Klf4 glutamylation is required for cell reprogramming and early embryonic development in mice

Temporal and spatial-specific regulation of pluripotency networks is largely dependent on the precise modifications of core transcription factors. Misregulation of glutamylation is implicated in severe physiological abnormalities. However, how glutamylation regulates cell reprogramming and pluripotency networks remains elusive. Here we show that cytosolic carboxypeptidases 1 (CCP1) or CCP6 deficiency substantially promotes induced pluripotent cell (iPSC) induction and pluripotency of embryonic stem cells (ESCs). Klf4 polyglutamylation at Glu381 by tubulin tyrosine ligase-like 4 (TTLL4) and TTLL1 during cell reprogramming impedes its lysine 48-linked ubiquitination and sustains Klf4 stability. Klf4-E381A knockin mice display impaired blastocyst development and embryonic lethality. Deletion of TTLL4 or TTLL1 abrogates cell reprogramming and early embryogenesis. Thus, Klf4 polyglutamylation plays a critical role in the regulation of cell reprogramming and pluripotency maintenance.

R eprogramming resets differentiated somatic cells to a pluripotent state, which can be achieved by nuclear transfer, cell fusion, and transduction of transcription factors 1 . Somatic cells can be reprogrammed to induced pluripotent cells (iPSCs) by expressing pluripotency factors Oct4, Sox2, Klf4, and c-Myc (termed OSKM) 2,3 . The generation of iPSCs can be derived from patient tissues and has great potential for regenerative medicine and cell replacement therapies 4,5 . Several hurdles, including low frequency of iPSC induction and genomic instability of iPSCs, need to be solved prior to development of a safe iPSC technology. However, the molecular mechanisms underlying reprogramming still remain ill-defined. The temporal and spatial-specific regulation of pluripotency networks largely depends on precise modifications and interaction controls of the core transcriptional factors [6][7][8][9] . These reprogramming factors are highly modified post-transcriptionally at the levels of mRNA stability, translation and protein activity 7,10 .
Protein post-translational modifications (PTMs) such as phosphorylation, acetylation, glycosylation, and ubiquitination play important roles in the regulation of activities of target proteins by changing their chemical or structural properties 11,12 . Indepth quantitative and dynamic proteomic studies reveal that PTMs occur on core transcription factors during the process of pluripotency maintenance and reprogramming 7 . Transcriptional and DNA-binding activities of Oct4 and Sox2 are regulated by phosphorylation, which exert considerable effect on pluripotency maintenance and iPSC generation 7,13 . Acetylation of Sox2 is critical for pluripotency control by regulating its nuclear export and protein stability 14 . O-GlcNacylation directly regulates transcriptional activities of Oct4 and Sox2 in maintaining pluripotency and cell reprogramming 9,15 . Moreover, ubiquitination of Klf4 and Oct4 modulates their half-life and subsequent protein stability 16 , 17 . It has been reported that B cells treated with C/EBPα can be efficiently reprogrammed into iPSCs by OSKM induction through enhancing chromatin accessibility and Klf4 stability 18 . Therefore, PTMs of reprogramming factors play critical roles in determining the cell fate decision of stem cells.
Glutamylation, a unique PTM, adds glutamate side chains onto the γ-carboxyl groups of glutamic acid residues in the primary sequence of target proteins [19][20][21] . Polyglutamylation of tubulins, well-known targets of glutamylation, regulates the interaction between microtubules (MTs) and their partners 19 , modulating MT-related processes such as ciliary motility, neurite outgrowth and neurodegeneration [21][22][23][24] . Glutamylation is catalyzed by polyglutamylases, also called as tubulin tyrosine ligase-like (TTLL) enzymes 25,26 . Glutamylation is a reversible modification that can be hydrolyzed by a family of cytosolic carboxypeptidases (CCPs) 23 . Misregulation of glutamylation causes several physiological abnormalities. CCP1 deficiency causes hyperglutamylation of tubulins, resulting in Purkinje cell degeneration 23,27,28 . We recently demonstrated that CCP6 deficiency induces hyperglutamylation of Mad2, leading to underdevelopment of megakaryocytes and abnormal thrombocytosis 29 . In addition, mutations of TTLL polyglutamylases are also implicated in severe disorders such as retinal dystrophy 30 and pancreatic oncogenesis 31 . However, how glutamylation regulates cell reprogramming and pluripotency maintenance remains elusive.
The Kruppel-like factor (Klf4), together with other three Yamanaka pluripotency factors, is able to reprogram adult fibroblasts into iPSCs 2,3,32 , which initiates somatic gene suppression in an early phase and pluripotency gene activation in a late phase during reprogramming 33,34 . Klf4 is highly expressed in mouse embryonic stem cells (ESCs) and rapidly downregulated in the early stage of differentiation 35 . Klf4 can be phosphorylated by ERKs, whose phosphorylation recruits βTrCPs (components of a ubiquitin E3 ligase) to ubiquitinate Klf4 for its degradation 16 . However, the precise modification crosstalk of Klf4 in reprogramming has not clearly defined yet.
In this study, we show that CCP1 or CCP6 deficiency substantially promotes iPSC generation. Klf4 is polyglutamylated by TTLL4 or TTLL1 that impedes its lysine (K) 48-linked ubiquitination to maintain Klf4 stability. Deletion of TTLL4 or TTLL1 impairs cell reprogramming and early embryogenesis.

