Background
Unprecedented insights into the global interaction of transcription factors with DNA, often in a tissue-specific context, have become available consequent to next generation sequencing technologies. It is necessary to understand the complex interplay between DNA sequence, protein structure, and protein-protein interactions (PPIs) in determining gene regulatory pathways. The
Ets (
E-
twenty-
six) family of transcription factors, characterised by the presence of an evolutionarily conserved 85 amino acid (aa)
Ets DNA-binding domain, utilises a range of factors to govern target specificity.
Ets proteins are classified into subfamilies based on sequence similarity in the
Ets domain and by flanking domains, which can determine whether they act positively or negatively as transcriptional regulators. In humans, 27 members of the
Ets family have been characterised, and many function as critical mediators of a wide variety of cellular processes, which include embryonic development, differentiation, growth, apoptosis, and oncogenic transformation [
1‐
3].
The
Ets domain forms a winged helix-turn-helix structure that binds the core
Ets motif 5′-GGAA/T-3′ [
4,
5]. Outside of the core sequence, the
Ets domain has high tolerance of variations in its target sequence [
6]. A key question is how
Ets proteins orchestrate DNA binding specificity to regulate specific biological processes. Analysis of individual
Ets family member DNA binding sites has indicated that specific as well as redundant occupancy may occur at
Ets sites throughout the genome [
7]. Subtle differences in
Ets sites, tissue-specific expression of
Ets factors and their co-factors, and differential signalling responses may all contribute to their distinct functions, but makes identifying true targets both problematic and challenging [
8,
9].
Certain
Ets proteins are known to play important roles in haemopoietic development via transcriptional regulation. Knockout mouse models have helped unravel the functional importance of
Ets proteins in haemopoiesis. Loss of PU.1 (SPI1) has a profound effect on haemopoietic development by affecting myeloid and B cell development [
10,
11]. Other
Ets gene knockout mouse models with defects in haemopoietic cells include
Ets1 [
12,
13],
SpiB [
14],
Fli1 [
15], and
Etv6 [
16]. Members of the ELF (E74-like factor) subfamily of Ets transcription factors including ELF1, ELF2, and ELF4 also play important roles in the development of lymphocytes and regulate numerous haemopoietic-specific genes. ELF1, which regulates genes involved in T cell development such as CD4 [
17], CD3ζ [
18], and IL-2 [
19], also plays a restricted role in natural killer T cell development [
20]. ELF4 (MEF; myeloid ELF-1-like factor) distinctly plays a critical role in the development and function of natural killer cells [
21]. ELF2, also known as NERF (new Ets-related factor), is the least characterised member of this subfamily, despite its identification by two independent groups over 20 years ago [
22,
23]. ELF2 binds to the regulatory regions of genes involved in lymphocyte development and function including B and T cell co-receptor proteins, tyrosine kinases, and enhancer regions [
23‐
25]; and in many instances, is shown to modulate their expression levels. A knockout mouse model for ELF2 has not been reported, so little is known about its functional role in haemopoietic development.
Two major isoforms of ELF2 arise from alternative promoter usage, ELF2A (NERF-2), and ELF2B (NERF-1) [
23]. These major isoforms of ELF2 can exhibit opposite regulatory effects, ELF2A activates whilst ELF2B represses expression of its target genes [
24]. Importantly, both isoforms interact with the master haemopoietic regulators RUNX1 and LMO2 [
22,
24]. Whilst both isoforms can bind the same
Ets target sites in DNA and bind common co-factors, little is known about what functional differences these ELF2 isoforms may have.
In this report, we established reagents to distinguish between ELF2 isoforms and showed that ELF2 isoforms are differentially expressed. Our overexpression studies comparing between the ELF2 isoforms and the related ELF family members ELF1 and ELF4 in primary and transformed cell lines demonstrated a proapoptotic role for ELF2B which was modulated through its N-terminus. We then explored the role of ELF2 isoforms in haemopoietic development using an in vivo bone marrow reconstitution model in Rag1
-/- mice. Our results show a defined effect on B and T cells as well as granulocytes, consistent with a potential role for ELF2 in regulating haemopoietic development.
