Lysine-specific histone demethylase 1 inhibition promotes reprogramming by facilitating the expression of exogenous transcriptional factors and metabolic switch

Lysine-specific histone demethylase 1 (LSD1) regulates histone methylation and influences the epigenetic state of cells during the generation of induced pluripotent stem cells (iPSCs). Here we reported that LSD1 inhibition via shRNA or specific inhibitor, tranylcypromine, promoted reprogramming at early stage via two mechanisms. At early stage of reprogramming, LSD1 inhibition increased the retrovirus-mediated exogenous expression of Oct4, Klf4, and Sox2 by blocking related H3K4 demethylation. Since LSD1 inhibition still promoted reprogramming even when iPSCs were induced with small-molecule compounds in a virus-free system, additional mechanisms should be involved. When RNA-seq was used for analysis, it was found that LSD1 inhibition reversed some gene expression changes induced by OKS, which subsequently promoted reprogramming. For example, by partially rescuing the decreased expression of Hif1α, LSD1 inhibition reversed the up-regulation of genes in oxidative phosphorylation pathway and the down-regulation of genes in glycolysis pathway. Such effects facilitated the metabolic switch from oxidative phosphorylation to glycolysis and subsequently promoted iPSCs induction. In addition, LSD1 inhibition also promoted the conversion from pre-iPSCs to iPSCs by facilitating the similar metabolic switch. Therefore, LSD1 inhibition promotes reprogramming by facilitating the expression of exogenous transcriptional factors and metabolic switch.

Scientific RepoRts | 6:30903 | DOI: 10.1038/srep30903 consequently decreased global levels of DNA methylation 5 . LSD1 also affects DNA damage-induced cell death through modifications on E2F1, and affects the cell cycle via MYPT1 demethylation 6,7 . LSD1 plays critical functions during embryonic development and the differentiation of ESCs. LSD1 deletion in mice leads to embryonic lethality at approximately day 6 due to an aberrant developmental program 5,8 . LSD1 regulates the balance between self-renew and differentiation by silencing several developmental genes with bivalent domains 9,10 . Since LSD1 occupies the enhancer and core promoters of some Oct4/Sox/Nanog-regulated targets 10 , it is reasonable to propose interactions between LSD1 and pluripotency-related transcriptional factors (TFs). Both Oct4 and Nanog interact with LSD1 or the LSD1-NuRD complex 11,12 , and their specific interaction could form a bridge between pluripotent transcription factors and LSD1.
Because epigenetic regulation plays critical roles during the generation of induced pluripotent stem cells (iPSCs) 13,14 , LSD1 has also been studied extensively in this process. Lithium greatly enhances iPSCs generation via down-regulation of LSD1 15 . In addition, an LSD1 inhibitor is included in the cocktail of small-molecule compounds which induces mouse somatic cells to iPSCs (CiPSCs) 16 . However, opposite observations also exist, LSD1 inhibition leads to an increase of mTOR expression, which subsequently impairs the reprogramming 17 . As a subunit of the larger LSD1 complex, the co-repressor Rcor2 can substitute for Sox2 during reprogramming, which is consistent with the critical functions of Sox2 in regulating the sensitivity of cancer cells to LSD1 inhibition 18,19 .
The beneficial roles of LSD1 inhibition were further confirmed recently during a shRNA library screening with the hTERT-stabilized secondary fibroblast reprogramming system and during a C/EBPα -accelerated B cell reprogramming with four Yamanaka factors 20,21 . In addition, the ability of LSD1 to promote oxidative metabolism of white adipose tissue suggested its potential involvement in the metabolic switch from oxidative to highly glycolytic at early stage of reprogramming 22,23 . Therefore, we aimed to further explore the functions of LSD1 during iPSC generation by determining the time-dependent regulatory roles of LSD1 inhibition, identifying the key effectors down-stream the inhibition, and demonstrating the potential influences on the metabolic switch from oxidative to glycolytic.

LSD1 inhibition promotes iPSCs generation.
Since Lsd1 expression in iPSCs and R1 ESCs were significantly higher than that in MEFs (Fig. 1a,b), a possible beneficial role of LSD1 for reprogramming was suggested. However, retrovirus-mediated over-expression of Lsd1 did not affect the efficiency of reprogramming (Fig. 1c,  Supplemental Fig. S1). In contrast, down-regulating Lsd1 with shRNA promoted iPSCs generation significantly, to more than 2 folds of control level (Fig. 1c, Supplemental Fig. S1).