Results
CCP6 or CCP1 deficiency promotes somatic cell reprogramming. We previously demonstrated that Ccp6 (official gene name Agbl4, referred to here as Ccp6)-deficient mice exhibit underdevelopment of megakaryocytes and abnormal thrombocytosis 29 . We next wanted to further explore whether glutamylation modifications were involved in the regulation of cell reprogramming. We noticed that CCP6 deficiency caused higher litter size at birth (Supplementary Figure 1a), whereas Ccp6-deficient mice displayed similar sperm and ovulation numbers compared with littermate control mice, suggesting CCP6 could be implicated in the modulation of cellular reprogramming. To determine whether glutamylation is involved in somatic cell reprogramming, we first generated Ccp1 −/− (official gene name Agtpbp1, referred to here as Ccp1), or Ccp6 −/− mouse embryonic fibroblasts (MEFs) (Fig. 1a), and transduced Yamanaka factors OSKM for iPSC induction assays. We noticed that Ccp1 −/− and Ccp6 −/− MEFs produced more iPSC colonies compared with WT MEFs (Fig. 1b). These iPSCs generated from Ccp1 −/− or Ccp6 −/− MEFs expressed elevated pluripotent stem cell surface marker SSEA-1 and increased pluripotent genes without partial differentiation (Fig. 1c, and Supplementary Figure 1b, c). These reprogrammed iPSCs expressed similar levels of pluripotent genes compared to ESCs, validating the reprogramming system was efficient (Supplementary Figure 1b). Moreover, the iPSCs derived from Ccp1 −/− or Ccp6 −/− MEFs were capable of forming teratomas containing cells of all three germ layers and displayed faster growth rates (Fig. 1d). Importantly, Ccp1 and Ccp6 double knockout (DKO) MEFs showed higher reprogramming efficiency (Fig. 1b), as well as pluripotent gene expression than Ccp1 −/− or Ccp6 −/− MEFs alone (Supplementary Figure 1b). In addition, restoration of CCP1 or CCP6 into Ccp1 −/− or Ccp6 −/− MEFs still went back to a low frequency of iPSC generation ( Fig. 1b and Supplementary Figure 1b). Of note, deficiency of CCP1 or CCP6 in feeder-free iPSCs really exhibited elevated pluripotent gene expression, whereas deficiency of CCP1 or CCP6 in MEFs alone did not affect pluripotent gene expression (Supplementary Figure 1d). Overall, these results indicate that deficiency of CCP1 or CCP6 promotes somatic cell reprogramming.
To further determine the physiological role of CCP1 and CCP6 in the process of reprogramming, we silenced CCP1 and CCP6 expression in MEFs with transfection of OSKM, and found CCP1 and CCP6 depletion enhanced alkaline phosphatase (AP)-positive iPSC colony formation and pluripotent gene expression (Supplementary Figure 1e-g). By contrast, overexpression of CCP1 and CCP6 impaired iPSC colony formation as well as downregulated pluripotent gene expression (Supplementary Figure 1e-g). Of note, depletion and overexpression of CCP1 and CCP6 in MEFs did not affect growth rates of MEFs (Supplementary Figure 1h). We also treated MEFs with CCP family protein agonist CoCl 2 36 and inhibitor phenanthroline 23 after OSKM induction. Consistently, the agonist CoCl 2 abrogated iPSC formation, whereas the inhibitor phenanthroline remarkably enhanced iPSC generation ( Fig. 1e and Supplementary Figure 1i). These data further confirm that loss of CCP1 or CCP6 virtually enhances cell reprogramming.
Fertilization initiates cellular reprogramming in zygote and subsequent blastocyst development, which also requires the establishment of pluripotency 37,38 . Since homozygous Ccp1deficient mice are male sterile 27 , we then assessed the effect of Ccp6 deficiency on blastocyst development. We isolated 2-cellstage embryos from Ccp6 +/+ and Ccp6 −/− pregnant mice and cultured the embryos ex vivo to allow following development. CCP6 deficiency substantially promoted blastocyst development at embryo day (E) 3.5 compared to WT mice (Fig. 1f). In parallel, expression levels of pluripotent genes were dramatically increased in Ccp6 −/− embryos ( Supplementary Figure 1j-k). To further determine the physiological relevance of glutamylation, we isolated 2-cell-stage embryos from C57BL6/J mice and cultured them ex vivo with CoCl 2 and phenanthroline treatment. We observed that the agonist CoCl 2 treatment overtly inhibited embryonic development, while the inhibitor phenanthroline treatment remarkably enhanced blastocyst development (Fig. 1g). Taken together, glutamylation regulations are implicated in the early embryonic development.