Methods
Vector construction
Full-length human ELF1 and ELF4 cDNAs were obtained from cDNA prepared from human thymus total RNA whilst ELF2A and ELF2B cDNAs were obtained from cDNA prepared from human testis total RNA (FirstChoice® Human Total RNA Survey Panel, Ambion). Mouse Elf2 isoforms were amplified from cDNA prepared from mouse testis RNA. Each full-length cDNA sequence was then cloned into the pcDNA3.1-HA expression vector containing a haemagglutinin (HA) tag on the N-terminus using
NotI and
XbaI sites. ELF2Δ, representing the common 513 aa region of ELF2 isoforms was amplified from ELF2A using primers with
NotI-5’ and
ClaI-3’ ends and was cloned into pcDNA3.1-HA. To construct lentiviral vectors, each HA-tagged ELF gene was subcloned into pCCLteteGFP-2A lentiviral vector [
26] via
BmgBI and
ClaI sites. To construct retroviral vectors, each Elf2 isoform was subcloned into the pMIG retroviral vector upstream of the IRES sequence via
BamHI and
PmeI sites. Primer sequences used for cloning are available on request.
Cell culture
HeLa, HEK293T, MPRO, and GP + E86 ecotropic retrovirus packaging cells were cultured in DMEM (MPRO with 10% (v/v) conditioned DMEM medium from BHK-HM5 cells secreting GM-CSF). K562, Jurkat, A20 and CH12 cells were grown in RPM1 1640 medium (A20 and CH12 cells with the addition of 50 μM β-mercaptoethanol (Sigma-Aldrich). All basal media were supplemented with 10% FCS (v/v), penicillin (100 U/mL), and streptomycin (100 μg/mL). Human foreskin fibroblast (hFF) cells were grown in Ham’s F-12K (Kaighn’s) media supplemented with 50 μg/mL ascorbic acid (Sigma-Aldrich), 5 ng/mL basic fibroblast growth factor (PeproTech), 1 μg/mL hydrocortisone (Sigma-Aldrich), 5 μg/mL bovine insulin (Sigma-Aldrich), and 20% v/v FCS. All cell lines are routinely tested for Mycoplasma contamination by PCR screening of genomic DNA isolates.
Lentivirus and retrovirus production
Lentiviral particles were produced using a four plasmid tat-independent packaging system delivered into cells by calcium phosphate transfection [
27]. At approximately 16 h post-transfection, the medium was replaced with fresh DMEM supplemented with 5 mM sodium butyrate. The media was collected after 24 h, and the virus-containing media was filtered through a 0.45-μM filter (MillexHV Millipore) to remove cell debris. Viral concentration was achieved by centrifugation at 20,000
g for 2 h at 4 °C in a Beckman L8-70M Ultracentrifuge using an SW28 rotor (Beckman). Following centrifugation, the supernatant was removed, and the viral pellets were resuspended in 1/100th of the original volume in DMEM/10% FCS. Viral titres were determined by testing transduction levels on HeLa cells using serially diluted virus. Cells were collected 48 h post-transduction and analysis by flow cytometry using an LSR Fortessa (BD). Percentages of GFP-positive cells at each virus dilution were evaluated using FlowJo version 9.4 (Treestar).
Gene expression analysis
Total RNA was extracted from mouse tissues or immortalised cell lines using TRI Reagent (Astral Scientific). Each RNA sample was first treated with DNase I before generation of oligo dT cDNA by reverse transcription using SuperScript III (Invitrogen). After each RT reaction, the samples were treated with RNase H (New England BioLabs). Gene expression levels were quantified using the CFX96 Touch™ Real-Time PCR Detection System (BioRad) in 10 μL reactions, containing 25 ng of cDNA template, SYBR green-containing iQ Master Mix buffer (BioRad), 300 nM of forward and reverse primers (Additional file
1: Table S1), and UltraPure™ DNase/RNase-Free distilled water (Invitrogen). Reaction conditions include: denaturation at 95 °C for 2 min, 30 amplification cycles at 95 °C for 10 s, 60 °C for 20 s, and 72 °C for 20 s, and melt curve analysis at 72 °C for 10 min.
