Kinome-wide functional analysis highlights the role of cytoskeletal remodeling in somatic cell reprogramming

Abstract

The creation of induced pluripotent stem cells (iPSCs) from somatic cells by ectopic expression of transcription factors has galvanized the fields of regenerative medicine and developmental biology. Here, we report a kinome-wide RNAi-based analysis to identify kinases that regulate somatic cell reprogramming to iPSCs. We prepared 3,686 small hairpin RNA (shRNA) lentiviruses targeting 734 kinase genes covering the entire mouse kinome and individually examined their effects on iPSC generation. We identified 59 kinases as barriers to iPSC generation and characterized seven of them further. We found that shRNA-mediated knockdown of the serine/threonine kinases TESK1 or LIMK2 promoted mesenchymal-to-epithelial transition, decreased COFILIN phosphorylation, and disrupted Actin filament structures during reprogramming of mouse embryonic fibroblasts. Similarly, knockdown of TESK1 in human fibroblasts also promoted reprogramming to iPSCs. Our study reveals the breadth of kinase networks regulating pluripotency and identifies a role for cytoskeletal remodeling in modulating the somatic cell reprogramming process.

Figure 1. Kinome-wide RNAi screen for regulators of somatic cell reprogramming

(A) Experimental design. Oct4-GFP MEFs were transduced on day 0 (D0) with retroviruses encoding the four pluripotency factors, Oct4, Sox2, Klf4, and c-Myc (OSKM; 4F), to induce reprogramming. A total of 3,686 lentiviruses carrying shRNAs targeting the entire kinome were produced in 293FT cells. On day 3 after 4F transduction, MEFs were infected with individual lentiviruses in separate wells. ES medium was changed on day 4 and every other day thereafter until GFP+ colonies were quantified on day 20. The pLKO.1/TRC-Mm1.0 vector expresses shRNAs from a U6 promoter and includes downstream central polypurine tracts (cPPT). (B) Identification of barrier kinases from the primary screen. Dot-plot shows the result of the primary screen assessing the effects of 3,686 shRNAs targeting 734 kinase genes. Results are expressed as the fold change in GFP+ colony counts after normalization to the control pLKO lentiviral-infected cells. The red line indicates the two-fold threshold used to select barrier kinases as hits. Validation of 153 genes from the primary screen was performed in duplicate in a 12-well format in the secondary screen. Subsequently, 59 genes were further validated in a tertiary screen in a 12-well format in duplicate and repeated five times. Red dots indicate shRNAs targeting the six kinases that were selected for further investigation. See also Figure S1 and Tables S1–S3.

Figure 2. Lentivirus-based knockdown of seven kinases at different times during reprogramming reveals the existence of time windows critical for enhancing reprogramming and their role in MET

(A) The effects of silencing p53 and seven selected kinases on reprogramming efficiency were evaluated at different times after 4F transduction. Oct4-GFP MEFs were transduced with 4F and passaged at 2 days post-infection (dpi). Lentivirus-based shRNA knockdown was performed on the indicated days by adding fresh virus-containing supernatants. GFP+ colonies were counted on 18 dpi. Results show the fold increase in GFP+ colonies relative to numbers obtained with empty pLKO.1. Results are mean ± SD of two independent experiments performed in triplicate. (B) Knockdown of p53 and the seven kinases enhances MET in 4F-transduced MEFs. E-cadherin expression served as a marker for induction of MET during the initial stage of reprogramming. MEFs were transduced with shRNAs targeting p53 or the seven kinases in the presence of 4F. Empty vector or a non-targeting shRNA served as controls. Total RNA was harvested on day 3 after shRNA lentiviral infection and E-cadherin expression was measured by RT-qPCR. See also Figure S2 and Table S4.

Figure 3. shTESK1-iPSCs exhibit a fully pluripotent state

(A) TESK1 overexpression compromises reprogramming efficiency. TESK1 or HA-TESK1 were cloned and expressed during 4F transduction. Immunoblotting confirmed the protein expression and function by enhancing cofilin phosphorylation (Figure S4B). Reprogramming efficiency was decreased more than 80% in TESK1-overexpressing cells compared with pMX-dsRed control-transduced cells. (B) shTESK1-iPSCs switch on endogenous mESC markers. GFP+ shTESK1-iPSCs cultured on feeder layers and collected on 16 dpi show positive staining for SSEA-1, Nanog, and alkaline phosphatase (AP), all indicators of pluripotency. (C) shTESK1-iPSCs can differentiate into three germ layers in vitro. Embryoid bodies formed from shTESK1-iPSCs were collected on day 14, fixed with 4% PFA and immunostained for β-tubulin III (ectoderm), sarcomeric actinin (mesoderm), or α-fetoprotein (AFP; endoderm). Insets show DAPI staining of nuclei (blue) in a wider field of view. (D) shTESK1-iPSCs can differentiate into many lineages in vivo. shTESK1-iPSCs were injected subcutaneously into the backs of Nude mice. Teratomas were removed at 3–4 weeks and stained with H&E. (E) shTESK1-iPSCs show gene expression profiles similar to mESCs. Genome-wide mRNA expression of shTESK1-iPSCs was compared with mESCs and MEF controls. See also Figure S3.