Glutamylation is required for pluripotency maintenance. We further tested whether CCP1 or CCP6 regulates self-renewal and pluripotency maintenance in ES cells. We silenced CCP1 or CCP6  in mouse ESC D3 cells, and noticed that CCP1-or CCP6depleted D3 cells exhibited increased colony numbers with strong AP-positive staining and upregulated pluripotent gene expression compared with control shRNA (shCtrl)-treated cells ( Fig. 2a and Supplementary Figure 2a). Double knockdown of CCP1 with CCP6 induced much more colony numbers than that of CCP1 or CCP6 knockdown alone. By contrast, ectopic expression of CCP1 or CCP6 substantially declined AP-positive colony numbers and downregulated pluripotent gene expression ( Fig. 2b and Supplementary Figure 2b). In parallel, CCP1 or CCP6 overexpressing ES R1 cells displayed similar observations ( Supplementary Figure 2c). Of note, we noticed that depletion of CCP1 or CCP6 promoted ESC proliferation (Supplementary Figure 2d), whereas overexpression of CCP1 or CCP6 suppressed ESC proliferation (Supplementary Figure 2d). We then measured the glutamylation dynamic changes during the process of early embryonic development. GT335 antibody specifically recognizes the branching point of glutamate side chains and detects all glutamylated forms of target proteins 39 . We observed that GT335 staining signals were elevated during early embryonic development and GT335 staining signals mainly resided in the nucleus of 2-cell and 4-8-cell-stage embryos (Fig. 2c). TTLL members add glutamylation modifications and CCP members remove it of their substrates 19,23 . We observed that polyglutamylases TTLL1 and TTLL4 were highly expressed undergoing early embryonic development (Supplementary Figure 2e). However, both CCP1 and CCP6 were downregulated in 8-cell-stage embryos (Supplementary Figure 2f). We also performed transcriptome profile assays for CCP6depleted and shCtrl-treated ESCs. We noticed that CCP6 knockdown in ESCs caused upregulation of pluripotency transcriptional network (Supplementary Figure 2g). In addition, we analyzed RNAseq data set GSE45352 40 for OSKM-induced reprogramming. We found that Ccp1 was downregulated and Ttll4 was upregulated over doxycycline-induced OSKM expression (Supplementary Figure 2h). In addition, from RNAseq data set GSE52396 41 Figure 2l). Moreover, CCP1-or CCP6-depleted human iPSCs were able to form teratomas with three germ layer differentiation potential and grew faster compared with shCtrl-treated iPSCs (Fig. 2f). Whereas CCP1 or CCP6 overexpressing human iPSCs impaired AP-positive colony generation and lost teratoma formation (Fig. 2f). These data suggest that glutamylation also regulates ESC and iPSC pluripotency that is evolutionally conserved in human reprogramming.
Klf4 undergoes polyglutamylation in cell reprogramming. We then measured the glutamylation dynamic changes during the process of somatic cell reprogramming. We observed that GT335 staining signals were elevated over OSKM transduction (Fig. 3a,b). Furthermore, GT335 staining signals mainly resided in the nucleus of MEFs (Fig. 3a). We observed that polyglutamylases TTLL1 and TTLL4 were highly expressed during OSKM-induced reprogramming of MEFs (Supplementary  Figure 3c). Thus, we proposed that TTLL1 and TTLL4 may mediate substrate glutamylation that is required for iPSC induction.
We next wanted to identify candidate substrates for cell reprogramming. We noticed that Ccp6 −/− MEF lysates with OSKM transduction displayed an around 60 kDa strong differential band compared to those of Ccp6 +/+ MEFs (Supplementary Figure 3d). We then generated an enzymatically inactive mutant of CCP6 (CCP6mut) via H230S and E233Q mutations 29 , and performed affinity chromatography with CCP6mut-immobilized Affi-gel10 resin. Interestingly, the around 60 kDa band was identified to be Klf4 ( Fig. 3c and Supplementary Figure 3e), a novel candidate substrate for CCP6 during cell reprogramming.
We next monitored the colocalization of Klf4 and GT335 in the process of reprogramming with immunofluorescence staining. Colocalization of GT335 with Klf4 staining was observed in the nucleus of MEFs after OSKM induction (Supplementary Figure 4f). Consistently, Ccp1 −/− or Ccp6 −/− MEFs with OSKM transduction showed strong glutamylation signals ( Fig. 3f and Supplementary Figure 4g). More importantly, Klf4 polyglutamylation also appeared in the nucleus of early embryos and enhanced after the 4-to 8-cell stage (Supplementary Figure 4h). As expected, hyperglutamylation of Klf4 also appeared in Ccp6deficient embryos (Fig. 4g). Consistently, Klf4 polyglutamylation also appeared in mouse ESCs, and declined during the process of retinoic acid (RA)-induced differentiation ( Supplementary Figure 4i). We also assessed whether the reprogramming process induced by non-OSKM factors affected endogenous Klf4 polyglutamylation. We transduced factors Sall4, Nanog, Esrrb, and Lin28 (termed SNEL) into Ccp6 −/− MEFs for iPSC induction. In parallel, SNEL factor transduction in Ccp6 −/− MEFs really enhanced hyperglutamylation of Klf4 as well, and consequently generated much more iPSC colonies compared to that of Ccp6 +/+ MEFs (Supplementary Figure 4j). Taken together, the polyglutamylation modification of Klf4 plays a critical role in the modulation of cell reprogramming and pluripotency.