To determine the functions of LSD1 during reprogramming in detail, a LSD1 inhibitor, tranylcypromine with an IC 50 of less than 2 μ M 24 , was used. Tranylcypromine had a variable effect on reprogramming efficiency, increasing the number of GFP + colonies with increasing concentrations, but declining at higher levels possibly due to cell toxicity or off-target effects (Fig. 1d). The optimal concentration was 20 μ M (T20, about 10 folds of IC 50 ) which increased reprogramming efficiency to around 3 folds of the control level (Fig. 1d). In addition, LSD1 but not its mutant, K661A which has little demethylase activity 25 , impaired T20-or sh-LSD1-induced increases in reprogramming efficiencies (Fig. 1e,f). These observations not only confirmed the effects of tranylcypromine and shRNA were through LSD1, but also indicated the critical roles of LSD1's demethylase activity in promoting iPSCs generation.
We next examined the time-dependent functions of LSD1 during reprogramming. The beneficial roles of LSD1 inhibition (T20 treatment) was only observed at early stage of reprogramming, from Day 2 to 5 during a 12-day reprogramming (Fig. 1g). Similarly, shRNA against LSD1 only promoted reprogramming significantly at early stage though its suppression last until late stage of reprograming (Fig. 1h, Supplemental Fig. S1).
Since c-Myc is also frequently used for reprogramming 14 , the effects of LSD1 inhibition in a reprogramming system with Oct4, Klf4, c-Myc and Sox2 (OKMS) was also determined. As indicated in Supplemental Fig. S1, LSD1 inhibition promoted reprogramming at early stage.
In addition, another inhibitor of LSD1, phenelzine also enhanced reprogramming with or without c-Myc in a dose-dependent manner (Supplemental Fig. S1). To further exclude the possible off-target effects, GSK2879552, a potent, selective, and irreversible LSD1 inhibitor 26 , was also determined to facilitate reprogramming with or without c-Myc in a dose-dependent manner (Supplemental Fig. S1). In addition, the beneficial effects of both phenelzine and GSK2879552 were limited to the early stage of reprogramming.
Furthermore, T20 also promoted the reprogramming of human foreskin fibroblasts as indicated by AP staining (Fig. 1i). These consistent results generated from different reprogramming systems further supported the beneficial effects of LSD1 inhibition.
iPSCs colonies generated with OKS and T20 had normal karyotypes, proper demethylation on both Nanog and Oct4 promoters, and the expression of endogenous pluripotency markers such as Nanog and SSEA1 (Supplemental Fig. S2). These iPSCs were able to form chimeric mice with germ line transmission (Supplemental Fig. S2).
T20 does not affect cell cycle and global DNA methylation. Since both p53 and DNMT1 are reported LSD1 substrates 4,5 , LSD1 inhibition may facilitate reprogramming by modulating p53 methylation and AP staining was performed to determining the efficiencies (n = 5). One-way ANOVA with Dunnett's test as a post-hoc was used to compare the experimental with control groups. The gels used for different antibodies were run under the same experimental conditions (a). Representative images were cropped from the original images with no modification on the relative intensities. The full length blots were provided in Supplementary Figure  It is possible that the observed beneficial effects on reprogramming were due to the alterations in cell cycle or global DNA methylation because of their influences on iPSCs generation 27,28 . Although the methylation levels of p53 and DNMT1 were not determined because of the lack of specific antibody, the possible involvement of their demethylation under current paradigm was excluded by determining their down-stream effects, modulations on cell cycle and global DNA methylation. The similar proliferation rates and lengths of different cell cycle phases in cells treated with or without T20 suggested that LSD1 inhibition did not affect reprogramming via a pathway from methylated p53 to cell cycle (Supplemental Fig. S3). The unchanged global DNA methylation levels suggested the influences from LSD1 inhibition were not mediated by DNMT1 or global DNA methylation. However, the unchanged down-stream effects such as cell cycle and global DNA methylation might be resulted from the changes on expression and methylation levels simultaneously. To exclude this possibility, the expression of p53 and Dnmt1 were determined. The expression of p53 and Dnmt1 were not affected by LSD1 inhibition, suggesting that the methylation levels of these two protein were unchanged from another aspect (Supplemental Fig. S3). In addition, Apoptosis also seems unaffected by T20 treatment as the percentage of TUNEL + cells remained unchanged (Supplemental Fig. S3).

