The Short Isoform of NF-YA Belongs to the Embryonic Stem Cell Transcription Factor Circuitry§


  • Diletta Dolfini,

    1. Dipartimento di Scienze Biomolecolari e Biotecnologie, Università degli Studi di Milano, Milano, Italy
    Search for more papers by this author
  • Mario Minuzzo,

    1. Dipartimento di Scienze Biomolecolari e Biotecnologie, Università degli Studi di Milano, Milano, Italy
    Search for more papers by this author
    • Telephone: 39-02-50315005; Fax: 39-02-50315044

  • Giulio Pavesi,

    1. Dipartimento di Scienze Biomolecolari e Biotecnologie, Università degli Studi di Milano, Milano, Italy
    Search for more papers by this author
    • Telephone: 39-02-50315005; Fax: 39-02-50315044

  • Roberto Mantovani

    Corresponding author
    1. Dipartimento di Scienze Biomolecolari e Biotecnologie, Università degli Studi di Milano, Milano, Italy
    • Dipartimento di Scienze Biomolecolari e Biotecnologie, Università di Milano, Via Celoria 26, 20133, Milano, Italy
    Search for more papers by this author
    • Telephone: 39-02-50315005; Fax: 39-02-50315044

  • Author contributions: D.D.: conception and design and data analysis and interpretation; M.M.: conception and design; G.P.: data analysis and interpretation; R.M.: conception and design, financial support, data analysis and interpretation, manuscript writing, and final approval of manuscript.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS August 7, 2012.


Totipotency of embryonic stem cells (ESCs) is controlled at the transcriptional level by a handful of transcription factors (TFs) that promote stemness and prevent differentiation. One of the most enriched DNA elements in promoters and enhancers of genes specifically active in ESCs is the CCAAT box, which is recognized by NF-Y, a trimer with histone-like subunits—NF-YB/NF-YC—and the sequence-specific NF-YA. We show that the levels of the short NF-YA isoform—NF-YAs—is high in mouse ESCs (mESCs) and drops after differentiation; a dominant negative mutant affects expression of important stem cells genes, directly and indirectly. Protein transfections of TAT-NF-YAs stimulate growth and compensate for withdrawal of leukemia inhibitory factor (LIF) in cell cultures. Bioinformatic analysis identifies NF-Y sites as highly enriched in genomic loci of stem TFs in ESCs. Specifically, 30%–50% of NANOG peaks have NF-Y sites and indeed NF-Y-binding is required for NANOG association to DNA. These data indicate that NF-Y belongs to the restricted circle of TFs that govern mESCs, and, specifically, that NF-YAs is the active isoform in these cells. STEM CELLS2012;30:2450–2459


Totipotent embryonic stem cells (ESCs) have the capacity to generate a whole organism, by producing, under the proper developmental signaling, subsequent programs of progressive differentiation of tissues and organs. Transcriptional regulation is at the heart of the events leading to development, differentiation, and tissue maintenance: not surprisingly, a selected cohort of transcription factors (TFs) are key to ES stemness, by promoting their regulated growth and preventing differentiation. Among these, SOX2, OCT4, KLF4, SALL4, MYC, and NANOG are also involved in the process of reprogramming of differentiated cells to become ES, or ES-like induced pluripotent stem (iPS) cells; genome-wide analysis has established a network of coregulation among these TFs in ESCs [1–7].

Analysis of transcriptomes of human and mouse ESCs (mESCs) to find common regulated pathways has led to the identification of transcription factors binding sites (TFBS) enriched in important regulatory regions of ES-specific genes. One of these TFBS is the CCAAT box, highly enriched in active enhancers [8]. Although the CCAAT was found in some enhancers before [9, 10], it has been mainly considered as a promoter element: it is positioned between −60 and −100 with respect to the transcriptional start site, and the importance is underscored by mutations causing a drop, or complete elimination, of promoter activity [11]. The CCAAT box is recognized with high affinity and specificity by NF-Y, a heterotrimer composed of NF-YA, NF-YB, and NF-YC, all evolutionarily conserved [12]. NF-YB and NF-YC contain histone fold domains (HFD), structurally related to histones H2B and H2A, and forming a stable histone-like dimer [12]. NF-YA associates to the HFD dimer, conferring CCAAT specificity to the complex. All nucleotides of the pentanucleotide are critical for NF-Y binding, with immediate flanking sequences contributing substantially [11]. The early embryonic lethality of an NF-YA mouse knockout (KO) model due to defects in cell proliferation indicates that the TF is active in the very early stages of development [13]; more recently, tissue-specific KO experiments in liver and bone marrow confirm the importance of NF-YA in adult tissues as well [14, 15]. The NF-YA isoform is believed to be the limiting and regulatory subunit of the trimer, and, unlike the HFD dimer, is absent in some postmitotic cells and tissues [reviewed in16]. NF-YA has two major isoforms: NF-YA long (NF-YAl) and NF-YA short (NF-YAs), lacking 28 amino acids coded by Exon 3, as part of the Q-rich transcriptional activation (TA) domain [17].