Nanog and Esrrb are downstream target genes of Klf4 during the iPSC reprogramming 47,48 . DNase I digestion assay showed that CCP6 deficiency augmented chromatin accessibility to DNase I digestion at the promoters of Nanog and Esrrb (Fig. 5c), indicating an open status of chromatin region of these genes. Rescue of Klf4-E381A into Klf4-silenced MEFs reduced DNase I accessibility of these two gene promoters (Fig. 5c). By contrast, restoration of Klf4-K232R into Klf4-silenced MEFs increased DNase I accessibility of these two promoters. These results suggest that Klf4 stabilization by polyglutamylation initiates transcriptional activation of its downstream target genes.
Actually, Klf4-wt glutamylation was catalyzed by TTLL4, which remarkably reduced its ubiquitination modification (Fig. 5d). By contrast, Klf4-E381A overexpression abrogated the polyglutamylation of Klf4 and augmented its ubiquitination signals (Fig. 5d). Expectedly, Klf4-K232R overexpression declined its ubiquitination modification (Fig. 5d). Similar results were achieved by treatment with the counterparts of TTLL1 overexpression (Supplementary Figure 5o). We next wanted to exclude the possibility that mutants of Klf4 could affect its target gene expression by changing their DNA-binding capacities. Fig. 2 Glutamylation is required for the maintenance of mouse and human pluripotency. a Mouse ESC D3 cells were transfected with scrambled shRNA (shCtrl), shCCP1, shCCP6 or CCP1 and CCP6 overexpression (oe) plasmids as indicated and cultured in mouse ESC media. After 5 days, pluripotency was analyzed by AP staining. Colony numbers for undifferentiated, mixed or differentiated clones were calculated as means ± S.D. **P < 0.01. Scale bar, 100 μm. n = 4. b D3 cells were transfected with the indicated plasmids and assessed AP staining as in a. **P < 0.01. Scale bar, 50 μm.n = 6. c Embryos at the indicated stages were isolated and stained with GT335 antibody and PI, and visualized by confocal microscopy. Scale bar, 20 μm. Percentages of cells with GT335 signal localized in the nucleus were counted as means ± S.D. **P < 0.01. For 1-cell-stage embryos, n = 48. For 2-cell-stage embryos, n = 56. For 4-to 8-cell stage embryos, n = 62. d Human ESC H9 cells were infected with lentivirus expressing the indicated shRNAs and cultured in human ESC media for 3 weeks. Pluripotency was analyzed by AP staining. Scale bar, 100 μm. AP-positive colony numbers per well were calculated as means ± S.D. **P < 0.01. n = 5. e Depletion of CCP1 or CCP6 in human iPSCs increases AP + colony formation. Human iPSCs were infected with lentivirus expressing the indicated shRNAs and cultured in hiPSC media for 3 weeks, followed by AP staining. Scale bar, 50 μm. AP-positive colony numbers per well were calculated as means ± S.D. **P < 0.01. n = 5. f CCP1 or CCP6 depletion in human iPSCs promotes teratomas formation. Human iPSCs infected with the indicated lentivirus were injected subcutaneously into NOD/SCID mice. Six weeks later, teratomas were collected for H&E staining. Red arrow denotes teratomas in upper panels. Black arrow, endoderm; red arrow, mesoderm; red arrowhead, ectoderm. Scale bar, 200 μm. Total 6 teratomas from 6 mice were analyzed pre condition (n = 6). The teratomas were from 2 iPSC cell lines. Student's t test was used as statistical analysis. ns no significance Binding capacity of Nanog promoter with Klf4-wt, Klf4-E381A, and Klf4-K232R was measured by an electrical mobility shift assay (EMSA). We observed that Klf4-E381A and Klf4-K232R mutants showed comparable binding affinities to Klf4-wt protein (Fig. 5e). Unlabeled Nanog probes could competitively bind to the Nanog promoter. These data suggest that Klf4 mutation per se has no direct activation effect on Nanog gene. Therefore, polyglutamylation-mediated Klf4 stability directly regulates the transcriptional activation of its downstream target genes.
To further verify the role of Klf4 polyglutamylation in the reprogramming and pluripotency regulation in vivo, we generated Klf4-E381A mutant knockin (Klf4 E381A KI) mice by CRISPR/Cas9 editing technology. As expected, MEFs isolated from Klf4 E381A KI mice really abrogated iPSC formation by in vitro OSKM transduction (Fig. 6a). Of note, Klf4 E381A KI mice impaired blastocyst development at E3.5 (Fig. 6b). In addition, Klf4 with cell fate markers (Nanog for epiblast, Gata6 for primitive endoderm, and Cdx2 for trophectoderm) in early embryos from Klf4 E381A KI mice was substantially reduced and consequently underwent apoptosis (Fig. 6c,d). Finally, Klf4 E381A homozygous pups were embryonically lethal (Fig. 6e). Functions of Klf2 and Klf5 are redundant with Klf4 in ES cells 35 and iPSCs 49 . Of interest, we found that Klf4-E381A was dimerized with Klf5 in ESCs, while Klf4-wt protein did not (Supplementary Figure 5p). We propose that the dimerization of Klf4-E381A with Klf5 could inactivate the Klf5 function to cause preimplantation lethality in Klf4 E381A KI mice. It has been reported that a glutamylation site can shift to the next available glutamate residue when the main modification site is mutated in tubulin 50 . However, Klf4 E381A mutant knockin abrogated the Klf4 function in vivo, excluding the possibility of modification site shift for Klf4 glutamylation. Taken together, Klf4 polyglutamylation at Glu381 is required for cell reprogramming and early embryonic development.
