KLF3 promotes the 8‐cell‐like transcriptional state in pluripotent stem cells

Abstract Objectives Mouse embryonic stem cell (mESC) culture contains various heterogeneous populations, which serve as excellent models to study gene regulation in early embryo development. The heterogeneity is typically defined by transcriptional activities, for example, the expression of Nanog or Rex1 mRNA. Our objectives were to identify mESC heterogeneity that are caused by mechanisms other than transcriptional control. Materials and methods Klf3 mRNA and protein were analysed by RT‐qPCR, Western blotting or immunofluorescence in mESCs, C2C12 cells, early mouse embryos and various mouse tissues. An ESC reporter line expressing KLF3‐GFP fusion protein was made to study heterogeneity of KLF3 protein expression in ESCs. GFP‐positive mESCs were sorted for further analysis including RT‐qPCR and RNA‐seq. Results In the majority of mESCs, KLF3 protein is actively degraded due to its proline‐rich sequence and highly disordered structure. Interestingly, KLF3 protein is stabilized in a small subset of mESCs. Transcriptome analysis indicates that KLF3‐positive mESCs upregulate genes that are initially activated in 8‐cell embryos. Consistently, KLF3 protein but not mRNA is dramatically increased in 8‐cell embryos. Forced expression of KLF3 protein in mESCs promotes the expression of 8‐cell‐embryo activated genes. Conclusions Our study identifies previously unrecognized heterogeneity due to KLF3 protein expression in mESCs.

cultured mESCs are found to activate 2-cell embryo specific genes including retrotransposons MERVL and Zscan4. 6,7 Since then, the transition between ordinary mESCs and these special 2-cell-like cells have been adopted as a model system to study early events in 2-cell embryos. [8][9][10][11][12] These studies have provided significant insights on the transcriptional regulation during zygotic genome activation. However, an important question remains whether cell populations resembling other early embryonic stages prior to blastocyst development exist.
Embryonic stem cell fate is determined by multiple layers of regulation including epigenetics, transcription and translation.
Accordingly, previous studies have identified a number of key regulators including chromatin modifiers, transcription factors (TFs), RNA binding proteins and non-coding RNAs in ESCs. Sitting at the core of ESC transcriptional network are TFs Oct4 (also known as Pou5f1), Sox2 and Nanog. 13 In addition, protein interactome studies followed by functional dissection have revealed that many other TFs work together with these core TFs to modulate ESC fate. [14][15][16] Interestingly, the activity of promoters or enhancers of some TFs such as Nanog 2 and Rex1 (also known as Zfp42) 4 displays various degrees of heterogeneity in mESC cultures. While the relevance of heterogeneity in vitro to the cell fate determination in vivo is still in debate, studying the phenomenon has provided plenty insights on gene regulation in mESCs. Moreover, the inconsistency between RNA and protein expression is observed in around 50% differentially regulated genes during mESC differentiation, 17 suggesting high prevalence of translation and post-translational regulation in mESCs. Recent studies at systemic level or on individual genes find that post-translational regulation, [18][19][20][21] in particular protein degradation, may play important roles in mESC fate determination. Nevertheless, mESC heterogeneity due to protein stability control has not been reported.
Kruppel-like factor (KLF) family transcription factors play important roles in diverse physiologic and pathologic conditions. 22,23 The family has 17 members sharing highly conserved DNA-binding domain, which locates at the C terminal and consist of three C2H2 zinc fingers. In contrast, the N-terminal domain of KLF proteins is more variable, which recruit different coregulators to exert diverse biological functions. Klf2, Klf4 and Klf5 have been shown as essential transcription factors for pluripotency maintenance in mESCs. 21,24,25 Among them, Klf4 has been used as one of the four reprogramming factors in the first study of induced pluripotent stem cells. 26 However, except for Klf2, Klf4 and Klf5, the function for most other KLF proteins is still unknown in mESCs.
In the present study, we find that Klf3 mRNA is expressed but its protein is repressed in mESCs. KLF3 protein is translated but vigorously degraded due to its highly proline-rich sequence, which leads to an internally disordered structure. Surprisingly, we find that KLF3 protein is expressed in a very small subset of mESCs, which upregulates numerous genes that are initially activated during 4-to 8-cell embryo stages. Interestingly, KLF3 protein but not mRNA is also highly upregulated in 8-cell embryos. Importantly, KLF3 protein may play a functional role in activating these transcripts, supported by the evidence that forced expression of KLF3 protein upregulates 8-cell-embryo activated transcripts. Together, this study highlights the importance of post-translational regulation in early embryo development and identifies a distinctive population of mESCs marked by KLF3 protein expression. C2C12, Hepa1-6 and CT26 were cultured in DMEM/F-12 (Invitrogen, Cat. #C11330500BT) with 10% foetal bovine serum (Gemini, Cat. #900-108), penicillin-streptomycin (Gibco, Cat. #15140163) and 0.1 mmol/L β-mercaptoethanol. For EB differentiation, mESCs were cultured in ultra-low adherent plates. mESC medium without LIF (DMEM with non-essential amino acids, L-glutamine, foetal bovine serum, penicillin-streptomycin and β-mercaptoethanol) was used. For ESC differentiation without LIF, ESC medium without LIF was used.