There are at least two stem systems in which NF-Y is known to regulate genes of key TFs. (a) The expression of HOXB4, a master gene of hematopoietic stem cells (HSC), is driven by two important NF-Y sites: one in the promoter, which synergizes with nearby upstream transcription factor (USFs), and one in the intronic enhancer, which overlaps and synergizes with YY1 [9, 18, 19]. Retroviral infections of NF-YAs in mouse HSCs increased substantially their capacity to repopulate the bone marrow of immunocompromized animals, by enhancing the expression of several HOX genes, as well as Notch and Wnt signaling [19]. Most importantly, this experiment has been repeated with TAT-fusion recombinant NF-YAs added to the medium, showing a better engraftment and repopulation of bone marrow derived HSC in mice ex vivo [20]. (b) SOX2 is essential for ESCs stemness and indeed one of the TFs inducing reprogramming and totipotency upon overexpression in differentiated cells, along with OCT4 and KLF4 [21]. The Sox2 promoter contains important CCAAT boxes controlled by NF-Y [22]. Functional inactivation of NF-YA and NF-YB in mES led to cell death, confirming that NF-Y is essential [23].

A somewhat surprising result pointing at an important role of NF-Y in mESCs emerged recently: the search for genomic locations of the nuclear signaling kinases JNK (Jun N-terminal kinases) identified NF-Y sites, rather than the expected AP-1 [23]; comparison with NF-Y loci in mESCs, before and after differentiation, showed that JNK targets are part of the larger NF-Y ensemble, and indeed removal of the trimer from representative targets using a dominant negative (DN) NF-YA mutant led to elimination of JNK DNA-binding. The opposite was not true, namely, in JNK−/− fibroblasts NF-Y binding was unaffected. Because JNKs play a crucial role in mES differentiation, it was suggested that the NF-Y/JNK connection is important for the switch from stem to differentiated cells.

In this report, we studied the role of NF-YA isoforms in mESCs: we found that NF-YAs has a global role in maintenance of stemness, by activating directly key stem cells genes and, indirectly, by promoting the association of NANOG to a large part of its regulated ESCs targets.


Cell Culture and Differentiation

Mouse E14 ESCs were maintained in glasgow minimum essential medium (Sigma-Aldrich, Missouri, USA) supplemented with 10% ESC qualified fetal bovine serum (Gibco, Grand Island, NY,, 1 mM L-glutamine, 0.1 mM nonessential amino acids, 100 units/ml penicillin, 100 μg/ml streptomycin, 1 mM sodium pyruvate, 0.1 mM 2-mercaptoethanol, and 1,000 units/ml of recombinant leukemia inhibitory factor (LIF) (Chemicon, Temecula, CA, Embryoid bodies were formed by suspension culturing, chemical differentiation induction was performed in the absence of LIF with or without 0.5 μM all-trans-RA (Sigma).

Transfection Experiments

0.5 × 106 cells were plated on gelatin in six-well tissue culture plates 12 hours prior to transfection. The plasmids transfected were: GFP-pSG5, pSG5, NF-YA1wt (short isoform), NF-YA1m29 (DN short isoform), NF-YA13wt (long isoform), NF-YA13m29 (DN long isoform) using 5 μl of Lipofectamine 2000 (Invitrogen, Carlsbad, CA,, according to manufacturer's instructions. Cells were lysed 24 hours after transfection for RNA isolation and protein extract preparation. The experiments were repeated four times.

Protein Purification and Transduction

Plasmids encoding fusion proteins were a kind gift of Dr. S.G. Emerson, and the recombinant proteins were expressed in E. coli as soluble proteins, according to the protocol of Domashenko et al. [20]. GST-tagged fusion proteins were purified according to standard methods. Briefly, bacterial cells were lysed in Sodium Chloride-Tris-EDTA (STE) Buffer (10 mM Tris, pH 8.0, 150 mM NaCl, 1 mM ethylenediaminetetraacetic acid) and 1.4% sarcosyl. After sonication, Triton X-100 was added to 2%, and the lysate was left for 30 minutes at room temperature. Batch/gravity-flow protein purification was performed with glutathione Sepharose 4B (Pharmacia), according to instructions. Proteins were eluted and dialyzed against STE50 (10 mM Tris, pH 8.0, 50 mM NaCl, 1M methylenediaminetetraacetic acid), 10% glycerol, and stored at −80°C. Protein concentration was determined by the Bradford method (Bio-Rad Protein Assay). Cells were treated with the indicated amount of the fusion proteins in complete ES medium and kept at 37°C in 5% CO2. Fresh medium was added with the recombinant proteins at 24-hour intervals for several days. Alkaline phosphatase assays were performed with a colorimetric system using the AB0300 Kit (Sigma-Aldrich). The experiments were repeated three times.

Western Blot Analysis

Western blots of total and nuclear extracts of mES were performed according to standard procedures, with the primary antibodies and a peroxidase-conjugated secondary antibody (Amersham, Piscataway, NJ, Primary antibodies used: anti-NF-YB (Genespin), anti-NF-YA Mab1a, anti-NF-YC (home-made), anti-NANOG (Bethyl), anti-KLF4 (Genespin), anti-KLF5 (Santa Cruz Biotechnology, Santa Cruz, CA,, anti-SOX2 (Santa Cruz Biotechnology), anti-VINCULIN (Santa Cruz Biotechnology).