TTLL1 or TTLL4 deletion impairs embryonic development. To further explore the physiological role of TTLL1 and TTLL4 in the process of reprogramming, TTLL1 and TTLL4 were silenced in MEFs for iPSC formation assays. Depletion of TTLL1 or TTLL4 dramatically reduced iPSC colony numbers as well as expression levels of pluripotent genes ( Fig. 7a and Supplementary Figure 6a, b). Double knockdown of TTLL1 and TTLL4 displayed a synergistic inhibitory effect. Overexpression of TTLL1 or TTLL4 in accordance with silenced MEFs restored iPSC colonies comparable to shCtrltreated MEFs (Fig. 7a). TTLL1 or TTLL4 depletion in OSKMtransduced MEFs substantially declined Klf4 glutamylation signals (Supplementary Figure 6c). By contrast, overexpression of TTLL1 or TTLL4 in accordance with silenced MEFs rescued Klf4 glutamylation signals (Supplementary Figure 6c), which in turn promoted iPSC colony formation efficiency (Supplementary Figure 6d, e). Consequently, TTLL1 or TTLL4 depletion promoted Klf4 degradation (Fig. 7b). In contrast, overexpression of TTLL1 and TTLL4 sustained Klf4 stability. Finally, Ttll4-deficient MEFs impaired iPSC formation (Supplementary Figure 6f). These data indicate that TTLL1-or TTLL4-mediated Klf4 glutamylation is required for somatic cell reprogramming. We next analyzed the role of TTLL1 and TTLL4 in ESC pluripotency maintenance. We silenced TTLL1 or TTLL4 expression in mouse ESC D3 cells. As expected, TTLL1-or TTLL4-depleted ESCs exhibited decreased numbers of colonies with AP-positive staining and downregulated pluripotent gene expression compared with shCtrl-treated cells (Fig. 7c,d). Double knockdown of TTLL1 with TTLL4 synergetically lowered APpositive colony numbers and pluripotent gene expression in ESCs. Additionally, restoration of TTLL1 or TTLL4 into its corresponding TTLL1-and TTLL4-silenced cells rescued ESC pluripotency and pluripotent gene expression (Fig. 7c,d)  Reprogramming efficiency was assayed by Nanog staining after dox removal. Nanog-positive colony numbers per 10 4 cells were calculated and shown as means ± S.D. **P < 0.01,. n = 4. b D3 ESC cells were transfected with indicated plasmids, followed by AP staining. Scale bar, 50 μm. AP-positive colony numbers per well were calculated as means ± S.D. **P < 0.01. n = 5. c SSEA-1 + ES cells were isolated and an equal amount of cells were lyzed for nuclei extraction, followed by DNase I digestion. Total DNA was extracted and quantitated by qPCR with Nanog or Esrrb promoter-specific primers. n = 5. d Indicated plasmids along with TTLL4 were overexpressed into Klf4-silenced MEFs after OSKM transfection for 2 days, and cells were collected and immunoprecipitated with anti-Klf4 antibody followed by detection with anti-poly-Ub antibody. e Indicated proteins were incubated with labeled and unlabeled probes against Nanog promoter, followed by EMSA assay. Student's t test was used as statistical analysis. ns, no significance NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03008-2 ARTICLE NATURE COMMUNICATIONS | (2018) 9:1261 | DOI: 10.1038/s41467-018-03008-2 | www.nature.com/naturecommunications contrast, enforced overexpression of TTLL1 or TTLL4 increased ESC self-renewal as well as upregulated pluripotent gene expression (Fig. 7e,f). Similar results were obtained in TTLL1or TTLL4-depleted human ESC H9 cells (Fig. 7g,h).
To further test the role of TTLL1 and TTLL4 in early embryogenesis, we also generated TTLL1 knockout (KO) mice via a CRISPR-Cas9 approach 45 (Supplementary Figure 7a, b). We isolated fertilized eggs from WT, Ttll1 -/-, and Ttll4 -/pregnant mice to assess ex vivo development of embryos. Both TTLL1 and TTLL4 were deleted in embryos by immunofluorescence staining (Fig. 8a). TTLL1 and TTLL4 deficiency suppressed blastocyst development (Fig. 8b), leading to decreased blastocoel areas, disordered lineage marker expression, and increased apoptotic cells of early embryos (Supplementary Figure 7c-f). Finally, both Ttll4-deficient pups and Ttll1-deficient pups were born in extremely lower numbers than those that would be predicted by Mendelian inheritance after heterozygotes crossing (Fig. 8c and Supplementary Figure 7g). Altogether, TTLL1 and TTLL4 play a critical role in the regulation of iPSC formation and early embryonic development.

