The stem cell factor SALL4 is an essential transcriptional regulator in mixed lineage leukemia-rearranged leukemogenesis

The pMIG-MLL-AF9-GFP plasmid and the ψ-eco packaging vector were obtained from Dr. Scott Armstrong [27]. MEIS1 and HOXA9 retroviral vectors were purchased from Addgene (#21013 and #8515). The pCDH-CMV-3xFLAG-DOT1L vector was created in Baylor College of Medicine Genetic Core. All the plasmids were validated by DNA sequencing and or Western blotting (for 3xFLAG-DOT1L) or GFP fluorescence microscopy (for MLL-AF9-GFP). Detailed data are listed in Additional file 1.

The Sall4 ^flox/flox mice [17, 28] have been crossed with RosaCreER ^T2 mice (Jackson Laboratory) to generate Sall4 ^f/f /CreER ^T2 mice. For in vivo Cre-recombination, tamoxifen (Sigma-Aldrich) was administered via intraperitoneal injection every 2 days (100 μL of 10 mg/mL in corn oil) for totally five times. Primers used for genotyping are the following: wild type forward: cctcccggaattgcttatct, neo reverse: ctgtccatctgcacgagact, and flox-Sall4 Cre check: gcttctgcctctggtattgc [28]. All animal experiments were approved by the Institutional Animal Care and Use Committee at Baylor College of Medicine or Stony Brook University Medicine.

Lin-BM cells were isolated and transduced with recombinant MLL-AF9 using published protocols [17, 29]. Cells (0.5 × 10⁴) were plated in methylcellulose media (MethoCult™ M3234, STEMCELL Technologies), and replating was performed every 7–10 days. CFU (with > 50 cells) was scored during each round of plating. For subsequent cell culture, individual colonies were plucked using a P200 micropipettor and transferred to microcentrifuge tubes containing 500 μl of PBS. Pelleted cells were then resuspended in BM culture media [17] and maintained in a humid 37 °C 5%CO₂-incubator. For in vitro recombination, 4-OHT (Sigma-Aldrich) was resuspended in ethanol and added to cell culture at final concentration of 250 nM. Media was changed daily during culture. For in vivo transplantation, MLL-AF9-transduced cells (5 × 10⁵) were injected through a tail vein into lethally irradiated (900 cGY) mice. Recipient mice were maintained on antibiotics for 2 weeks.

Cells (1 × 10⁶/mL) were washed with ice-cold PBS-EDTA, centrifuged at 500g for 5 min, and fixed with 2 mL 70% ice-cold ethanol at 4 °C for overnight. After fixation, cells were washed again, resuspended in 500 μL PBS solution containing 1× RNAse and propidium iodide(Abcam), and then incubated at 37 °C for 30 min. Fluorescence was measured with a LSR II flow cytometer. The assay was carried out in triplicate, and 10,000 events were analyzed per experiment using the BDFACSDiva software. Data was analyzed with ModFit LT 5.0.

Propidium iodide (Molecular Probes) was used to exclude dead cells. Fluorescence-activated cell sorting (FACS) scheme used to isolate primary mouse BM cells are listed as follows [30]. The HSC and multipotent progenitor cells (MPP) were defined as HSC (Linneg Sca-1+ c-Kit+, LSK, CD34-CD135-CD150+ CD48−), MPP1 (LSK CD34+ CD135− CD150+ CD48−), MPP2 (LSK CD34+ CD135− CD150+ CD48+), MPP3 (LSK CD34+ CD135− CD150− CD48+), and MPP4 (LSK CD34+ CD135+ CD150− CD48+). For early hematopoietic precursors [31], BM cells were co-stained with antibodies against lineage markers, Sca1, c-kit, CD34, FcgRII/III, IL7Rα, and flk2.

ChIP assays were conducted using a One-Step ChIP kit (Epigentek) with ChIP-grade antibodies against FLAG (Sigma), H3K4me3, H3K79me2/3 (Abcam), and LSD1 (Cell Signaling). Chromatin enrichment was evaluated by qPCR relative to IgG-pulled input DNA. The sequences of the PCR primers were described previously [25, 26].

Total RNA was extracted using TRIZOL reagent and purification with QIAGEN RNeasy Mini Kit. Microarray analysis was performed using Agilent SurePrint mouse G3 Exon 4x180K. The arrays were hybridized and scanned, and data was extracted using Agilent Feature Extraction Software Version 11.0.1.1. The Bioconductor “limma” package (http://​bioconductor.​org) was used to analyze the microarray data. The background-corrected data were log2 transformed and quantile normalized. Moderated t statistics were used to test if genes were differentially expressed between Sall4 KO and control groups. Benjamini-Hochberg method was used to estimate false discovery rate (FDR). FDR < 0.05 were considered statistically significant.