ChIPs were performed essentially as described previously [24]. Briefly, 5 × 107 cells were crosslinked using 1% formaldehyde for 10 minutes, the reaction quenched with 1/20 volume of 2.5 M glycine and centrifuged at 1,350g for 5 minutes; the pellet was washed twice with PBS and resuspended in sonication buffer, sonicated to obtain fragments of approximately 300–600 bp, as verified on agarose gel electrophoresis. Reactions were centrifuged at 20,000g for 10 minutes, and the supernatants were used for incubations with antibodies overnight at 4°C. 5 × 106 equivalents of chromatin were immunoprecipitated with 5 μg of anti-YB (Genespin), anti-SOX2 (Santa Cruz Biotechnology), anti-OCT4 (Abcam, Cambridge, U.K.,, anti-NANOG (Bethyl), and anti-FLAG (Sigma-Aldrich) antibodies. Protein-G beads (KPL) were used for recovery of antibody-bound chromatin. Crosslinking was reversed by incubation at 65°C overnight. Reactions were digested with RNAse A and Proteinase K and DNA purified by phenol–chloroform extraction and ethanol precipitation. DNA was resuspended in TE and used in quantitative polymerase chain reaction (qPCRs) reactions. A region of Satellite DNA was used as internal negative control. ChIP experiments were repeated three times. Re-ChIP experiments were performed as previously described [24].

Reverse transcriptase PCR and Real-time PCR

RNA was isolated by the Tryzol protocol, treated with DNAse (Roche Diagnostics, Basel, Switzerland,, purified with RNeasy mini kit (Qiagen, Hilden, Germany, and reverse transcribed using reverse transcription system (Promega, Madison, WI, The cDNA reaction was diluted 1:20 in TE. PCR primers were designed to amplify 100–200 bp fragments: sequences are available upon request. In addition to Gapdh, two housekeeping, NF-Y-independent genes—Pten and Nubp1—were used to normalize the results. Quantitative PCR was performed using SYBR green IQ reagent (Biorad) in the iCycler IQ detection system (Biorad). Reactions were run in triplicates, and the relative sample enrichment was calculated with the following formula: 2ΔCtx − 2ΔCtb, where ΔCtx = Ct input − Ct sample and ΔCtb = Ct input − Ct control Ab. Q-PCR was performed in triplicate for each of the biological replicates.


A Switch of NF-YA Isoforms During Differentiation of mES

The regulatory NF-YA subunit of NF-Y exists in two distinct isoforms that differ of 28 amino acids (Fig. 1A) and are present at different levels in various cellular contexts [17, 25]. NF-YC is also differentially spliced, generating 37 and 50 kD isoforms. We checked the mRNA and protein levels of NF-Y subunits, by quantitative reverse transcriptase PCR (qRT-PCR) and Western blot in mESCs, before and after differentiation to embryoid bodies, with and without addition of retinoic acid (Fig. 1B, 1C): the predominant NF-YA isoform in undifferentiated cells is NF-YAs, with NF-YAl becoming the only isoform after differentiation, while NF-YAs essentially disappears. NF-YB remains constant, the short 37 kD NF-YC, mostly found in undifferentiated cells, is switched off, unlike the 50 kD isoform, which is upregulated after differentiation. As expected, the levels of Sox2, Oct4, Nanog, and Klf4 were used to monitor differentiation, decrease, at the mRNA and protein levels (Fig. 1B, 1C).

Figure 1.

Switch of NF-YA isoforms during differentiation of mESCs. (A): Description of the two isoforms of NF-YA originating from alternative splicing of exon 3. (B): Quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) analyzed mRNA levels of different stem cells genes and of the three subunits of NF-Y after 11 days of differentiation by embryoid bodies' (EB) formation and treatment with retinoic acid (RA). Error bars represent the SD of the q-PCR replicates. (C): Western blot analysis of the indicated protein levels in the conditions indicated in (B). Abbreviations: EB, embryoid body; mESCs, mouse embryonic stem cells; RA, retinoic acid.