Discussion
Transcription factors-mediated reprogramming was sufficient to reset terminally differentiated cells into induced pluripotent stem cells (iPSCs) 2 Fig. 6 A glutamylation-defective mutant of Klf4 impairs ESC pluripotency. a Klf4 E381A knockin mice were generated by a CRISPR/Cas9 approach. MEFs were isolated and induced for iPS formation as in Fig. 5a. Scale bar, 100 μm. AP-positive colony numbers per 10 4 cells were calculated and shown as means ± S.D. **P < 0.01, n = 4. b Fertilized eggs at E3.5 embryonic stage were isolated from pregnant mice after heterozygotes crossing. Embryo genotype was confirmed by both GT335 immunostaining and PCR of genomic DNA. Scale bar, 100 μm. Distribution of embryonic stage at E3.5 was counted as means ± S.D, **P < 0.01. For WT embryos, n = 101. For Klf4 E381A embryos, n = 87. c Indicated preimplantation embryos from WT and Klf4 E381A knockin mice were stained with cell fate markers (Nanog for epiblast (EPI), Gata6 for primitive endoderm (PrE), and Cdx2 for trophectoderm (TE)) together with Klf4. n = 34 for WT embryos and n = 28 for Klf4 E381A embryos (anti-Nanog staining), n = 41 for WT embryos and n = 25 for Klf4 E381A embryos (anti-Gata6 staining), and n = 31 for WT embryos and n = 29 for Klf4 E381A embryos (anti-Cdx2 staining). d Indicated preimplantation embryos were stained with cell fate markers as well as cleaved caspase 3 (cl-Casp3). The percentage of positive cell in each germ layer was counted as means ± S.D, **P < 0.01. Scale bar, 20 μm. n = 40 for WT embryos and n = 31 for Klf4 E381A embryos (anti-Nanog staining), n = 37 for WT embryos and n = 30 for Klf4 E381A embryos (anti-Gata6 staining), and n = 37 for WT embryos and n = 25 for Klf4 E381A embryos (anti-Cdx2 staining). e Klf4 E381A pups were genotyped after heterozygotes crossing. Student's t test was used as statistical analysis  Fig. 7 Deletion of TTLL1 or TTLL4 impairs iPSC induction and ESC pluripotency. a Depletion of TTLL1 or TTLL4 impairs iPSC formation. TTLL1 or TTLL4 was silenced in 4F2A MEFs as well as rescued with TTLL1 or TTLL4 into their silenced MEFs, followed by dox treatment for iPSC formation and stained with anti-Nanog antibody as in Fig. 1e. Scale bar, 50 μm. Nanog-positive colony numbers per 10 4 cells were calculated and shown as means ± S.D. **P < 0.01. n = 5. b Immunoblotting of Klf4 in above-treated MEFs as in a. Fold changes of relative expression of indicated proteins compared with β-actin were caculated as means ± S.D. The data represent four independent experiments. c TTLL1 or TTLL4 depletion enhances the pluripotency of mouse ESCs. D3 cells were transfected with scrambled shRNA (shCtrl), shTTLL1, shTTLL4, or TTLL1 and TTLL4 overexpression (oe) plasmids and cultured in mouse ESC media. After 5 days, pluripotency was analyzed by AP staining. Colony numbers for undifferentiated, mixed, or differentiated clones were calculated as means ± S.D. n = 5. Scale bar, 100 μm. d D3 cells were transfected with indicated plasmids. mRNA levels of the indicated genes were analyzed by real-time qPCR. Relative gene expression fold changes were counted as means ± S.D. **P < 0.01. n = 5. Primer pairs are shown in Supplementary Table 1. e Indicated plasmids were transfected into mouse D3 cells as in c. Colony numbers for undifferentiated, mixed, or differentiated clones were calculated as means ± S. D. n = 5. Scale bar, 100 μm. f mRNAs levels of the indicated genes were analyzed by real-time qPCR as in d. Relative gene expression fold changes were counted as means ± S.D. **P < 0.01. n = 5. g TTLL1 or TTLL4 depletion in human ESC H9 cells was confirmed by real-time qPCR. n = 5. h TTLL1 or TTLL4 depletion in human ESC H9 cells increases AP + colony formation. Human H9 cells were infected with lentivirus expressing the indicated shRNAs and cultured in human ES media for 3 weeks. ESC pluripotency was analyzed by AP staining. AP + colony numbers were calculated as means ± S.D. n = 4. Scale bar, 100 μm. Student's t test was used as statistical analysis. oe overexpression reliable and beneficial for treatment of human-related diseases 52 .