RNAseq data was trimmed by Trim Galore using the default Illumina Adapter Sequences. The trimmed reads were mapped to the Ensembl mouse transcriptome GRCm38.73 (mm10) using the default settings of TopHat 2.0.8. The FPKM was then calculated using mapped reads by Cufflinks v2.1.1 under default settings. Analysis of genomic regions surrounding the transcription start site of ELF2 isoforms for putative transcription factor binding sites was performed using MatInspector (Genomatix). Experimentally validated transcription factor binding sites were obtained from UCSC and Ensembl browsers by viewing publicly available ChIPseq datasets. Alignments to determine conservation in genomic DNA and protein sequences were performed using orthologous sequences obtained from Ensembl and aligned using the ClustalW algorithm within MacVector. Prediction of NLS sequences was performed with SeqNLS. Protein disorder analysis was performed using the PONDR server.
Chromatin immunoprecipitation (ChIP)
For each ChIP, 5 × 10
6 HEK293T cells transfected with pCCLteteGFP, pCCLteteGFP-2A-HAELF2A, and pCCLteteGFP-2A-HAELF2B were cross-linked with 1% (w/v) formaldehyde for 10 min and were quenched with 1 M glycine to a final concentration of 20 mM. Nuclear lysates were sonicated for 25 cycles, 30 s on, 30 s off using a Bioruptor sonicator (Diagenode). Antibodies for immunoprecipitating protein/DNA complexes include: acetylated H3K9/K14 (#9677, Cell Signaling); CTCF (07-729, Millipore); and HA (ab9110, Abcam). Protein G-conjugated agarose beads (Millipore) were used to immunoprecipitate antibody-bound chromatin complexes, and all subsequent steps were performed according to the manufacturers’ instructions. After de-crosslinking, phenol/chloroform extraction, and ethanol precipitation, PCR was performed on genomic DNA targets using Phusion polymerase with GC buffer (Finnzyme). Primer sequences are in Additional file
1: Table S1.
Antibody production and purification
Isoform-specific antibodies were raised to recognise the N-termini of ELF2A (aa 2-19) and ELF2B (aa 2-19). Each peptide was synthesised and conjugated to keyhole limpet hemocyanin (KLH) by Mimotopes (Victoria, Australia) and then sent to the Institute of Medical and Veterinary Science (IMVS, Adelaide, Australia) for a series of rabbit immunisations performed according to their standard operating procedures and approved institutional animal ethics protocols. The antiserum collected from the final bleed was used for subsequent affinity purification procedures. Antibodies were purified from crude rabbit serum using thiopropyl sepharose 6B (GE Healthcare) according to the manufacturers’ instructions.
Western analysis
Cell lysates were prepared using a whole cell lysis buffer (20 mM Tris-Cl pH 7.6, 150 mM NaCl, 1% (v/v) Triton X-100, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS). Nuclear and cytoplasmic fractionation was performed using the NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Scientific), as per the manufacturers’ instructions. Protein samples were denatured at 90 °C for 10 min with 100 mM DTT in NuPAGE® LDS sample buffer. Samples were separated using a 4–12% NuPAGE® Novex® Bis-Tris mini gel (Invitrogen) and transferred onto PVDF membrane (Millipore) using Trans-Blot® SD Semi-Dry Transfer Cell (BioRad). Each blot was then probed with antibodies specific to the protein of interest (Additional file
2: Table S2).
FACS
To prepare cells for fluorescence-activated cell sorting (FACS), single cell suspensions of cultured cells were filtered to remove cellular debris and aggregates and then were resuspended in 400 μL of PBS containing 2% (v/v) FCS and 5 μg/mL PI. Transduced GFP-positive cells were purified (to > 95% purity) using a BD Influx into sterile 5 mL polystyrene FACS tubes.