NF-YA DN Mutants Differentially Affect Expression of Stemness Genes

Our lab has generated a DN NF-YA (YAm29) in which a mutation in the DNA-binding subdomain, while still allowing the binding to the HFD subunits, renders the trimer unable to recognize the CCAAT box. Upon overexpression, YAm29 leads to elimination of endogenous NF-Y binding and thus lack of activation of CCAAT-dependent promoters. YAm29 was originally based on NF-YAl and used in several studies proving to be diagnostic of specificity of NF-Y involvement [reviewed in11, 23]. To assess the role of NF-YA isoforms in transcription of stem genes, we first generated the NF-YAs version of the YAm29 and transfected mES E14 cells with wt NF-YAs, NF-YAl and their respective DN counterparts. First, we assessed the expression of endogenous Sox2, Oct4, Nanog, and Klf4 mRNAs by qRT-PCR. Figure 2A shows that the genes are affected, with a more pronounced decrease following YAs-m29. This was even more evident on the expression of several important mES regulatory genes, such as Arid1a, Fgf4, Sall4, Klf5, and Jarid2, showing a reduction only with YAs-m29 (Fig. 2B). All these genes have CCAAT boxes in canonical promoter positions and are bound by NF-Y in Chip-Seq experiments (D. Dolfini, R. Mantovani, unpublished data). Expression of Suz12, Rif1, and Ccnb1 decreased with both DN isoforms, the CCAAT-less Mll1 and Check1 with neither. Note that the expression levels of the two transfected DNs were similar (Fig. 2C) and that the differential behavior was observed at different doses of transfected DNAs (D. Dolfini, R. Mantovani, not shown). The DN effect on SOX2 protein was confirmed in extracts of transfected cells, whereas the levels of NANOG were minimally affected (Fig. 2C). Only Sox2 was previously shown to have a bona fide NF-Y site in the promoter, hence we verified whether the DN effect was direct on the other genes. Figure 2D shows ChIP analysis of Sox2, Oct4, and Nanog promoters with antibodies against NF-YB, SOX2, OCT4, and NANOG: NF-Y was positive on Sox2 and, to lesser degree, on Oct4, not on Nanog; SOX2 was found on Oct4 and Nanog, NANOG on Sox2 and Oct4, data which are in accordance with many previous studies. These findings indicate that Sox2 and Oct4 are direct targets of NF-Y in mESCs and suggest an indirect effect of NF-Y on expression of Nanog.

Figure 2.

NF-YAs dominant negative regulates transcription of stem cells genes. (A): Mouse embryonic stem cells transfected with NF-YAs (upper panel) and NF-YAl (lower panel) and their dominant negative versions. mRNA levels of Sox2, Oct4, Nanog, and Klf4 were evaluated by qRT-PCR 24 hours after transfection. (B): mRNA levels of additional stem cells genes. Error bars represent the SD of the q-PCR replicates (18 analyses). (C): Western blot analysis of protein levels of overexpressed NF-YA 24 hours after transfection with wt NF-YAs, NF-YAsm29, wt NF-YAl, and NF-YAlm29. The levels of SOX2, KLF4, and KLF5 are indicated. (D): Chromatin immunoprecipitation was performed with the indicated antibodies (control FLAG, NF-YB, SOX2, OCT4, NANOG) in mES cells. Quantitative polymerase chain reaction (qPCR) analysis was performed with primers that monitor regions in the core promoters of Sox2, Oct4, and Nanog. Error bars represent the SD of the qPCR replicates (nine analyses)

Transduced NF-YAs Stimulates Expression of Stemness Genes and mES Proliferation

The effect of YAs-m29 on expression of key stem cell genes prompted us to focus on these subunits in overexpression assays. There are two major drawbacks in conducting these experiments in mESCs: (a) NF-YA half-life upon transfections is short, with the overexpressed protein disappearing within 24 hours [26]; (b) attempts at stable expressions were repeatedly unsuccessful. An alternative, appealing approach was transfecting a recombinant TAT-fusion protein, a strategy used in HSCs by the Emerson's group: a functional GST-TAT-NF-YAs was delivered in the nucleus thanks to the TAT epitope [20]. We produced GST-TAT-NF-YAs and the control GST-TAT proteins from E. coli and purified them (supporting information Fig. S1). We treated several types of human and mouse cells, and indeed observed rapid entry into nuclei, in accordance to the Domashenko et al.'s protocol (D. Dolfini, M.M., unpublished data). There are two issues in mES transfections: the first is that the original protocol prescribed multiple additions—six, over a period of 90′—in serum-free medium, which would be unpractical in mESCs; the second is that repeated transfections over several days, untested so far, would be desirable. Figure 3A shows a dose-response of repeated transfections of a single dose of TAT-NF-YAs in complete mES medium, adding the protein at 24-hour intervals for 4 days; nuclear extracts were prepared and assayed in Western blots: the recombinant proteins translocated efficiently in the nucleus at nanomolar concentrations and the effect was observed for days. At 100 nM concentrations, a significant amount of TAT-NF-YAs was found in the cytoplasm as well (not shown), and we decided to pursue our experiments at the concentration of 50 nM. After 4 days of treatment of mESCs with the control GST-TAT and TAT-NF-YAs, we checked the mRNA expression of endogenous stem cell genes by qRT-PCR: essentially all CCAAT-dependent genes were specifically upregulated by TAT-NF-YAs (Fig. 3B). The magnitude of the increase was well in line with many studies using DNA transfections [see11, 16 for review]. Note the increase in the genes of the esBAF complex, NF-YB/NF-YC and NF-YAs, the latter in accordance to the increased protein levels observed in the Western blots of Figure 3A. The control CCAAT-less Nes gene was not affected. Next, we checked the protein levels of SOX2, NANOG, KLF4, and KLF5 at 96 hours post-TAT-NF-YAs treatment: the increase was widespread and quite relevant for SOX2 and NANOG (Fig. 3C; quantitation in supporting information Fig. S2). Because of these data, we assessed the effect of TAT-NF-YAs on mESCs proliferation: compared to addition of GST-TAT, TAT-NF-YAs progressively increased proliferation rates (Fig. 3D). In conclusion, protein transfections with TAT-NF-YAs are feasible in complete medium, the protein enters nuclei efficiently and rapidly, it is functional, as it activates important CCAAT-dependent stem cells genes, and it stimulates growth of mESCs.