T20 increases exogenous OKS expression by preventing H3K4 demethylation. LSD1 is essential
in maintaining low H3K4 methylation on endogenous retrovirus elements and subsequently suppresses their expression in ESCs 29 . Thus the abilities of T20 to affect retrovirus-delivered exogenous expression of Oct4, Klf4, and Sox2 were determined. Although retrovirus delivery of OKS increased their expression to different extents, LSD1 inhibition with T20 further boosted their expression to about three folds.
To explore the connection between increased exogenous expression of OKS and LSD1-mediated histone demethylation, shRNAs against Dpy30 and Eset were used (Supplemental Fig. S4). Since Dpy30 and Eset are critical for H3K4 and H3K9 methylation respectively 30,31 , blocking their expression should affect different parts of LSD1 inhibition. As indicated in Fig. 2a, increased exogenous expression of OKS could be blocked by sh-Dpy30 but not sh-Eset. Therefore, LSD1 inhibition facilitates OKS expression by suppressing related H3K4 demethylation.
In order to determine the contributions of increased exogenous OKS expression to reprogramming, different amounts of viruses were used to infect MEFs at the beginning of reprogramming. The abilities of T20 to increase exogenous OKS expression were between those of 4-fold and 8-fold of viruses (Fig. 2b). Although both 4-fold and 8-fold of viruses increased iPSCs generation efficiency significantly, the reprogramming efficiencies induced by them were still lower than that of OKS+ T20 (Fig. 2c), suggesting that increased OKS expression only accounted for part of T20's effects.
To further confirm the observation above, the effects of T20 on reprogramming were determined in a virus-free system. When using the previous reported protocol to generate CiPSCs 16 , T20 also greatly promoted iPSCs generation at early stage ( Fig. 2d). Therefore, LSD1 inhibition facilitates reprogramming via another pathway.
In addition, immunoblotting data suggested that T20 increased global levels of H3K4me2/3 and H3K9me2/3 (Supplemental Fig. S4). Furthermore, reducing the expression of Dpy30 or Eset with shRNA significantly impaired but not fully blocked the beneficial roles of T20 (Fig. 2e). Therefore, LSD1 inhibition with T20 promotes iPSCs generation by regulating both H3K4 and H3K9 demethylation and through multiple pathways.

T20 partially reverses gene expression changes induced by OKS.
Since the beneficial roles of T20 on reprogramming was mostly at early stage, we performed RNA-seq analysis in MEFs and cells on Day 3 during reprogramming with (OKS+ T20) and without T20 (OKS) (Supplemental Table S1). Gene expression profiles were compared between any two of these three samples, and totally three comparisons were performed (Fig. 3a, Supplementary Fig. S5). The difference between the expression profiles of OKS group and OKS+ T20 group was smaller than the difference between MEF and OKS or between MEF and OKS+ T20. Although the difference between OKS and OKS+ T20 was the smallest in the three comparisons, it indicated the function of T20 and was focused for further analysis.
The 9399 genes with detectable expression in current RNA-seq were divided into 14 groups basing on their expression ratios, OKS vs MEF. It was found that OKS preferred to up-regulate genes with relatively low expression in MEFs, and down-regulate those with relatively high expression (Supplemental Fig. S5). This phenomenon was a typical pattern for gene expression changes. However, when compared the gene expression in OKS+ T20 with those in OKS, T20 preferred to increase the expression of genes that were down-regulated by OKS, while to decrease those up-regulated by OKS (Fig. 3b). To be more accurate, within the 1074 genes that were down-regulated by more than 50% after OKS delivery, 509 genes had their expression changes un-affected by T20, 514 genes had their expression changes been significantly reversed, and only 51 genes were further reprogramming. Exogenous expression of OKS were determined by qPCR on Day 3 during reprogramming and normalized to those in OKS group ((b), n = 5). GFP + colonies numbers on Day 12 were summarized in ((c), n = 5). (d) CiPSCs were generated as reported previously and outlined in Methods. Tranylcypromine was used at 20 μ M and in indicated time periods during reprogramming. GFP + colonies were counted for efficiencies (n = 5). (e) Retrovirus encoding sh-Eset and sh-Dpy30 were used in different combinations with T20, retrovirus encoding sh-LSD1 and LSD1 for MEFs reprogramming. GFP + colonies were counted to represent the efficiencies (n = 6). One-way ANOVA with Dunnett's test as a post-hoc was used for (a-d), while Two-way ANOVA with Bonferroni's test was used for (e).