In this study, we show that CCP1 or CCP6 deficiency substantially promotes iPSC generation and sustains ESC pluripotency. During the process of reprogramming, the core pluripotency factor Klf4 is polyglutamylated by TTLL4 or TTLL1 that impedes its K48-linked ubiquitination to maintain Klf4 stability (Fig. 8d). TTLL1-or TTLL4-mediated polyglutamylation of Klf4 is required for activation of its downstream pluripotency genes leading to somatic reprogramming, which maintains ESC pluripotency as well as drives embryogenesis. Deletion of TTLL4 or TTLL1 impairs cell reprogramming and early embryogenesis. Therefore, Klf4 polyglutamylation plays a critical role in the regulation of cell reprogramming and pluripotency maintenance. Protein levels of Oct4 and Sox2 are precisely modulated in order to sustain self-renewal and pluripotency of ESCs 53,54 . A moderate increase of Sox2 causes differentiation of ESCs primarily into neural ectodermal cells, mesodermal, and trophectoderm-like cells 53 , while decreased expression of Sox2 induces differentiation of ESCs into trophectoderm-like cells 55 . Similarly, Klf4 activity is also tightly regulated by PTMs. For example, phosphorylation of Klf4 by ERKs at Ser123 leads to inhibition of its transcriptional activity 16 . In addition, Klf4 also undergoes sumoylation that enhances Klf4 transactivation activity 56 . However, it is unknown whether core transcription factors undergo glutamylation in cell reprogramming and ESC pluripotency. The well-known substrates of polyglutamylation are tubulins and nucleosome assembly proteins 43,57,58 . A recent study delineates a structural microtubule recognition basis by catalysis with TTLL7 19 . TTLLs have different expression patterns in diverse tissues and their functions are not entirely redundant 59 . We recently reported that TTLL4 and TTLL6 are most highly expressed in megakaryocytes 29 , both of which catalyze polyglutamylation of Mad2 to modulate megakaryocyte maturation. In this study, we demonstrate that TTLL1 and TTLL4 are constitutively elevated in ESCs and early embryos, and dramatically upregulated in iPSCs during cell reprogramming. Both TTLL1 and TTLL4 can catalyze polyglutamylation of Klf4, which impedes its K48-linked ubiquitination to maintain Klf4 stability. We showed that both Ttll4-deficient pups and Ttll1deficient pups were born in extremely lower numbers than those that would be predicted by Mendelian inheritance after heterozygotes crossing. These observations suggest that TTLL1 and TTLL4 might synergistically modify the same glutamate on Klf4 to regulate early embryonic development. However, the different biochemical activities of TTLL family enzymes in the During the process of reprogramming and pluripotency maintenance, the core pluripotency factor Klf4 is polyglutamylated by TTLL4 or TTLL1 that impedes its K48-linked ubiquitination to maintain Klf4 stability. Deletion of TTLL4 or TTLL1 impairs cell reprogramming and early embryogenesis. Klf4 glutamylation is hydrolyzed by CCP1 or CCP6. CCP1 or CCP6 deficiency substantially promotes iPSC generation and sustains ESC pluripotency. Glutamylation of Klf4 is required for activation of its downstream pluripotency genes leading to somatic reprogramming, which maintains ESC pluripotency as well as drives embryogenesis regulation of cell reprogramming still need to be further investigated.
Protein glutamylation is a reversible modification, whose deglutamylation is hydrolyzed by a family of cytosolic carboxypeptidases (CCPs) 23,60 . CCP family members harbor enzymatic specificities to carry out deglutamylation. Of these CCP members, CCP1, CCP4, and CCP6 remove the shortening of penultimate polyglutamate chains of α-tubulin, while CCP5 specifically hydrolyzes the branching site glutamate 23 . We previously showed that Mad2 polyglutamylation can be hydrolyzed by CCP6 but not CCP1 23 . We provide the example that CCP1 and CCP6 harbor nonredundant roles in physiological conditions. Consequently, deficiency of CCP1 or CCP6 exhibits different abnormalities 27,29 . Herein, we show that both CCP1 and CCP6 are highly expressed in differentiated cells and remarkably declined in ESCs and iPSCs. Both CCP1 and CCP6 can hydrolyze the glutamylation of Klf4 during cell reprogramming and ESC maintenance. CCP1 and CCP6 could synergistically shorten longer glutamate chains of glutamyated Klf4, while CCP5 may be needed to deglutamylate Klf4 at a branching glutamate site. Alternatively, CCP1 and CCP6 could also remove a branched glutamate of glutamyated Klf4, whose biochemical activities may be different from tubulins 23 .
Glutamylation is highly conserved in all metazoans and protists, performing the assembly and function of cilia and flagella 58 . For example, TTLL7, the most abundantly expressed in the mammalian nervous system, is conserved from acorn worm to primates, where it modulates neurite outgrowth and localization of dendritic MAPs. In addition, CCP and TTLL members exert their functions by enzymatic activities against their respective substrates, and their enzymatic activities are also modulated by post-translational modifications and other regulatory factors. Here, we show that Klf4 polyglutamylation mediated by TTLL1 and TTLL4 enhances mouse and human reprogramming and pluripotency. In parallel, the CCP inhibitor phenanthroline dramatically promotes somatic cell reprogramming and embryonic development. Thereby, we strongly believe that it is necessary to develop specific inhibitors or agonists for each individual polyglutamylase and cytosolic carboxypeptidase. Manipulating polyglutamylation profiles by using these compounds, we may potentially improve iPSC efficiency for future clinical applications. In sum, Klf4 glutamylation plays a critical role in the regulation of cell reprogramming and pluripotency maintenance. Our findings provide mechanistic insights into how glutamylation modulates cell fate determination.
Methods expression was induced by doxycycline treatment. After 3 weeks, reprogramming efficiency was assayed by Nanog staining after dox removal.
Embryo collection and culture. A total of 5 IU of pregnant mare serum gonadotropin (PMSG) was intraperitoneally injected for superovulation and 5 IU of human chorionic gonadotropin (hCG) was injected 48 h later. Embryos were flushed with M2 medium (Millipore) from oviducts or uteri at the indicated stages and cultured ex vivo in KSOM medium (Millipore) supplied with 1 mg/ml BSA at 37 ℃ with 5% CO 2 . Concentrations of 10 μM CoCl 2 or 1 μM phenanthroline were used for embryo culture. For each group, more than 100 typical embryos were observed.