Figure 3.

Effects of TAT-NF-YAs transduction on expression of stem genes and cell proliferation. (A): Titration of recombinant TATNF-YAs transfected in mouse embryonic stem cells (mESCs). The addition of the recombinant protein was performed daily at the indicated concentrations. The levels of nuclear protein were monitored by Western blot analysis with anti-NF-YA antibodies in nuclear extracts. (B): mRNA levels of stem cells genes, NF-Y subunits, and esBAF subunits genes were monitored by quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) 4 days after daily treatment with recombinant TAT-NF-YAs and the control GST-TAT at the concentration of 50 nM. Error bars represent the SD of the qPCR replicates (nine analyses). (C): Effects of transduction of different concentrations of TAT-NF-YAs on the levels of SOX2, NANOG, KLF4, and KLF5 checked by Western blot analysis with the respective antibodies, after 96 hours of daily treatment. (D): Growth curves of mESCs treated daily with 50 nM control GST-TAT or TAT-NF-YAs. The experiment was repeated three times and error bars represent the SD of the biological replicates. Abbreviation: TFs, transcription factors.

NF-YAs Counteracts LIF Withdrawal

LIF is required for ESCs to grow and maintain an undifferentiated state. We therefore analyzed whether NF-YAs sustains the undifferentiated state in absence of LIF. We treated mESCs with recombinant GST-TAT and TAT-NF-YAs for 6 days in the absence of LIF: as shown in Figure 4A, SOX2, NANOG, and KLF5 decreased after withdrawal of LIF, with or without addition of the control GST-TAT; cells treated with TAT-NF-YAs, instead, showed normal expression of the three stem cells markers (Fig. 4A; quantitation in Fig. 4A). LIF withdrawal indeed changed dramatically the morphology of the cells, with typical mES colonies disappearing, and addition of GST-TAT had no significant effect on the resulting flattened phenotype (Fig. 4B, upper panels). However, daily treatment with TAT-NF-YAs in absence of LIF visibly prevented the change in morphology, maintaining a stem-like phenotype. As a further indication, we performed alkaline phosphatase assays under the abovementioned conditions: positivity was scored in untreated mESCs grown in LIF; in LIF-less conditions, mESCs lost positivity, unless TAT-NF-YAs was added (Fig. 4B, lower panels). To further substantiate these findings, we evaluated the mRNA levels of stem cells genes by qRT-PCR (Fig. 4C): essentially all showed a decrease in expression under conditions of LIF withdrawal and recovered normal expression in cells grown with TAT-NF-YAs. We checked the levels of expression of components of the esBAF complex, shown to be important to maintain the pluripotent state by conditioning the genome for LIF/STAT3 signaling [27]: the CCAAT Smarca4 and Smarcc1 indeed increased upon transduction of cells with TAT-NF-YAs. The effects were relatively modest only on Sall4 and Check1 and negligible on the CCAAT-less Nes and Mll1 genes. As further controls, we evaluated the expression of two bona fide CCAAT genes, bound by NF-YA in vivo only in differentiated cells, IGF2 and MSX1 [23]: their levels were extremely low in undifferentiated mESCs, as expected, increased upon LIF withdrawal, with or without GST-TAT, but were kept very low in the presence of TAT-NF-YAs; therefore, the activity of the short NF-YA isoform is exerted on the NF-Y regulome active in mESCs. Taken together, these data indicate that NF-YAs play a crucial role in the maintenance of stemness of mESCs.

Figure 4.

TAT-NF-YAs counteracts withdrawal of LIF. (A): Protein levels of NF-YA, SOX2, NANOG, and KLF5, 6 days after withdrawal of LIF and treatment with control GST-TAT recombinant protein or with TAT-NF-YAs at a concentration of 50 nM. (B): Morphology of colonies of mouse embryonic stem cells (mESCs) treated as indicated (upper panel). Alkaline phosphatase (AP) staining of mES colonies 8 days after treatment as indicated (lower panel). (C): mRNA levels of stem cells regulators were evaluated by quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) 6 days after withdrawal of LIF and treatment with the recombinant proteins, control GST-TAT, or TAT-NF-YAs. Error bars represent the SD of the qPCR replicates (nine analyses). Abbreviation: LIF, leukemia inhibitory factor.