| MATERIAL S AND ME
For mESC differentiation induced by retinoic acid (RA), 1 µmol/L RA was added to mESC medium without LIF.
The KLF3 rabbit polyclonal antibody was made in Biodragon Immunotech Company.

| Western blot
Cells were collected and directly lysed in lysis buffer containing RIPA Buffer (Beyotime) with PMSF (Sigma) and phosphatase inhibitor (Roche).  Table S3. Source data for qPCR analysis are provided in Data S1.

| Polysome fractionation assay
Cells were treated with 100 μg/mL cycloheximide for 5 minutes and then scraped with ice-cold PBS containing 100 μg/mL cy- Before the extraction of RNA from each fraction, tagRFP mRNA was added as spike-in. For qPCR data analysis, the spike-in RFP mRNA was used as control.

| IF staining
Cells were fixed with 4% paraformaldehyde for 20 minutes at room temperature. After the fixation, cells were permeabilized with 0.25% Triton X-100 for 20 minutes at room temperature and blocked with 3% FBS in PBS for 1 hour at room temperature. Cells were then in-

| Vector construction
Doxycycline-inducible plasmids were constructed from pBlue-Script II; PciI and PsiI were used to remove unnecessary sequence.
Plasmids expressing various fusion proteins of KLF3 and Renilla were modified from psiCheck-2 (Promega). Codon optimization of Klf3 was performed by the software at https://www.gensc ript.com/tools/ rare-codon-analysis. Codon optimized Klf3 and proline-mutated Klf3 sequences were synthesized by GENEWIZ Company.

| Luciferase assay
Cells were seeded in 96-well plates and grown for 16 hours, and then transfected with plasmids and cultured for another 48 hours before lysis. Protein were extracted and processed for luciferase assay using the Dual-Luciferase Reporter Assay System (Promega). Source data for luciferase assay are provided in Data S1.

| Flow cytometry
Cells were trypsinized and re-suspended in ice-cold PBS containing 2% FBS. Sorting was performed using Aria SORP (Becton, Dickinson and Company). During sorting, cells were collected in culture medium and kept at 4°C. The fraction of KLF3-GFP-positive cells was analysed by BD LSRFortessa SORP (Becton, Dickinson and Company).
Data were analysed by FlowJo software.