This was performed by Active Motif (Carlsbad, CA). Briefly, 10 × 10⁶ cells were fixed with 1% formaldehyde and quenched with 0.125 M glycine. Chromatin was prepared and sonicated, and the DNA sheared to an average length of 300–500 bp. Genomic DNA (input) was prepared by treating aliquots of chromatin with RNase, proteinase K, and heat for de-crosslinking, followed by ethanol precipitation. An aliquot of chromatin (30 μg) was precleared with protein A agarose beads (Life Technologies). Genomic DNA regions of interest were isolated using 12 μg of anti-SALL4 antibody (Abcam #ab29112). The obtained complexes were eluted from the beads with SDS buffer and subjected to RNase and proteinase K treatment. Crosslinks were reversed by incubation overnight at 65 °C, and ChIP DNA was purified by phenol–chloroform extraction and ethanol precipitation. Illumina sequencing libraries were prepared from the ChIP and input DNAs by the standard consecutive enzymatic steps of end-polishing, dA-addition, and adaptor ligation. After a final PCR amplification step, the resulting DNA libraries were quantified and sequenced on NextSeq 500 (75 nt reads, single end). Reads were aligned to the mouse genome (mm10) using the BWA algorithm. Duplicate reads were removed and only uniquely mapped reads (mapping quality ≥ 25) were used for further analysis. Alignments were extended in silico at their 3′-ends to a length of 200 bp, which is the average genomic fragment length in the size-selected library, and assigned to 32-nt bins along the genome. Peak locations were determined using the MACS algorithm (v1.4.2) with a cutoff of p value = 1e−7. Peaks that were on the ENCODE blacklist of known false ChIP-Seq peaks were removed. Signal maps and peak locations were used as input data to the Active Motifs proprietary analysis program. Quantitative PCR (qPCR) reactions were carried out in triplicate using SYBR Green Supermix (Bio-Rad) on a CFX Connect™ Real Time PCR system. The resulting signals were normalized for primer efficiency by carrying out qPCR for each primer pair using input DNA. Primer sequences are available upon request.

To characterize the effect of SALL4 loss on MLL-AF9-induced transformation, we first generated an inducible knockout system by crossing Sall4-floxed mice with mice harboring a Cre-estrogen-receptor-T2 (Cre-ER^T2) allele at the ubiquitous ROSA26 locus. Previous studies demonstrated that Cre-recombination of the Sall4 allele removes exons two and three, which contain all zinc-finger domains found in Sall4 [28]. Thus, upon tamoxifen administration, the Cre-ER^T2 system allows for translocation of Cre recombinase into the nucleus resulting in efficient deletion of Sall4 exons. Initially, we isolated lineage negative (Lin-) BM cells from 5-fluorouracil-primed Sall4 ^f/f CreER^T2 mice and transduced them with a retroviral pMSCV-IRES-MLL-AF9 vector by spinoculation on three consecutive days. The cells were then subjected to four rounds of plating in methylcellulose media, with or without 4-hydroxytamoxifen (4-OHT) administration (Fig. 1a). As illustrated in Fig. 1b, c, during the fourth round of plating, 4-OHT administration induced Sall4 exon sequence excision, which nearly abolished the clonogenic capacity of MLL-AF9-transformed cells. This was in sharp contrast to the compact, large number of “blast-like” colonies observed in ethanol (control)-treated cells. Next, individual colonies from the control group dishes have been plucked, dissociated into single cells, and grown in BM culture media. In these cultures, 4-OHT administration likewise substantially inhibited proliferation of MLL-AF9-transformed cells (Fig. 1a, c). Thus, these studies suggest that proper expression of SALL4 is required in MLL-AF9-mediated leukemic transformation. To further validate these findings with a separate model, we used our previously validated doxycycline-inducible lenti-TRIPZ-7410 Sall4 shRNA vector [13] and transduced it into MLL-AF9-transformed BM cells. After 5 days of puromycin (1 μg/mL) selection, the recovered cells were subjected to proliferation assays and treated with doxycycline (1 μg/ml) every 2 days for 7 days. As illustrated in Fig. 1d, doxycycline-mediated Sall4 knockdown also markedly reduced viable cell numbers in culture.