NF-Y Is Required for NANOG DNA-Binding in CCAAT Promoters

To understand the transcriptional role of NF-Y in ESCs, we felt important to correlate genomic binding of NF-Y to that of mESCs TFs. We analyzed the sites of several TFs—E2F1, NANOG, KLF4, SOX2, ESRRB, STAT3, OCT4, and CTCF—for which ChIP-Seq data in mESCs are available [6, 27, 28], including NF-Y [23]. We retrieved the promoters of genes bound by these TFs and scanned them with Pscan; using the latest NF-Y positional sequence frequency matrix, we have gathered from genomic studies [11], at a score of 0.8, which recovers most bona fide NF-Y sites. The percentage of NF-Y+ promoters is high for E2F1 (35%), NANOG (47%), KLF4 (30%), and SOX2 (43%); p-values were quite relevant for the first three TFs and the majority of CCAAT boxes were indeed bound by NF-YA in mESCs in vivo (Fig. 5A). Note that this is most likely an underestimate of NF-Y binding, since using a more efficient NF-YB antibody, two to three times more NF-Y sites can be recovered in ChIP-Seq experiments (R. Mantovani, unpublished data). The interplay with E2F1 was expected, as this TF is genetically linked to NF-Y [reviewed in29]. The correlation with NANOG, instead, was new, and considering that almost half of NANOG targets overlapped an NF-Y bound in proximity, we decided to pursue the analysis. TFBS search in NANOG peaks with the TRANSFAC and JASPAR databases confirmed the presence of the NF-Y site at the top of the list (Fig. 5B, left panel). We then turned to the larger collection of NANOG sites described in human ESCs [28]: as shown in Figure 5B (right panel), the NF-Y matrix is also enriched, with E2F1, MIZF, ELK4, and GABPA: while not at the top of the list, it does have a very significant p-value. To further substantiate and define these data, we performed de novo motif discovery with the Weeder software [30], interrogating sequences within 100 bp of the NANOG ChIP-Seq promoter peaks: the results in Figure 5C indicate that the CCAAT sequence, in either orientation, comes as third in the ranking of the mESCs dataset (upper panel) and first in the larger dataset of hESC NANOG peaks (lower panel). These results were also confirmed by Meme analysis (data not shown). Tentatively, at least a third of NANOG sites in hESCs have an overlapping NF-Y site. These bioinformatics data strongly indicate an excellent in vivo correlation between the NF-Y and NANOG promoter sites in mouse and human ESCs.

Figure 5.

NF-Y and NANOG genomic correlation. (A): Promoters bound by mouse embryonic stem cells (mESCs) TFs [6] were scanned with the NF-Y matrix [11] using the Pscan alghoritm. On the left part of the Table, the number of promoters bound by the TFs with a NF-Y consensus with a score higher than 0.8 [11] and the calculated p-values. On the right part, comparison of predicted in silico data with in vivo ChIP-Seq data of NF-YA derived from Tiwari et al. [21], and the calculated p-values. (B): TFBS analysis of mouse [6] and human [28] NANOG loci in promoters with Pscan, using JASPAR and TRANSFAC matrices. (C): De novo motif discovery analysis using Weeder, on 100 bp large NANOG peaks located in promoters of human and mESCs. The NF-Y site emerged in the two species with slightly different scores. Abbreviations: TF, transcription factor; TFBS, transcription factors binding sites.

NF-YA is known to synergize with several TFs, by helping their DNA binding [11]. To help define the relationship between NF-Y and NANOG, we transiently transfected mES with wt NF-YAs and YAs-m29 and used chromatin to perform ChIPs with anti-NF-YB and NANOG antibodies. The results show the expected negative effect of the DN-NF-YAs on binding of NF-YB to CCAAT promoters that are positive for NANOG in ChIP-Seq data (Fig. 6A, Upper Panels). As for NANOG binding, it is decreased with YAs-m29 in all NF-Y-bound promoters tested, with the exception of Fam33a (Fig. 6A, Lower Panels). In the CCAAT-less Lefty promoter, NANOG binding was unaffected. To further substantiate the coresidency of NF-Y and NANOG on common promoters, we performed sequential ChIPs—Re-ChIPs—on chromatin from mESCs: the results shown in Figure 6B indicate that NF-YB-NF-YB Re-Chip was enriched in CCAAT, with respect to the NF-YB-Flag control, as expected since it is our internal control. Moreover, in NF-Y and NANOG common sites, the NF-YB-NANOG Re-ChIP was also enriched, whereas the CCAAT-less Lefty was negative. Taken together, our data indicate that NF-Y binding to CCAAT boxes is required for binding of NANOG to up to half of its targets in mESCs.

Figure 6.

NANOG binding to CCAAT promoters requires NF-Y. (A): ChIP analysis with indicated antibodies of mouse embryonic stem cells (mESCs), transfected with wild-type NF-YAs (light gray bars), or with the dominant negative NF-YAsm29 (black bars). The targets analyzed are based on ChIP-Seq data of Nanog [6] and NF-YA [23]. (B): NF-YB and Nanog ReChIP of mESCs. Quantitative polymerase chain reaction (qPCR) data were normalized for the NF-Y-negative control Lefty (see (B)) and satellite DNA. Error bars represent the SD of the qPCR replicates (nine analyse s). Abbreviation: ChIP, chromatin immunoprecipitation.