Scientific RepoRts | 6:30903 | DOI: 10.1038/srep30903 down-regulated by T20 (Fig. 3c). Similarly, within the 1134 genes that were up-regulated by more than 100% after OKS delivery, 614 genes had their expression changes un-affected by T20, 339 genes had their expression changed been significantly reversed, and only 181 genes were further up-regulated by T20 (Fig. 3c). Therefore, T20 partially reverses gene expression changes induced by OKS. Since the expression profile of OKS+ T20 group was still closer to OKS group than to MEF group, the ability of T20 to modulate gene expression could not overwhelmed those of Oct4, Klf4 and Sox2.
Considering the interactions between LSD1 and pluripotent transcriptional factors like Oct4, Sox2. Klf4 and Nanog (Fig. 3d), OKS might down-regulate genes with relatively high expression via H3K4 demethylation, while up-regulated those with low expression via H3K9 demethylation, with help from LSD1 at early stage of reprogramming. When LSD1 was inhibited, OKS-induced expression changes might be reversed. T20 assists in a metabolic switch during iPSCs generation. To connect the abilities of T20 to reverse OKS-induced expression changes to its beneficial roles for reprogramming, genes with more than two-fold expression differences between OKS and OKS+ T20 groups were analyzed with DAVID online service (http:// david.abcc.ncifcrf.gov/) to identify the enriched Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways 32 . Genes related to mitochondrial and oxidative phosphorylation pathway (OX,  4 0 . 4~0 . 6 0 . 6~0 . 8 0 . 8~0 . 9 0 . 9~1 . 0 1 . 0~1 . 1 1 . 1~1 . 2 1 . 2~1 . 5 1 . 5~2 . 0  (c) 2208 genes with significant expression differences (over two folds) between MEF and OKS groups were analyzed separately, downregulation in left panel and up-regulation in right. Within these genes, genes with expression ratios (OKS+ T20 vs OKS) between 0.67 and 1.50 were classified as not significant (N/S, in Blue), genes with ratios over 1.50 were as increased (Incre. in Red), while genes with ratios below 0.67 were as decreased (Decre. in Green). Genes with increased expression were classified as reversed changes in left panel but as consistent changes in right panel. The classification of genes with decreased expression was just opposite. (d) Oct4, Klf4, Sox2 and Nanog were over-expressed in MEFs via retrovirus system for three days, antibody against LSD1 was used for immuno-precipitation (n = 5). One-way ANOVA with Dunnett's test as a post-hoc was used. The gels used for different antibodies were run under the same experimental conditions (a). Representative images were cropped from the original images with no modification on the relative intensities. The full length blots were provided in Supplementary Figure S8.
Genes in OX pathway and another related pathway, glycolysis pathway (KEGG: 00010), were then analyzed basing on the information collected from KEGG website (http://www.genome.jp/kegg/). 98 out of 116 genes in OX pathway had detectable expression in the current RNA-seq. About 80% of these 98 genes were up-regulated by OKS on Day 3 during reprogramming and experienced down-regulation when T20 was included in reprogramming system simultaneously (Supplemental Table S3). On the other hand, 34 out of 60 genes in glycolysis pathway were detectable in the current RNA-seq. About 40% of these 34 genes were down-regulated by OKS on Day 3 and up-regulated when T20 was used (Supplemental Table S3). 61 OX genes and 8 glycolysis genes were further selected out because of the larger expression differences among MEF, OKS and OKS+ T20 samples. Their expression levels were presented in Fig. 4c. Therefore, T20 reverses OKS-induced expression changes of genes in OX and glycolysis pathways.