Cell biology assays
For colony-forming assays, FACS-enriched cells were plated in triplicate at 1000 cells/10 cm plate and were incubated for 14 days with media replaced every 5 days. Cells were fixed with 5 mL of ice-cold methanol for 10 min. Plates were air-dried and stained for at least 2 h with Giemsa solution diluted 1:20 in distilled water. Colonies were scored on a digital colony counter (Labserv Technologies). For cellular proliferation assays, FACS-enriched cells were seeded in triplicate wells at 200–1000 cells/well in a 96-well plate in 100 μL of media. Proliferation was assessed every 2 days for a total of 10 days or daily for 4 days. At each time-point, proliferation was measured by MTT assay (Chemicon) according to the manufacturer’s instructions and absorbance was measured by spectrophotometry at 572 nm using a POLARstar Omega microplate reader (BMG Labtech). For cell cycle analysis, BrdU (150 μg/mL diluted in medium) was added to approximately 1 × 106 cells and incubated for 4 h. Cells were rinsed twice in PBS, detached from plates using TrypLE™, fixed, and stained for BrdU incorporation using the APC BrdU Flow Kit (BD Bioscience), following the manufacturer’s instructions. The cells were subsequently incubated in 7-AAD (BD Bioscience) and were analysed on a Canto-II flow cytometer (BD). For cell division analysis, approximately 3 × 105 cells were labelled with 10 μM of CFSE Cell Trace Violet (Invitrogen), according to the manufacturers’ instructions. Following CFSE labelling, cells were resuspended in media and were divided equally into plates containing media with or without doxycycline (1 μg/mL). Cells were allowed to proliferate for 4 days and were subsequently analysed on a flow cytometer. All flow cytometry data was analysed using FlowJo software (Treestar).
Apoptosis assays
For assessment of Annexin V staining, approximately 1 × 106 cells were labelled with Annexin V reagent conjugated to Pacific Blue or APC fluorophores (BioLegend). Cells were incubated on ice for 1 h, rinsed, resuspended in 200 μL of binding buffer (10 mM HEPES pH 7.4, 140 mM NaCl, 25 mM CaCl2) containing PI, and analysed on an LSR Fortessa (BD) flow cytometer. To measure caspase activation, cells were seeded at 1 × 104 cells/well in a 96-well plate and incubated in a 5% CO2-humidified 37 °C incubator for 24 h. Apoptosis was measured using the Caspase Glo 3/7 Assay (Promega) according to the manufacturers’ instructions. The luminescent signal was measured using a POLARstar Omega microplate reader. The luminescent signals measured were normalised to untransduced HeLa control, set as 1.0, for each experiment to account for signal variation between experiments.
Immunofluorescence staining
HeLa cells were seeded (2 × 10
4 cells/well) in 8-well chamber culture slides (BD Biosciences) and incubated overnight. The cells were fixed in 4% (w/v) PFA, permeabilised in 0.2% (v/v) Triton X-100, and blocked with 20% (v/v) BlokHen (Aves Laboratories). Cells were then stained with antibodies at optimal dilutions (Additional file
2: Table S2). The cells were stained with DAPI (1 μg/mL) before visualisation using a DM6000 microscope (Leica Microsystems).
Generation of retrogenic mice and haemopoietic cell analysis
Establishment of GP + E86 NIH3T3-based ecotropic packaging cell lines expressing pMIG retroviral vectors containing eGFP empty vector (control) or HA-tagged ELF2 isoforms was performed as described [
28]. Transduction of mouse bone marrow and generation of retrogenic mice was performed as described [
28]. For analysis of haemopoietic cell populations, single cell suspensions were stained with relevant antibodies diluted in FACS buffer (PBS + 2% (v/v) FCS) with or without 5 μg/mL propidium iodide (PI). Cells were stored on ice until flow cytometry analysis on an LSR Fortessa (BD). The antibody-fluorochrome conjugates and flow cytometry filter sets used to identify specific haemopoietic populations are described in Additional file
3: Table S3.
ATRA-induced MPRO differentiation
MPRO cells (2.5 × 105 cells/mL) were treated with 10 μM ATRA. After 72 h treatment, the cells were stained with FITC-conjugated anti-Gr-1 (Ly6G/Ly6C) antibodies (BioLegend), and the cells enriched by FACS for different stages of differentiation based on Gr-1 staining. RNA was isolated from sorted cells for subsequent RT-qPCR analysis. Each FACS-enriched population was cytospun onto glass slides using the Shandon CytoSpin III Centrifuge (GMI), according to the manufacturer’s instructions. May-Grunwald Giemsa staining was performed by the NATA-accredited Haematology Laboratory at Royal Prince Alfred Hospital (Sydney, Australia).