Alternative Splicing of NF-Y Subunits

Almost all mammalian genes are differentially spliced. It is no exception for TFs, of course, and it has long been known that a gene can produce an activator of transcription, as well as a repressor, if the TA domain is omitted, or if a repressor domain is incorporated. In other cases, protein–protein interaction domains are alternatively included, so that heterodimerizing or interacting partners are alternatively chosen. In the case of NF-YA, the difference between the two isoforms is rather modest: 28 amino acids, composed of Gln and hydrophobic residues within a large—180 amino acids—TA domain with a similar amino acids composition [17]. This is one of the two TA domains of the trimer, the other being incorporated in the C-terminal domain of NF-YC. Western blots of different lines, mostly derived from transformed tumor cells, showed the prevalence of one isoform, but never a clear-cut absence of the other, and it was impossible to rationalize the relative abundance relating it to tissue- or cell-type distribution, cell-cycle phase, growth or differentiation status. The two isoforms are identical in terms of interactions with the HFD dimer and affinity for CCAAT boxes. In GAL4-based assays and in vitro transcription, the extent of NF-YAl and NF-YAs brute activation potential, which is really the only discernible readout of those assays, was about the same [31, 32]. Truly, when constructs were transfected in different cellular contexts, each isoform performed better in a given cell type, but this was subtle and certainly not an all-or-nothing effect [17]. The first evidence that the two isoforms are not functionally identical came from studies on the Cystathionine synthase promoter: expression of Sp1, another Q-rich TF, showed clear cooperativity in activation only with NF-YAl [33]. Further strong evidence of a specific function of NF-YAs in mouse HSCs followed [19].

We provide here four lines of experiments pointing at a crucial role of NF-YAs in mESC biology. First, we confirmed the switch between NF-YAs and NF-YAl upon differentiation of mESCs [8]: NF-YAl is present at low levels in undifferentiated mESCs, possibly because these cultures are not entirely homogeneous, and it somewhat increases upon differentiation; the sharpest difference is observed in NF-YAs, which is the major isoform before and drops to undetectable levels after differentiation. Second, we show that the two DN NF-YA proteins have differential behaviors on the expression of several key stem cells genes, notably a profound negative effect of YAs-m29 on Klf4, Klf5, Arid1a, Fgf4, Sall4, and Jarid2. For other genes, the effect of the DNs was similar. We take this as an indication that the two isoforms are not interchangeable and that they have target preference within the large NF-Y regulome. This conclusion is further substantiated by the lack of activation by NF-YAs of bona fide NF-Y targets active after differentiation. Third, protein transfections of TAT-fusion NF-YAs lead to rapid and efficient nuclear uptake of the recombinant protein; TAT-NF-YAs is functional, as it activates CCAAT-containing stemness genes, stimulating accelerated growth of mESCs in normal conditions. This effect might be a reflection of a role of NF-YAs in preventing cells from exiting from the stem pool within the culture. Fourth, TAT-NF-YAs transfections lead to the maintenance of stem features in the absence of LIF, also as a result of activation of CCAAT stemness genes. The latter result is largely assumed to be typical of TFs belonging to the “core” ESC circuitry, as similar results were obtained with KLF4, TCF3, and MYC, [34–36]. It is therefore reasonable to conclude that NF-YAs is crucial for mES stemness, a result that could have consequences beyond this totipotent system: the precedent of mouse HSCs suggests that other tissue- or organ-specific stem cells use this isoform predominantly.

Another relevant observation comes from the analysis of the NF-YC subunits. NF-YC has two major isoforms of 50 and 37 kDa, also generated by alternative splicing in the TA domain, whose expression varies in different cell lines [25]; we previously reported that NF-YC 50 and NF-YAs levels are linked, and, similarly, NF-YC 37 is linked to NF-YAl, in several cell lines of different origin. Furthermore, the two combinations are differentially able to activate reporter genes, teaming up with the appropriate partner. In mESCs, however, we find that NF-YC 37 is coupled to NF-YAs, both disappearing upon differentiation: the NF-YAs/NF-YC 37 combination is apparently unique, and behaving in a coordinate manner. It is therefore possible that, fine tuning of the NF-YC isoforms is as important as NF-YA to maintain stemness.