Figure 4. TESK1 regulates cytoskeletal remodeling during reprogramming

(A) TESK1 is highly expressed in MEFs and regulates cofilin phosphorylation. TESK1, phosphorylated (P-) cofilin, and total (T-) cofilin expression levels were detected by immunoblotting of extracts from Oct4-GFP MEFs, mESCs, or MEFs transfected with TESK1-targeting siRNA (siTESK1) or control siRNA (siNT). GAPDH served as an internal control. (B) Visualization of polynucleation of actin filaments. Actin filaments were detected by staining MEFs and mESCs with rhodamine-labeled phalloidin. (C) TESK1 knockdown disrupts actin filamentous structure. Actin cytoskeleton organization was visualized by confocal microscopy of MEFs transfected with control siRNA or sRNA targeting TESK1 or COF. The top, middle, and bottom panels show the basal section, middle section, and Z-projection, respectively. (D) TESK1 regulates filamentous actin network organization. MEFs expressing non-targeting shRNA (top row) or shTESK1 (bottom row) were imaged by light microscopy (column 1) or TEM (columns 2–4). The white boxes in the left panel mark the areas shown enlarged in the right panels. On top the view is rotated by ~45 degrees anticlockwise from the left panel to the right panels, on the bottom the view is rotated by ~20 degrees anticlockwise. Blue arrows indicate an isotropic network and small vertical ruffles, red highlighting indicates actin bundles. The experiment was repeated at least three times for each condition, with similar results. Scale bars = 20 μm in B and D and 5 μm in C. See also Figure S4.

Figure 5. Cofilin phosphorylation modulates reprogramming

(A) Cofilin phosphorylation (P-COF) is dynamically regulated during the reprogramming process. 4F transduced MEFs were harvested on different days (D3, D6 and D9) and the cell lysates were used to analyze P-COF and total cofilin (T-COF) levels by immunoblotting. GAPDH served as loading control while E-CAD served as marker for MET and ES cell states. (B) Immunoblot analysis of MEF cells overexpressing GFP-fusion proteins, wild-type cofilin (COF-WT-GFP) or a mutant cofilin (COF-S3A-GFP), in the presence of non-targeting (siNT) or cofilin-targeting (siCOF) siRNA. Empty expression vector was used as a transfection control and GAPDH was used as loading control. Arrowhead indicates exogenously expressed cofilin-GFP and arrow indicates endogenous cofilin. (C) Phosphorylation incompetent cofilin promotes iPSC generation. Quantification of SSEA1+ iPSC colonies obtained from MEFs transduced with OSKM plus empty vector, wild-type cofilin (COF-WT-GFP), mutant cofilin (COF-S3A-GFP), COF-WT-GFP+HA-TESK1 or COF-S3A-GFP+HA-TESK1. Results are the means ± SD of three independent experiments. (D) Knockdown of TESK1 decreases ERK phosphorylation. Immunoblotting of TESK1, ERK1/2, phospho-ERK1/2, and Spry2 in Oct4-GFP MEFs, mESCs, and non-targeting (NT) shRNA- and shTESK1-transduced Oct4-GFP MEFs 4 days post-transduction. GAPDH served as an internal control. (E) Activation of the ILK signaling pathway results in ERK and cofilin phosphorylation. MEFs were grown on plates coated with fibronectin (Fbn) or a fibronectin analogue (Fbn-Anlg) to induce the ILK signaling pathway. MEFs seeded on gelatin-coated plates served as controls. Immunoblot analysis of phospho- and total ERK and cofilin in cell lysates prepared on 4 dpi. See also Figure S5.

Figure 6. TESK1 knockdown promotes reprogramming of human fibroblasts

(A) Left panel, number of AP+ clones derived from human BJ fibroblasts transfected with non-targeting (siNTC) or TESK1-targeting (siTESK1) siRNA. Right panel, RT-qPCR of TESK1 mRNA levels in control and TESK1-KD cells. (B) Representative AP+ clones derived from reprogramming of human BJ cells expressing siNT or siTESK1. (C) Expression of ESC pluripotent markers in iPSCs derived from human BJ cells expressing siTESK1. Staining for SSEA-3, NANOG, SOX2, Tra-1-81, Tra-1-60, and OCT4 was performed. (D) Teratoma formation shows the pluripotency of iPSCs. Cells were injected into SCID mice and tumors were harvested 8 weeks later and stained with H&E. (E) Normal karyotype of a human siTESK1-iPSC clone. See also Figure S6.

Figure 7. Interactome network of kinases and cytoskeletal remodeling in reprogramming

(A) Interactome network of kinases and transcription factors in iPSC generation. The interactome network depicts the functional associations of 24 bridge proteins that connect 28 barrier kinases and four transcription factors. The predicted network was generated by STRING (version 9.0) on the basis of protein interactions of high confidence (score >0.7). Each bridge protein is associated with at least two barrier kinases and one transcription factor. Top-ranked functional annotations generated by Ingenuity IPA suggests the bridge proteins are likely to be involved in the EMT and/or MET programs. Nodes and edges represent proteins and predicted functional associations, respectively. The colored lines indicate the seven types of evidence used in predicting the associations: red for fusion, green for neighborhood, blue for co-occurrence, purple for experimental, yellow for text mining, light blue for database, and black for coexpression evidence (see Table S5). (B) Proposed mechanism for TESK1 and LIMK2 function in cytoskeletal remodeling and iPSC generation. TESK1 or LIMK2 phosphorylates cofilin to promote the formation and/or stabilization of the cytoskeletal structures involved in EMT and cellular differentiation. RNAi-mediated silencing of either kinase inhibits cofilin phosphorylation, which in turn disrupts actin polynucleation, promotes the MET transition step of reprogramming in 4F-infected MEFs, and thus enhances iPSC generation.