In vivo assay of teratomas. Human iPSCs were generated by transducing OSKM factors into human urothelial cells. Urine samples were recruited from three healthy donors with informed consent based on approval by the Institutional Ethical Committees at the Institute of Biophysics, Chinese Academy of Sciences. Ccp1 -/or Ccp6 -/mouse iPS cells or human iPS cells infected with shCCP1, shCCP6, oeCCP1, and oeCCP6 lentiviruses were collected, washed twice with PBS, and then subcutaneously injected into the bilateral inguens of male NOD/SCID mice (2 × 10 6 cells per injection). After 4 weeks for mouse iPSCs and 6 weeks for human iPSCs, mice were killed. Tumors were weighed, fixed in 4% paraformaldehyde, and sectioned for staining with hematoxylin and eosin (H&E) 61 . The maximal sizes of tumors permitted were 2500 mm 3 according to the ethical approval by the Institutional Animal Care and Use Committees at the Institute of Biophysics, Chinese Academy of Sciences.
Immunofluorescence assay. MEFs were fixed with 4% paraformaldehyde (PFA, Sigma-Aldrich) for 20 min after retrovirus infection. Embryos were collected and treated with acidic tyrode buffer for 1 min to remove zona pellucida followed by 4% PFA fixation. Cells were permeabilized with 1% Triton-X 100 in PBS for 20 min, and blocked with 10% donkey serum. Cells were then incubated with indicated primary antibodies at 4 ℃ overnight followed by staining with Alexa405-, Alexa488-, Alexa594-, and Alexa649-conjugated secondary antibodies. Nuclei were stained with DAPI or PI. Images were obtained with Olympus FV1000 laserscanning confocal microscopy (Olympus, Japan). The software ImageJ was used for colocalization analysis 61 .
Mass spectrometry. MBP-CCP6-wt and MBP-CCP6-mut 23,29 proteins were expressed in E. coli and purified using the amylose resin (New England BioLabs, Ipswich, USA) according to the manufacturer's instruction. Proteins were immobilized with Affi-gel10 resin to go through OSKM-transduced MEF lysates for affinity chromatography. Eluted fractions were visualized by SDS-PAGE followed by silver staining. Differential bands in SDS-PAGE gels were trypsinized for mass spectrometry with LTQ Orbitrap XL (Thermo Finnigan) 29 .
In vitro glutamylation assay. A total of 293 T cells were transfected with expression plasmids of TTLL4, CCP1, and CCP6 for 48 h. Cells were collected and lyzed with PBS-0.2% NP40. Supernatants were incubated with recombinant GST-Klf4 protein at 37 ℃ for 2 h. GST-Klf4 was pulled down with Glutathione Sepharose 4B beads, followed by immunoblotting 29 . Uncropped scans of results were shown as Supplementary Figure 8.
Gene expression assay by real-time qPCR. iPSCs and ESCs were sorted by flow cytometry after SSEA-1 staining. A total of 5 × 10 5 cells were lyzed in each sample for the analysis below. Total RNAs were extracted with the RNA miniprep kit (LCsciences, Houston, USA) according to the manufacturer's manual. M-MLV reverse transcriptase (Promega, Madison, USA) was used to synthesize cDNA. Quantitative PCR analysis and data collection were performed on the ABI 7300 qPCR system using the primer pairs listed in Supplementary Table 1. Quantitation was normalized to an endogenous β-actin gene 64 .
Microarray analysis. Mouse ES cell line D3 cells were transfected with the shCCP6 or shCtrl plasmid and cultured in mouse ESC media for 3 days. RNAs from mouse shCCP6 ESCs and shCtrl ESCs were prepared for Affimatrix microarray analysis according to the manufacturer's instruction. Briefly, total RNA was extracted by Trizol reagent. Double-strand cDNA was synthesized by an Invitrogen SuperScript cDNA synthesis kit and labeled according to the Affimatrix protocol. Microarrays were hybridized, washed and scanned according to the manufacturer's instruction. Differentially expressed genes were identified as fold change cutoff >2.0, FDR<0.05.
Electrophoretic mobility shift assay (EMSA). Biotin-labeled double-strand probes were generated by annealing complementary single-stranded oligonucleotides. The probe sequence used for detection was from Nanog promoter region: 5′-GCCGCCTGGGTGCCTGGGAGAATAGGGGGTGGGTAGGGTAGGAGG CTTG-3′. Recombinant GST-Klf4 and mutants were expressed by E.coli and purified as described. LightShift Chemiluminescent EMSA Kit (Thermo Scientific) was used for shift assay according to the manufacturers' instructions. Unlabeled (1fold or 100-fold of labeled probe) probe was used for competitive reaction 61 . The uncropped scan of the result was shown as Supplementary Figure 8.
Statistical analysis. Unless otherwise stated, data are presented as means ± S.D. Student's t test was used as statistical analysis by using Microsoft Excel as described.
Data availability. The authors declare that all data supporting the findings of this study are available within the article and its supplementary information files or from the corresponding author upon reasonable request. Microarray data that support the findings of this study are available in GEO datasets with the accession code GSE106809.