| Low-input RT-qPCR
For each sample, 10 cells were sorted by flow cytometry into 2 μL mild hypotonic lysis buffer composed of 0.2% Triton X-100 and 2 U/ μL of RNase inhibitor (Ambion, Cat. #AM2684). Reverse transcription reaction and PCR pre-amplification were performed as previously described. 12,28 The qPCR was performed using AceQ qPCR SYBR Green Mater Mix (Vazyme, Cat. #Q141). protocol was applied to construct RNA libraries. 28 The RNA-seq libraries were sequenced on Illumina HiSeq platform (Genewiz). All sequencing reads were aligned to the mouse genome (mm10) with STAR (version 2.5.0) using the GENCODE transcript annotation as transcriptome guide. All programs were processed following default settings except for special annotation. The FPKM values generated by Cufflinks (version 2.2.1) were used to quantify the expression level. Differentially expressed genes were determined by DESeq2.

| RNA-seq and bioinformatics analysis
GO enrichment analysis was differentially expressed genes with fold change > 2 or < 0.5 and P-value < .01, and was performed with DAVID v.6.8. The enrichment of selected gene sets was calculated by java GSEA Desktop Application. R software (v3.5.1) was used for the generation of scatter plot, box plot and Venn plot.
The 2C-specific ZGA genes are genes activated during ZGA (the 2C stage) that are also enriched in 2C::tdTomato + cells. 7 Single cell RNA-seq data from Ref 39 were used to define 8-cell-embryo activated genes, which are upregulated at least 16-fold from 4-cell to 8-cell stage. For KLF3 motif enrichment analysis, KLF3 binding motif was obtained from MEME suite (http://meme-suite.org/db/motifs).
The promoter region was defined as −1 to + 1kb around transcription start sites. RNA-sequencing data have been deposited in the Gene Expression Omnibus (GEO) under accession code GSE157790.

| Quantification and statistical analysis
Data were presented as mean ± SD except where indicated otherwise. Statistical analyses were performed using the GraphPad Prism v6 software and R software (v3.5.1). Statistical significance was assessed by two-tailed t test. For multiple comparison, the P-value was calculated by one-way or two-way ANOVA with Dunnett's test.

| The Klf3 mRNA but not protein is expressed in mESCs
We first analysed the expression of Klf family TFs in mESCs based on previously published RNA-seq in mESCs 29-34 ( Figure S1A). As expected, pluripotency regulating Klf2, Klf4 and Klf5 is expressed at high levels in mESCs. Interestingly, we found that Klf3, Klf9, Klf10 and Klf16 are also expressed at relatively high levels, although lower than these canonical pluripotency regulating KLF genes. Previous studies have reported that Klf3 plays important roles in erythropoiesis, 35 adipogenesis, 36 lymphopoiesis 37 and cardiovascular development. 38 We then focused our analysis on Klf3. We first checked the protein expression in V6.5 mESCs with commercial and in-house antibodies. Surprisingly, KLF3 protein was barely detectable in V6.5 mESCs ( Figure 1A). In contrast, in skeletal muscle (SM) which expressed mRNA at similar levels as mESCs, KLF3 protein was highly expressed. We further checked KLF3 protein level in Hepa1-6, CT26, C2C12, brain, liver and heart. Interestingly, even though Klf3 mRNA was found to be expressed at higher levels in these samples than in SM, only heart expressed detectable KLF3 protein ( Figure 1A,B).
These results demonstrate that Klf3 protein is repressed in mESCs and various other cells.

| The inhibition of KLF3 protein expression is relieved in differentiating myoblasts but not mESCs
C2C12 can be differentiated into SM lineage in vitro. We then checked whether KLF3 protein becomes expressed during C2C12 differentiation. Consistent with successful differentiation into SM lineage, SM markers Ckm and Myl1 were gradually but strongly upregulated at both RNA and protein levels ( Figure 1C,D). Interestingly, although Klf3 mRNA was not significantly varied at different time points, KLF3 protein was extensively upregulated during SM differentiation ( Figures 1C,D and S1B). Next we checked whether KLF3 protein level is upregulated during mESC differentiation. We failed to detect any KLF3 protein in embryoid bodies, mESCs differentiated in media without LIF supplement or mESCs differentiated in the presence of all trans-retinoid acid (RA) ( Figure 1E-G and S1C), suggesting that KLF3 protein is also repressed in differentiating mESCs.
Likewise, we also failed to detect KLF3 protein in a Dgcr8 knockout mESC line ( Figure 1H and S1C), indicating that the repression of KLF3 protein was not due to miRNA-mediated regulation. Together, these results demonstrate that the expression of KLF3 protein is suppressed in many different cells but can be dynamically upregulated during myoblast differentiation.