Discussion
Historically, distinct ELF2 isoforms were described (NERF-1a, NERF-1b, NERF2) with some functional attributes [
22‐
24,
35,
36]. However, more recent studies on ELF2 function have been challenging to interpret as the exact ELF2 isoform used in overexpression studies, targeted in knockdown studies or detection methods were often not specified [
37‐
39]. ELF2 isoforms arise from distinct conserved loci, suggesting they may have evolved to play specific functional roles. The unique antibody reagents we developed enabled us to distinguish whether either ELF2 isoform is mutually or exclusively expressed. ELF2A is the major isoform expressed in the testis, whilst ELF2B is preferentially expressed in the thymus. The exact role ELF2 plays in different tissues may be impacted by N-terminal functional differences between isoforms and their proportionate expression. In a similar manner, three OCT1 isoforms that differ at their N-termini, can elicit variable transactivation of the same target genes and can control a different but overlapping set of target genes [
40].
A distinguishing feature of ELF subfamily members as opposed to other Ets proteins is the centrally positioned
Ets DNA-binding domain, and the intrinsically disordered N and C termini that flank it. Intrinsically disordered regions are more prone to forming protein interaction scaffolds or undergo post-translational modifications [
41]. Typically, intrinsically disordered proteins can also form central interaction hubs in signalling pathways [
42]. Deletion mutant studies have identified an acidic transactivation domain in ELF1 [
43], ELF4 [
44], and ELF2A [
24]. The lack of an intact transactivation domain in ELF2B supports the functional differences in ELF2B we observed. ELF2B’s unique and conserved N-terminus may interfere with normal ELF2 protein-protein interactions or may recruit unique binding partners that augment its inhibitory function. Deletion of this 19 aa N-terminal domain abrogates ELF2B’s inhibitory function, reversing the anti-proliferative and apoptosis-inducing effects of ELF2B.
Our comprehensive functional analysis of ELF subfamily members was necessary due to the similarities in their
Ets domain and structure. ELF1, ELF2, and ELF4 all have haemopoietic-specific expression, considerable redundancy in DNA binding [
8], and some common binding partners. As a result, ELF family members may compete for binding or exhibit a co-ordinated transactivation program in a stage- or temporally specific manner. ELF2 is able to competitively inhibit ELF1 transactivation of
Tie1 and
Tie2 target sites in chicken blood vessels [
45]. Other
Ets factors can compete with ELF1 for high affinity
Ets sites, but low affinity sites were still available for ELF1 binding in co-operation with other co-factors [
46]. With redundancy in DNA binding between
Ets factors, binding partners may significantly influence site preference for individual
Ets proteins in a context-specific manner.
Both ELF2B and ELF4 dramatically reduced cellular proliferation and clonogenicity in primary and transformed cells, whilst ELF1 and ELF2A had negligible effects. Antiproliferative functions have previously been suggested for ELF transcription factors, particularly for the candidate tumour suppressor protein ELF4 [
37,
47]. ELF2 (and ELF1) overexpression also resulted in reduced cellular proliferation in transformed cells (T3M-1 CI-10, HT1080, and MCF10A) [
37]. However, in contrast to these studies, ELF2 overexpression in hepatoma cells actually enhanced tumour cell proliferation, whilst conversely ELF2 knockdown repressed cell growth [
38]. Our data clearly reaffirms the antiproliferative effects of ELF2B and ELF4 and attributes a proapoptotic function to ELF2B.
The disparate effects on cell cycle kinetics observed between ELF2B and ELF4 may arise from differential DNA occupancy. ELF2 knockdown in ES cells resulted in up- or downregulation of fewer than 100 genes [
48], whilst ELF4 overexpression in T3M-1 CI-10 cells revealed 95 strongly regulated genes implicated in G1 cell cycle phase regulation and apoptosis [
37]. Previous studies have linked a role for ELF4 in cell cycle kinetics with transactivation activity largely limited to the G1 phase [
49]. ELF4 regulates the quiescent state of haemopoietic stem cells by facilitating their transition from G0 to G1 [
21]. ELF4 overexpression induced an accumulation in the G1 cell cycle stage, consistent with a previous report [
37]. Knockdown of ELF2 in SK-Hep1 cells led to an accumulation of cells in G1 using shRNAs that target both ELF2 isoforms [
38]. Consistent with these data, ELF2B overexpression resulted in a decrease in G1 phase, but this may have resulted from loss of cells due to apoptosis. As ELF2B can bind
Ets sites and therefore compete with ELF2A, it may be acting as a dominant negative protein by preventing canonical activation of growth-promoting ELF2A targets. As a consequence of inhibiting normal ELF2A activation, growth suppression and apoptosis may result. Furthermore, as a dominant negative protein, ELF2B may also recruit co-repressors or disrupt protein complexes that normally interact with ELF2A. ELF2B acts as a putative tumour suppressor protein in its ability to decrease proliferation, clonogenic capacity, and induce apoptosis in vitro, however, its ability to inhibit tumour formation in vivo needs to be definitively tested.
