Specific lncRNAs play critical roles in regulating hPSC self-renewal [36]. hPSCs can differentiate into all cell types in the body, making them an important cell source for basic research and future cell-based therapies. Understanding the regulatory mechanisms of hPSC self-renewal and differentiation is therefore critical for advancing stem cell therapy. However, the key lncRNAs and their downstream effectors that govern these processes remain largely unknown.

Notably, the lncRNA ESRG is highly expressed in undifferentiated hPSCs but exhibits minimal or no expression in normal tissues. Recent studies suggest that ESRG marks early reprogrammed cells [37] and undifferentiated hPSCs and can identify residual undifferentiated hPSCs in differentiated populations [38, 39]. ESRG knockdown induces differentiation, increases apoptosis, and reduces pluripotency gene expression [6, 7]. Thus, ESRG may play a central role in maintaining the self-renewal and pluripotency network of hPSCs.

This study shows that ESRG directly binds to NPM1, inhibits its ubiquitination and degradation, and thereby maintains NPM1 protein stability. These results suggest that NPM1 acts as a downstream effector in ESRG-mediated regulation of hPSC self-renewal. As a member of the nucleophosmin family, NPM1 influences cell proliferation, nucleocytoplasmic transport, and apoptosis [40]. NPM1 knockout in cells results in diminished proliferative capacity [41–43]. For example, NPM1 forms a complex with chromobox protein homolog 3 (CBX3) to promote peroxiredoxin 6 (PRDX6) transcription, thereby promoting the proliferation of colorectal cancer cells [44]. NPM1 also exhibits important functions in hESCs; a previous study showed that lncRNA-NNT-AS1 upregulates NPM1 expression by sponging miR-30c, thereby promoting hESC proliferation [45]. In the present study, silencing NPM1 induced obvious morphological changes in hPSCs, including an elongated, fibroblast-like appearance and loose cell arrangement. NPM1 knockdown also decreased ALP activity, downregulated the expression of core pluripotency factors including OCT4, SOX2, and NANOG, and reduced cell proliferation.

Multiple studies using microarray or omics sequencing analyses have suggested that activation of the TGF‑β signaling pathway is associated with self‑renewal and dedifferentiation [46, 47], and in pluripotent stem cells, these analyses have further shown that TGF‑β signaling activates a set of transcription factors—including OCT4, NANOG, and SOX2—that are crucial for self‑renewal and the prevention of premature differentiation, thereby maintaining pluripotency [48]. Consistent with these reports, our RNA‑seq data showed that the TGF‑β signaling pathway was significantly enriched in response to NPM1 knockdown. Knockdown of ESRG or NPM1 decreased the protein levels of TGF‑β1, Smad2, and Smad3. Activation of TGF‑β signaling (SRI-011381) reversed these changes and partially restored the pluripotency defects caused by ESRG or NPM1 depletion.

Through a combination of bioinformatic prediction and experimental validation, we identified that NPM1 interacts with BMP4. As a member of the TGF-β superfamily, BMP4 regulates the differentiation and proliferation of stem cells. During in vitro directed differentiation of hPSCs, a common strategy is to optimize the concentration gradient of BMP4 to control differentiation and induce specific cell lineages [49–51]. TGF-β and BMP4 often function antagonistically to regulate diverse cellular processes, and this balance is important for tissue morphogenesis, organogenesis, and adult tissue homeostasis [52]. For example, in vascular remodeling, BMP4 reduces Smad2/3 protein levels and inhibits TGF-β1 activity to influence cell proliferation [53]. In pulmonary fibrosis, BMP4 suppresses the proliferation and anti-apoptotic ability of primary lung fibroblasts induced by TGF-β1, thus exerting an anti-fibrotic effect [54]. Our study showed that knockdown of ESRG or NPM1 increased BMP4 protein levels and decreased the level of PTBP1, a protein that stabilizes TGF-β1 mRNA, thereby resulting in reduced TGF-β1 mRNA stability.

In conclusion, this study suggests that lncRNA ESRG contributes to the regulation of hPSC self-renewal via direct interaction with NPM1. Our results indicate that ESRG binds to NPM1, which in turn modulates BMP4 abundance and TGF-β pathway activity. These findings provide new insights into the molecular network underlying hPSC self-renewal, and identify PTBP1 as a potential regulatory node through which BMP4 counteracts TGF-β signaling. We therefore propose the ESRG-NPM1-BMP4-PTBP1-TGF-β1 module as a candidate regulatory axis that may inform future strategies for optimizing hPSC expansion.

Despite the findings of this study, several limitations must be acknowledged. First, this study primarily focuses on the regulatory mechanism by which ESRG maintains hPSC self-renewal via stabilizing NPM1; however, NPM1 overexpression can only partially restore the reduced self-renewal capacity induced by ESRG knockdown, indicating that additional mechanisms through which ESRG regulates hPSC biological functions remain to be explored. Second, our findings that ESRG stabilizes NPM1 protein were primarily obtained using CHX chase and MG132 treatment assays. In future investigations, the application of specific small-molecule inhibitors targeting NPM1, the dTAG system, or other targeted protein degradation strategies will be valuable to further validate these conclusions and enable more precise and acute modulation of NPM1 protein levels in human pluripotent stem cells. Third, our results suggest that NPM1 regulates BMP4 at both the mRNA and protein stability levels, while also modulating its extracellular secretion, and the detailed molecular mechanisms underlying this regulation warrant further investigation. Additionally, although the ESRG-NPM1-BMP4-TGF-β1 regulatory axis is proposed as a potential target for optimizing hPSC expansion, its practical feasibility in clinical settings remains unclear, primarily due to the lack of investigations related to clinical translation—such as the optimization of hPSC expansion protocols and safety evaluations of expanded cells. Although further investigation is required to resolve the remaining issues, our work enriches insights into how lncRNAs orchestrate the signaling balance necessary for sustaining hPSC self-renewal (Fig. 10).

Fig. 10.

ESRG maintains hPSCs self-renewal by binding NPM1 to regulate BMP4 protein and TGF-β signaling pathway

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

CPR and XJJ conceived and designed the study. ZLL and CZ collected the data and wrote the manuscript. LW, WDL, JJT, WX, JHOY, CYY, BZ, XL, YXFand QPH revised the manuscript. All authors read and approved the final manuscript.

Funding

This review was supported by the National Natural Science Foundation of China (82471424, 82071399, 30871246).

Data availability

The data used to support the findings of this study are available from the corresponding author upon request. Microarray data have been deposited in the NCBI GEO under accession GSE317438.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no affiliations with or involvement in any organization or entity with any financial interest in the subject matter or materials discussed in this manuscript.