| KLF3 protein is translated but repressed through degradation in mESCs
Next we sought to understand mechanism underlying KLF3 protein repression in mESCs and C2C12. The repression of protein expression could happen at translational or post-translational level. We first performed polysome fractionation assay to check whether Klf3 mRNA is translated. The results showed that Klf3 mRNA was bound by polysome in a similar pattern to highly translated Gapdh mRNA ( Figure 2A), suggesting that Klf3 is translated in ESCs. Like in ESCs, Klf3 was also bound by polysome in C2C12 cells, and there was no difference in the binding pattern between C2C12 cells and differentiated C2C12 cells ( Figure S2). These data suggest that KLF3 protein is translated but repressed through post-translational regulation in mESCs and C2C12.
Since KLF3 protein was successfully translated, we then checked whether KLF3 protein is degraded in mESCs. We tested this possibility through a sensitive assay using KLF3 and Renilla luciferase fusion constructs ( Figure 2B). Compared to unfused Renilla control, Klf3-Renilla was repressed around 500-fold in mESCs ( Figure 2C). This result confirmed the repression of KLF3 protein in mESCs. To distinguish degradation versus translational repression, we introduced T2A peptide sequence between Klf3 and Renilla luciferase ( Figure 2B). Because T2A peptide can induce 'ribosome skipping' between last 2 amino acids, 39 KLF3 and Renilla will be produced as two independent proteins, therefore excluding destabilization effects due to KLF3 peptide sequence on Renilla. In other words, if KLF3 is repressed through degradation, the split between KLF3 and Renilla will likely lead to higher luciferase expression. Indeed, we observed ~10-fold increase in Renilla level upon the introduction of T2A peptide ( Figure 2C). We further analysed the expression difference of Renilla between Gapdh-Renilla and Klf3-Renilla constructs, and between Gapdh-T2A-Renilla and Klf3-T2A-Renilla constructs ( Figure 2D). Theoretically, the difference between GAPDH-Renilla F I G U R E 1 The expression of KLF3 protein is repressed in mESCs. A, Western blotting analysis of KLF3 protein in wild-type mESCs (V6.5), Hepa1-6, CT26, brain, heart, liver and skeletal muscle. Two antibodies were shown including in-house antibody and commercial antibody from Abnova. For each sample, 20 μg protein was loaded. B, Relative RNA expression of Klf3 gene in wild-type mESCs (V6.5), Hepa1-6, CT26, brain, heart, liver and skeletal muscle. The Hsp90ab1 gene was used as a control. For each gene, data were normalized to the mRNA level of wild-type mESCs (V6.5). Shown is mean with range of two biological replicates. C, Relative RNA expression of Klf3, Ckm and Myl1 during C2C12 differentiation. The Gapdh gene was used as a control. For each gene, data were normalized to the mRNA level in day 12 differentiated C2C12. Shown is mean ± SD, n = 3 biological replicates. D, Western blotting analysis of KLF3 and CKM during C2C12 differentiation. For each sample, 20 μg protein was loaded. Data were quantification of protein level normalized to HSP90β and then to day 12 differentiated C2C12. Results from another independent experiment are shown in Figure S1B. E-H, Western blotting analysis of KLF3 protein level in embryoid body (E), differentiated mESCs in media without LIF (F), differentiated ESCs induced by RA (G) and Dgcr8 knockout mESCs (H). For each sample, 20 μg protein was loaded. An empty lane