Elf2 overexpression drastically affected early T cell development in the thymus, but not peripheral T cells. We observed a significant accumulation of TCRβ
+ immature single positive (ISP) T cells and concomitant reduction in DP T cells in reconstituted mice after overexpression of ELF2 isoforms. After recombination at the TCRα locus, thymocytes assemble the mature TCR and co-express the co-receptor proteins CD4 and CD8 to form the pool of DP αβ-TCR expressing immature thymocytes, which constitute ~90% of the lymphoid compartment [
50]. The decrease in immature DP cells is reflected in an overall decrease in the proportions of surface TCRβ
lo-expressing cells. Further analysis of ISP DN cells attributes this to an accumulation of DN1 thymocytes. Given the significant decrease in DN4 and DP cells, we postulate that the increase of DN1 immature thymocytes results from an increased T cell lymphoid progenitor recruitment from the bone marrow to compensate for the reduced DP thymocyte output.
We postulate that ELF2 overexpression perturbs T cell development by interfering with pre-TCR assembly and activation. This could arise via interference with the expression of co-receptors, specialised adaptor molecules, and transcriptional regulators. Regulation of apoptosis is key in promoting survival of DN4 and DP thymocytes during pre-TCR signalling and subsequent positive and negative selection following engagement of self-antigen with major histocompatibility complex (MHC)-peptide ligands. Therefore, we also propose that Elf2 overexpression in T cells may impact on negative selection of DP thymocytes, leading to subsequent decreased DN4 and DP populations. For example, mice deficient in
Ets1, a regulator of pre-TCR signalling, have an impaired development of DN3 to DP cells, which is coupled with increased apoptosis but normal cell proliferation [
51]. Similar to Elf2 overexpression, this also resulted in reduced DN4 and DP populations.
During B cell development, common lymphoid progenitors in the bone marrow differentiate into pro-B cells and then transition to pre-B cells. The progression from pro-B cells to pre-B cells involves pre-B cell receptor (BCR) rearrangement via V(D)J recombination. Failure to assemble the BCR complex results in cell death at the first checkpoint in B cell development. In B cells, V(D)J recombination involves an orchestrated cleavage, rearrangement, and joining of DNA segments, which is tightly linked to the cell cycle, particularly in G0 and G1 phases [
52]. As Elf2b overexpression in particular reduces the number of cells in G1 phase, this may explain the reduction in precursor B cells progressing to pre-B cells, suggesting pre-BCR development is directly impacted. Alternatively, Elf2 may compete with other
Ets factors for binding at
Ets sites, which consequently affect the survival, cell cycle, or DNA rearrangement of pre-B cells. Elf2 also binds IgH enhancers π and μB, which are involved in V(D)J recombination [
23,
53] as does ELF1 and PU.1 [
54,
55]. ELF2 also regulates the expression of specialised adaptor molecules required for signal transduction after B cell activation, such as BLK and LYN [
24], as well as components of the BCR complex such as Igα and Igβ [
23].
In this study, we have demonstrated that ELF2B reduces cell proliferation, colony-forming ability, cell cycle progression, and survival. This has an impact in vivo with ELF2 isoforms disrupting the tightly regulated development of B and T cells. The significant reduction in DN4 and DP T cell populations in the thymus and in the pre-B cell population in the bone marrow of mice overexpressing Elf2 is consistent with disruption of key developmental checkpoints. Developing lymphocytes produce specific B and T cell receptors through V(D)J gene rearrangement and recombination events, a process crucial in generating receptor diversity. A functional receptor will confer cell survival and proliferation signals that enable these lymphocytes to progress in their development, whilst failure to form a functional receptor will trigger apoptosis in the developing lymphocyte. ELF2 may therefore play an important role in regulating key effectors involved in V(D)J gene rearrangement for TCR and BCR assembly in early lymphocytic development.