To confirm the RNA-seq results, qPCR was performed to determine the expression of several representative genes. Nuclear respiratory factor 1 (Nrf1) expression was not affected significantly by OKS or T20, but representative genes in the OX pathway were up-regulated by OKS and down-regulated by T20 (Fig. 5a). Hypoxia-inducible factor 1α (HIF1α) expression was down-regulated by OKS while reversed by T20, which was consistent with its down-stream glycolysis genes, Pdk1, Pdk2, and Pkm2 (Fig. 5b). Seahorse XF24 extra-cellular flux analyzer (Seahorse Bioscience) was then used to assess cellular energy metabolism on Day 3 during reprogramming. According to the OCR and ECAR readings in Fig. 5c,d, T20 treatment shifted cells from oxidative to glycolytic metabolism. Since transition from oxidative to glycolytic metabolism is required during iPSC generation 33

Intracellular membrane-bounded (328)
Organelle ( Similarly, shRNA against Lsd1 and LSD1-specific inhibitor, GSK2879552, also decreased the expression of representative genes in the OX pathway, increased the expression of glycolysis genes, and facilitated the metabolic switch from oxidative to glycolytic when used together with Oct4, Klf4 and Sox2 during iPSCs generation (Supplementary Figure S6).  Table S4). Binding motifs of Klf4 and HIF1α were enriched in the corresponding regions of the 60 glycolysis genes. Thus NRF1 and HIF1α were selected for further studied. Since significant expression changes were only observed on Hif1α but not Nrf1 under current paradigm (Fig. 5a,b), whether T20 facilitated metabolic switch by modulating Hif1a expression was further determined.
Although Hif1a expression in iPSCs/ESCs were lower than that in MEF, its expression on Day 3 during reprogramming were suppressed to a level significantly lower than those in iPSCs and ESCs (Fig. 6a). This over down-regulation of Hif1a was resulted from the exogenous expression of Oct4, Klf4, and Sox2 (Fig. 6a). Considering the interaction between LSD1 and these transcriptional factors (Fig. 3d), it was reasonable to proposed that Oct4, Klf4, and Sox2 induced H3K4 demethylation on Hif1a gene locus with the help from LSD1 and subsequently decreased Hif1a expression. When LSD1 was inhibited, such H3K4 demethylation was impaired and Hif1a expression was reversed.
To test the hypothesis above, ChIP-qPCR was used to determine the binding of OKS and LSD1 to Hif1a gene locus and to explore the changes on H3K4 methylation in the same locus. As indicated in Fig. 6b, Oct4, Klf4, Sox2 and LSD1 bound to Hif1a gene locus on Day 3 during reprogramming. In addition, ChIP-qPCR with antibodies against H3K4me2 and H3K4me3 suggested that the H3K4 methylation level on Hif1a gene locus decreased after OKS delivery, and could be reversed partially with T20 treatment (Fig. 6c,d). Therefore, Oct4, Klf4, and Sox2 bind to Hif1a gene locus and regulate related H3K4 methylation with the help from LSD1.
Then the contributions of Hif1a expression modulation and metabolic switch to T20's functions during reprogramming were determined. As expected, an shRNA against Hif1a counteracted with T20 in regulating expression of representative genes in OX and glycolysis pathways (Fig. 6e), and reduced the ability of T20 to promote iPSCs generation (Fig. 6f). Additionally, the oxidative phosphorylation inhibitor, rotenone, increased reprogramming efficiency, which could be blocked by LSD1 over-expression at least partially (Fig. 6f). These observations further supported the conclusion that LSD1 inhibition assists the metabolic switch from oxidative to glycolytic by regulating Hif1a expression. T20 also promotes pre-iPSCs conversion by assisting metabolic switch. In addition, when analyzing current RNA-seq together with previous studies on pre-iPSCs, iPSCs, and ESCs 36-38 , LSD1 inhibition seemed to promote the conversion from pre-iPSCs to iPSCs (Supplemental Tables S1 and S3). Thus two already established pre-iPSCs lines, pre-2-2 and pre-3 39 , was used to determine the effects of T20 on the Vitamin C (Vc)-induced conversion from pre-iPSCs to iPSCs.
As expected, both shRNA against Lsd1 and treatment with T20 or GSK2879552 promoted the conversion by about two folds (Fig. 7a-d). To determine potential interaction between Vc and T20, the concentrations of Vc in the medium were reduced to different levels. Although T20 could not convert pre-iPSCs to iPSCs without Vc, it promoted these conversions significantly even at a much lower concentration (0.5 μ M) of Vc (Fig. 7e,f). In addition, pre-treating pre-iPSCs with T20 for 4 days increased the following conversions no matter whether the cells were continued to be treated with Vc alone or with both Vc and T20 on Day 5 to 11 (Fig. 7g). Thus the beneficial roles of T20 on this conversion were independent from the functions of Vc.