was added between SM and other samples to avoid cross contamination. Ponceau S was used as input control and KLF3-Renilla will reflect both degradation and translational repression, while the difference between GAPDH-T2A-Renilla and KLF3-T2A-Renilla will only reflect translational repression since Renilla protein is detached from KLF3 or GAPDH during translation process. While KLF3-Renilla activity is ~2800-fold lower than GAPDH-Renilla, KLF3-T2A-Renilla activity is only 6-fold lower than GAPDH-T2A-Renilla ( Figure 2D). These data suggest that degradation may be the major pathway inhibiting KLF3 protein expression in mESCs, but translational repression could also play a role. Next we tested through which pathway KLF3 is degraded in ESCs. Proteins are usually degraded through the proteasome pathway or lysosome pathway, and these two pathways can be inhibited by MG-132 and F I G U R E 2 KLF3 protein is translated but degraded in mESCs. A, Ribosome profiling of mESCs. Up, monitoring of absorbance at 254 nm for 12 fractions collected in polysome fractionation assay; lower two panels are percentage of Gapdh and Klf3 in each fraction. For each gene, data were normalized to spike-in mRNA. B, Schematic representation of the structure of Klf3 gene in psiCheck-2 plasmid used for luciferase reporter assay. C, Luciferase reporter assay of mESCs transfected with Renilla, Klf3-Renilla and Klf3-T2A-Renilla. For each construct, data were normalized to firefly control and then to Renilla construct. Shown is mean ± SD, n = 3. The P value was calculated by unpaired one-way ANOVA with two-sided Dunnett's test compared with Klf3-Renilla. D, Luciferase reporter assay of mESCs transfected with Gapdh-Renilla, Klf3-Renilla, Gapdh-T2A-Renilla and Klf3-T2A-Renilla. Data were normalized to firefly control then to respective Gapdh construct. Shown is mean ± SD, n = 6. The P value was calculated by unpaired two-tailed Student's t test. E, Luciferase reporter assay of mESCs transfected with KLF3-Renilla and treated with MG-132, CQ and CHX. For each sample, data were normalized to firefly control and then to KLF3-Renilla treated with DMSO control. Shown is mean ± SD, n = 4. The P value was determined by unpaired one-way ANOVA with two-sided Dunnett's test compared with KLF3-Renilla treated with DMSO control. F, Luciferase reporter assay of mESCs transfected with KLF3-Renilla and KLF3-T2A-Renilla and treated with bortezomib and carfilzomib. For each sample, data were normalized to firefly control and then to KLF3-Renilla treated with DMSO control. Shown is mean ± SD, for KLF3-Renilla, n = 8, for KLF3-T2A-Renilla, n = 4. The P value was determined by unpaired two-way ANOVA with two-sided Dunnett's test compared with KLF3-Renilla treated with DMSO control chloroquine (CQ), respectively. Adding MG132 but not CQ led to ~3fold increase in KLF3-Renilla level ( Figure 2E). Moreover, when the protein synthesis was inhibited by the treatment of cycloheximide (CHX), KLF3-Renilla was further decreased. However, MG132 significantly rescued KLF3-Renilla level in the presence of CHX, supporting KLF3 being degraded by proteasome pathway ( Figure 2E).
In contrast, none of them increased Renilla activity in KLF3-T2A-Renilla construct ( Figure 2F). Together, these data show that KLF3 protein is translated but degraded through proteasome pathway.
However, our data do not exclude the role of other mechanisms such as translational repression in inhibiting KLF3 protein expression.