Mechanisms of NF-Y Activity in Stem Cells

It is clear that the stemness enhancing properties of NF-YAs reside in the targeted genes. As there are no isoform-specific antibodies available to perform ChIP-Seq or is it not possible to functionally inactivate either isoform selectively (D. Dolfini, R. Mantovani, unpublished data), we related the NF-Y regulome to those of other key TFs. This exercise indicates a strong link between NF-Y and E2F1, SOX2, and NANOG. The intersection with E2F1 sites was largely expected, given the (a) the correlation of the two TFBS in large datasets of tumor profilings [37], (b) the genetic interplay between NF-YA and E2F1 [reviewed in28]. Finding NF-Y sites in SOX2 peaks was also not surprising, since many times the TFs targeted by NF-Y at the transcriptional level are also synergistic partners in specific developmentally regulated or stress-inducible groups of genes. The interplay with NANOG, instead, is new, and somewhat unexpected. The NF-Y logo was at the top of the list of TFBS analysis in NANOG peaks and clearly emerged with unbiased methods of sequence sorting (weeder, meme). The best matrix emerging from de novo motifs analysis in mESCs and hESCs is manifestly an NF-Y site. There might be some skewing, as two purines at the 5′ of CCAAT, which are usually well represented in NF-Y sites [11, 38] and required for efficient in vitro binding [38] are not enriched, unlike two/three nucleotides at the 3′ end, CA(G), which are clearly enriched. NANOG is an homeodomain TF shown to recognize the motif TAAT(G/T)(G/T) [39], which is one nucleotide—underligned—different from the core NF-Y consensus ATTGG: interestingly, in one of the ChIP-on-Chip location analysis reported, the NANOG motif was almost identical to an inverted CCAAT box [2]. The subtle difference in the NF-Y logo at NANOG sites might suggest that the latter binds immediately flanking, at the 5′ or 3′ of NF-Y, selecting for additional sequences that are only present in a subset of CCAAT sites. This is not unprecedented for NF-Y, which appears to be a promoter architectural organizer, as well as a facilitator of nearby TFs association in many different systems [11]. Interestingly, the Nanog promoter is not a direct NF-Y target, but it is known to be affected by SOX2 and OCT4, which are controlled by NF-Y. In general, the NF-Y-NANOG interplay is observed at two levels: NF-YAs indirectly elevates Nanog mRNA and protein levels, and, most importantly, favors NANOG binding to a sizeable cohort of promoters: in fact, elimination of NF-Y binding by the YAs-m29 leads to removal of NANOG from the large cohort of genomic locations containing a CCAAT box, thus establishing a hierarchy in which NF-Y binding is a prerequisite. It is unclear how the interplay occurs mechanistically, whether the two proteins interact physically or indirectly via other polypeptides. The availability of NANOG [39] and NF-Y crystal structures with DNA [M. Nardini, N. Gnesutta, G. Donati, M. Bolognesi, R. Mantovani, manuscript in preparation] could enable us to expand on these findings in the future.

Finally, we notice a good parallel between NF-YAs and SALL4, another TF important for ESCs [3, 40, 41], shown to have effects similar to NF-YA on HSCs, including after protein transfections [40, 42]. SALL4 is also present in two isoforms, and one is specifically important in ESCs as part of the stem TF circuitry [43], which involves co-occupation of NANOG sites [44]. SALL4 is involved in leukemogenesis as well as in other types of cancer [45]. Interestingly, SALL4 is also capable to enhance reprogramming of differentiated cell into ES-like iPS cells [46], an aspect of NF-YA that has not been studied so far: the data presented here should certainly spur investigations in this direction.

Implication for Cancer?

An accepted model of cancer development is that it originates from the transformation of tissue or organ stem cells into cancer stem cells. Many of the ES TFs are known to play a crucial role in maintenance of cancer stem cells. By comparing gene expression patterns in tumors versus normal tissues, de novo motifs discovery identified CCAAT and E2F sites as highly enriched in regulatory regions of genes overexpressed in tumors [37]; studies on TFBS in profilings of diverse types of tumors came to the same conclusions, particularly in the most aggressive cohorts [reviewed in16]. It seems therefore fair to conclude that the NF-Y regulome is important for cancer cells. Might the data presented here have implications for cancer cells? One could hypothesize that the NF-YAs isoform might be important for cancer stemness. The recent results with conditional KO mice in liver and hematopoietic cells indeed indicate that NF-YA is important for tissue maintenance, possibly by regulating stem cells [14, 15]. Hematopoietic cancers arise mostly not from HSCs but from lineage-specific stem cells, which presumably already underwent a partial switch in isoforms: hence the presence of both NF-YA isoforms in hematopoietic-derived cell lines. Within these cells, there could be a partition of duties in activation of the NF-Y regulome between NF-YAl (differentiation genes) and NF-YAs (stemness genes).

Finally, we introduced important practical aspects on protein transfection experiments, particularly concerning the feasibility and efficiency of procedures with TAT-NF-YAs [20]: we confirm and extend these data by showing that TAT-NF-YAs enters nuclei very efficiently—in the nanomolar range—and rapidly—minutes after addition—, and that treatment can be pursued for days at 24-hour intervals in complete medium. Transfections of proteins fused to cell penetrating peptides are an increasingly popular way to deliver gene products without gene transfer [47]. The previously described expansion of the HSCs compartment by treatment with TAT-NF-YAs and better engraftment of bone marrow transplantations (BMT) into immunocompromised donor mice [20] holds great promise for therapeutic intervention. In fact, BMT is currently the medical treatment of choice in several hematologic malignancies, so the potential clinical exploitation of NF-YAs transfections is considerable. At the same time, however, this could be a two-edged sword with potential dangers: TAT-NF-YAs could enhance the proliferative potential of residual cancer stem cells in the treated population, by activating “stemness” genes, some of which have been identified here. The procedures and reagents described here should enable us to explore this key concept in the future.


In summary NF-YA belongs to the restricted circle oPTFs that govern mESCs and its short isoforms play a crucial role in maintaining stemness.


We thank S. Emerson (U. of Pennsylvania, USA) for the TAT-NF-YAs constructs, Dr. V. Broccoli for the E14 mES line, and member of the lab for helpful discussions. The work is supported by Grant Nepente from Regione Lombardia.


The authors indicate no potential conflicts of interest.