In addition, the beneficial roles of T20 on the metabolic switch were also observed during the conversion of pre-iPSCs. shRNA against Hif1a significantly inhibited the beneficial roles of T20, while Lsd1 over-expression also impaired rotenone-increased conversion efficiency (Fig. 7h).

Discussion
The current study demonstrated the two mechanisms employed by LSD1-inhibition to promote iPSCs generation at early stage of reprogramming. On the one hand, LSD1 inhibition increases retrovirus-delivered exogenous expression of Oct4, Klf4, and Sox2 by preventing related H3K4 demethylation. The increased exogenous expression of Oct4, Klf4, and Sox2 contribute to the T20-faciliated reprogramming at least partially. On the other hand, because of the interactions between LSD1 and Oct4, Klf4 and Sox2, LSD1 inhibition impairs OKS-induced performed. Three sets of primers on Hif1α gene locus were used for qPCR. IgG was used as negative control and the results were normalized to the qPCR results from 1% inputs (n = 5). (e) qPCR results on genes in OX pathway and glycolysis pathway on Day 3 of reprogramming with OKS+ sh-Luc, OKS+ sh-Luc+ T20, or OKS+ sh-Hif1α + T20 (n = 5) (f) Retrovirus encoding sh-Hif1a was used with or without T20, and oxidative phosphorylation inhibitor, rotenone (Rot), was used with retrovirus encoding Flag or LSD1 during iPSCs generation. GFP + colonies were counted (n = 6). One-way ANOVA with Dunnett's test as a post-hoc was used.
H3K4 demethylation on genes with relative high expression and H3K9 demethylation on genes with relative low expression (Supplemental Table S1). As indicated by the RNA-seq analysis, LSD1 inhibition partially impaired OKS-induced expression changes, like the expression changes of genes in OX and glycolysis pathways. Thus, although the two mechanisms seem to be controversial from each other, they both promote reprogramming and account for the beneficial roles of LSD1 inhibition. Since LSD1 inhibition also promotes reprogramming even in a virus-free system, and LSD1 inhibition increases iPSCs generation by impairing some OKS-induced expression changes, some OKS-induced expression changes may be unfavorable for iPSCs generation. The typical unfavorable expression changes identified in the current studies were those of genes in OX and glycolysis pathways. Although cells transit from oxidative to glycolytic metabolism during iPSCs generation 33 , genes in both OX and glycolysis pathways were up-regulated during reprogramming when analyzing previous studies on pre-iPSCs, iPSCs, and ESCs [36][37][38] (Supplemental Tables S1   pre-iPSC: 2 (e,f) The effects of T20 on pre-iPSCs conversion were determined in pre-2-2 ((e), n = 6) and pre-3 ((f), n = 6) with reduced Vc concentrations. (g) Pre-2-2 were pre-treated with PBS (control), Vc and T20 from Day 0 to Day 4, and then treated with Vc or Vc+ T20 from day 5 to Day 11. GFP + colonies were counted.
(h) Retrovirus encoding sh-Luc or sh-Hif1α was used with or without T20, and Rot was used with retrovirus encoding Flag or LSD1 during iPSCs generation. GFP + colonies were counted.Two-way ANOVA with Bonferroni's test as a post-hoc test was used. and S3). OKS over up-regulated genes in OX pathway, but down-regulated genes in glycolysis pathway at early stage of reprogramming, which subsequently impaired metabolic switch. Inhibiting LSD1 partially reversed these expression changes, facilitated metabolic switch, and subsequently contributed to the beneficial roles of T20. Then the question was to determine what is the percentage of such unfavorable expression changes in OKS-induced expression changes. Our hypothesis was that more than 50% OKS-induced expression changes was unfavorable. After analyzing current RNA-seq data together with previous studies on pre-iPSCs, iPSCs, and ESCs [36][37][38] (Supplemental Tables S1 and S3), 2208 genes whose expression changed by more than 2 folds after OKS delivery in current RNA-seq were used for further analysis. Three expression ratios, OKS vs MEF, iPSC&ESC vs MEF, and pre-iPSC vs iPSC&ESC, were used to define "consistent", "un-necessary", "opposite", and "over-regulated" changes induced by OKS (Supplemental Fig. S7). 693 genes (31%) had OKS-induced changes consistent with those from MEFs to iPSCs/ESCs, 1144 genes (52%) had opposite or un-necessary changes, and 203 genes (9%) seemed to be over-regulated (Supplemental Table S1 and Fig. S7). If considering opposite, un-necessary, and over-regulated changes as unfavorable changes, majority (61%) of expression changes induced by OKS were unfavorable for reprogramming.