| The proline-rich sequence promotes the degradation of KLF3
To gain further insights on why KLF3 is degraded in mESCs, we predicted disorder structure in KLF3 through PONDR program. 40 Interesting, we found that N terminal of KLF3 was highly unstructured ( Figure 3A). Proline often causes disordered structure due to its constrained conformation. We noticed that there were 46 prolines in N terminal of KLF3 protein ( Figure S3). When these prolines were mutated to alanine, the proportion of disordered structure was largely reduced ( Figure 3B). We then experimentally tested the hypothesis that disordered structure in N terminal of KLF3 leads to its degradation in mESCs. Three lines of evidence supported the hypothesis. First, fusing a stable protein GFP to the N terminal of KLF3-Renilla significantly increased the Renilla activity ( Figure 3C); second, truncation of 150 amino acids or more at KLF3 N terminal significantly increased the expression of KLF3-Renilla ( Figure 3D); finally, mutation of all prolines to alanines in the N terminal also significantly increased the expression of KLF3-Renilla ( Figure 3E).
These data suggest that disordered structure of KLF3 at the N terminal made it vulnerable to degradation in mESCs.

| KLF3 protein is expressed in 8-cell stage mouse embryos
Next we checked whether KLF3 protein expression is dynamically regulated during early embryo development from zygote to blastocyst stage. Analysis of previously published single cell RNA-seq data [41][42][43][44] indicates that Klf3 mRNA was activated at 2-cell stage, and remained at high level until blastocyst stage ( Figure 4A). To F I G U R E 3 Intrinsically disordered structure promotes the degradation of KLF3 protein in mESCs. A-B, PONDR score of KLF3 (A) and proline-mutated KLF3 (B) predicted by PONDR program. C, Luciferase reporter assay of ESCs transfected with KLF3-Renilla, GFP-KLF3-Renilla and KLF3-T2A-Renilla. For each sample, data were normalized to firefly control and then to KLF3-Renilla. Shown is mean ± SD, n = 3. The P value was determined by unpaired one-way ANOVA with two-sided Dunnett's test compared with KLF3-Renilla. D, Luciferase reporter assay of mESCs transfected with N-terminal truncated KLF3-Renilla. Number of truncated amino acids at the N terminal is indicated. For each sample, data were normalized to firefly and then to full-length KLF3-Renilla. Shown is mean ± SD, n = 4. The P value was determined by unpaired one-way ANOVA with two-sided Dunnett's test compared with full-length KLF3-Renilla. E, Luciferase reporter assay of mESCs transfected with wild-type KLF3-Renilla and proline-mutated KLF3-Renilla. For each sample, data were normalized to firefly and then to Klf3-Renilla. Shown is mean ± SD, n = 8. The P value was determined by two-tailed Student's t test check KLF3 protein level, we performed immunofluorescence staining (IF) for KLF3 protein from fertilized egg to blastocyst stage.
Interestingly, we found that KLF3 protein is barely detectable before or after 8-cell stage ( Figure 4B,C). However, KLF3 protein was detected at high levels at 8-cell stage embryo ( Figure 4B,C). These data suggest that KLF3 protein is post-translationally regulated in preimplantation mouse embryos with highest expression at 8-cell stage.

| KLF3 protein is expressed in a small subset of mESCs
Previously, a rare population of cells in ESC culture was found to

| The transcriptome difference between KLF3positive and KLF3-negative mESCs
We then performed qPCR analysis for Klf3 and key pluripotency genes in KLF3-GFP-positive mESCs. Consistent with regulation at protein level, Klf3 mRNA was similar in GFP-positive and GFPnegative cells ( Figure 6A). Moreover, pluripotency genes including Nanog, Oct4, Esrrb, Rex1 and Sox2 were slightly repressed in GFPpositive cells ( Figure 6A). To fully characterize KLF3-GFP-positive cells, we performed RNA-seq. Overall, there were 862 and 1606 genes that are downregulated and upregulated in GFP-positive versus GFP-negative mESCs ( Figure 6B and Table S1), respectively.
These differentially regulated genes were enriched in lysosomal, protein processing and metabolic pathways ( Figure S4A,B). In addition, analysis of 2-cell specific genes indicates that these rare