The 61% above was calculated on Day 3 in current reprogramming system, another microarray datasets of a 21-day reprogramming system was analyzed 38 . Genes with unfavorable regulations by OKS decreased during at late stage of reprogramming (Supplemental Fig. S7). In addition, when calculating the percentages of unfavorable expression changes at different time points in the 21-day reprogramming system, the percentages peaked on Day 5 and gradually decreased (Supplemental Fig. S7), suggesting a significant portion of expression changes induced by OKS were unfavorable especially at early stage. Therefore, LSD1 inhibition might promote reprogramming by reversing some of these unfavorable changes.
The "unfavorable" hypothesis above still had additional challenges. These unfavorable changes might be a necessary stage for reprogramming and might be not really "unfavorable". In addition, RNA-seq and microarray data were generated from large population of cells during reprogramming and were contaminated with cells that were not on their way to iPSCs. Thus further testing on this hypothesis should be based on addition single-cell tracing and expression profile analysis.
In addition, another recent report on autophagy and mechanistic target of rapamycin complex 1 (mTORC1) supported current hypothesis from another aspect 40 . OKS down-regulated mTORC1 expression, but up-regulated genes involved in the elongation and expansion of autophagy like Atg5, Atg7, Atg4b and Atg16l. T20 might also facilitate reprogramming by reversing these gene expression and suppressing autophagy.

Methods
Animal Studies. Our studies were performed in accordance with the guidelines for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocols were approved by the Committee on the Ethics of Animal Experiments of the Guangzhou Institutes of Biomedicine and Health. All efforts were made to minimize the number of animal and the animal discomfort.
Generation of iPSCs. MEFs were derived from 13.5-day mouse embryos carrying the Oct4-GFP transgenic allele 41 . These MEFs were maintained in DMEM (Gibco) supplemented with 10% FBS (Gibco), nonessential amino acids (NEAA, Gibco), penicillin/streptomycin (P/S, Thermo), and GlutaMAX (Gibco). Retrovirus was produced with Plat-E cells and pMXs-based retroviral vectors as previous reported except calcium phosphate transfection protocol was used 42 . MEFs within two passages were split into twelve-well plate (1.5 × 10 4 cells/well). After adding polybrene to 4 μ g/ml, the viral supernatant was used for infection. Oct4, Klf4, c-Myc and Sox2 or Oct4, Klf4 and Sox2 were introduced into cells twice on Day 0 and Day 1 respectively, while mES or mES-Vc [high glucose DMEM (Hyclone), NEAA, P/S, GlutaMAX, leukemia inhibitory factor (LIF), β -mercaptoethanol, pyruvate, and 10% FBS without or with Vc (Sigma, 49752)] was used on Day 2. Medium was replaced with freshly prepared medium every day. If inhibitor, like GSK2879552 or T20, was used, it was also added to the fresh medium from the stock solution every day. iPSC colonies were counted or picked according to their Oct4-GFP expression and ES-like morphology, on Day 12-14. R1 ESCs and iPSCs were cultured on MEF feeder cells in Knockout DMEM supplemented with Knockout Serum Replacement (KSR, Gibco), LIF, PD0325901, CHIR-99021, NEAA, P/S, GlutaMAX, and β -mercaptoethanol.
HFFs were cultured in DMEM with 10% FBS, NEAA, GlutaMAX and P/S. 1 × 104 HFFs were plated into every well of 24-well plate. On Day 0 and 1, HFFs were incubated with retrovirus-containing supernatant with 4 μ g/ml polybrene for 12 hours followed with a 12-hour rest. On day 2, medium was changed to DMEM with 20% defined FBS, and bFGF (Shenzhen Symmix Industry). On Day 4, Vc and VPA (1 mM; Merck, Germany) was added to the medium. Conditioned medium, the supernatant collected and filtered 12 hours after MEF feeder cells were cultured in current medium, was used from Day 10 until the AP staining on Day 24.