| KLF3-GFP-positive ESCs upregulate 8-cellembryo activated genes
Since KLF3 protein is also highly expressed in 8-cell embryos, we then checked the similarity in transcriptome of 8-cell embryos and KLF3-positive mESCs. GSEA analysis showed that genes activated F I G U R E 4 KLF3 protein is highly expressed in 8-cell embryos. A, Relative expression level of Klf3 from zygote to blastocyst. For each stage, FPKM were normalized to blastocyst. Centre line, median. B, IF staining of KLF3 protein in mouse embryos from zygote to blastocyst stage. Commercial antibody anti-KLF3 from Abnova was used. Shown are representative images. Scale bars, 20 μm. C, Quantification of fluorescence intensity in nucleus. Shown is mean ± SD. Each dot represents one nucleus. n = 8 to 16 embryos for each stage from 4-to 8-cell stage 41 were significantly enriched in KLF3-GFPpositive mESCs ( Figures 6C and S5A,B). RT-qPCR analysis also confirmed the upregulation of selected 8-cell-embryo activated genes in KLF3-GFP-positive mESCs (Figures 6D and S5C). Taken together, these data show that KLF3-GFP expression marks a small subset of mESCs which upregulates 8-cell-embryo activated genes.

| Forced expression of KLF3 protein upregulates 8-cell-embryo activated genes
The specific expression of KLF3 protein in 8-cell embryos and the subset of mESCs suggest a potential function for KLF3 in these cells. To understand its impact on gene expression, we decided to overexpress KLF3 protein in mESCs. Expression of native Klf3 sequence did not lead to any detectable KLF3 protein in mESCs, consistent with the posttranslational regulation of KLF3 expression ( Figure 7A). The steady level of a protein depends on both production and degradation rate.
Since we lack means to specifically inhibit the degradation of KLF3, we reasoned that we may achieve higher KLF3 protein expression by increasing translation efficiency through codon optimization. Indeed, after codon optimization, we were able to detect trace amount of KLF3 protein in mESCs in a doxycycline-inducible fashion. The amount of KLF3 protein in mESCs overexpressing codon optimized Klf3 (oKlf3) was around 15% as in skeletal muscle ( Figure 7A). We noticed that overexpressed KLF3 protein migrated to a position of higher molecular weight than in skeletal muscle. Mass spectrometry analysis indicates that multiple serine of KLF3 was phosphorylated ( Figure S6A). Treatment with λ protein phosphatase (λpp) shifted KLF3 protein back to the same migration rate in mESCs as in skeletal muscle ( Figure S6B  Our study reveals that KLF3 protein is translated but significantly degraded in mESCs and C2C12 cells. However, the current data do not exclude other mechanisms inhibiting KLF3 protein expression, for example, translational repression. How KLF3 is degraded warrants further investigations. A potential explanation supported by our data is that the intrinsic disordered structure at the N terminal of KLF3 may cause it inherently sensitive to proteasome degradation. 48,49 How many other transcription factors are regulated in the same manner as KLF3 is another interesting question. Answering this question also requires the elucidation of exact regulatory pathway that leads to the repression of KLF3. Furthermore, maybe an equally interesting question is how KLF3 is stabilized in cell lineages of skeletal muscle and heart. A 'nanny protein' 50 may be co-expressed with KLF3 in these cells. Therefore, protein pull-down experiments in these cells will likely provide hints for how KLF3 is stabilized in certain cells. Finally, chromatin immunoprecipitation followed by high-throughput sequencing needs to be performed for KLF3 in 8-cell embryos and tissues such as heart and SM. These studies will provide insights not only for the physiologic role of KLF3 in these cells but also for why KLF3 has to be repressed in many other cells.

ACKNOWLEDGEMENTS
We would like to thank members of Wang laboratory for critical reading and discussion of the manuscript. We thank the mass spec-

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.

AUTH O R CO NTR I B UTI O N S
JH performed all experiments except these indicated below with help from other authors and bioinformatics analysis. XY initiated the project and performed experiments in Figures 1B,C

DATA AVA I L A B I L I T Y
All data generated or analysed during this study are included in the manuscript and its supplementary information files. RNA-seq data will be deposited in NCBI's Gene Expression Omnibus and the accession number will be provided before publication. All data that support the findings of this study are available from the corresponding authors upon request.