Based on our intriguing in vitro findings, we then sought to determine whether Sall4 deletion also affects MLL-AF9-mediated transformation by BM transplantation—which would establish SALL4 requirement for MLL-AF9-mediated leukemogenesis in vivo. We isolated lineage Lin- BM cells from Sall4 ^f/f CreER^T2 mice and transduced them with a retroviral MLL-AF9 vector and then injected the cells intravenously into lethally irradiated syngeneic mice. Four days later, the mice were injected intraperitoneally with either corn oil (control) or tamoxifen for 10 days to activate Cre-mediated Sall4 excision. The mice were monitored for disease development by circulating blood cell (CBC) analysis and symptoms such as abnormal gait and labored breathing. Interestingly, while tamoxifen-administered mice failed to develop leukemia over an 8-month observation period, oil (control)-administered mice all died of AML within 102 days, as evidenced by highly elevated white blood cell (WBC) counts (not shown), marked splenomegaly, and extensive blast infiltration of BMs (Fig. 2a–c). Thus, SALL4 is clearly required for MLL-AF9-induced leukemia initiation in vivo. Next, we further asked if Sall4 deletion at a later time point similarly affects MLL-AF9 leukemogenesis. We then generated the same MLL-AF9 leukemogenesis model, and the transplanted recipient mice were allowed to establish leukemia for up to 4 weeks. At this stage, injection of tamoxifen was found to significantly prolong the survival of the mice to a median of 80 days, compared with the median survival of 52 days of the control mice (Fig. 2d), although all mice in both groups eventually died of leukemia. These data indicate that inactivation of SALL4 is sufficient to delay MLL-AF9-induced AML progression.

The apparent SALL4 requirement for MLL-AF9-mediated transformation both in vitro and in vivo prompted us to test how loss of SALL4 affects MLL-AF9 cellular properties. As described in Fig. 1, 4-OHT or doxycycline treatment in MLL-AF9-transformed cells largely reduced colony-forming units and/or viable cell numbers. We next designed flow cytometric assays for cells that underwent SALL4 knockdown. We have focused on cell cycle and apoptosis markers as these have been related to SALL4 functions in AML cells [12]. As shown in Fig. 3a, propidium iodide (PI) staining and fluorescence-activated cell sorting (FACS) analyses indicate that doxycycline-mediated SALL4KD caused cell cycle arrest in cells at G1 phase (65% ± 4.8 for TRIPZ7410 vs 48% ± 2.1 for TRIPZ non-silencing vector-treated control groups, p = 0.005), but significantly reduced G2-phase cell fractions (0.7% ± 0.7 for TRIPZ7410 vs 7.8% ± 0.4 for control groups) and the S-phase cell populations (34% ± 3.8 for TRIPZ7410 vs 44% ± 1.7 for control groups, p = 0.02). In addition, PI and annexin V staining revealed that approximately 10% of the SALL4 reduction cells were apoptotic (Fig. 3b), whereas only ~ 2.5% of the control cells were apoptotic (more than a fourfold change). Consistent with the FACS data, real-time quantitative qRT-PCR analysis following SALL4 reduction showed a substantial upregulation of the transcript levels of Cdkn1a/p21, Trp53inp1, Glipr1, and Zmat3/ Wig1 (Fig. 3c). Cdkn1a has been known to be a potent cyclin-dependent kinase inhibitor that blocks cell cycle progression in the G1 phase [32], while Trp53inp1 and Glipr1 are p53-induced pro-apoptotic genes that also mediate apoptosis and G1 cell cycle arrest [33, 34]. Similarly, Wig1 regulates cell cycle arrest and cell death through the regulation of p53 targets FAS and 14-3-3σ mRNA levels [35]. In contrast, transcript levels of some genes that are related with leukemia prognosis/aggressiveness, such as Ambp and Thbs1, were drastically repressed. Thus, these data suggest that deletion of Sall4 disrupts MLL-AF9 cellular proliferation mainly by inducing apoptotic events and blocking cell cycle progression in the G1 phase.

In MLL-r leukemias, HOXA9 and its dimerization partner MEIS1 are the most well-characterized direct targets, which in turn can replace MLL-fusion proteins in overexpression experiments [21]. Consistent with previous findings from LSK HSPC [7] or leukemic cell [19] studies, 4-OHT-induced Sall4 deletion in MLL-AF9-transformed BM cells decreased the transcription levels of these two factors when normalized to tested RNA levels (Fig. 4a). Additionally, multiple MLL-AF9-related genes were also downregulated, which contrasts to upregulated apoptotic genes Bat3 and Dapk3 (Fig. 4a). These results thus demonstrate that proper expression of MLL-AF9 target genes requires constant activation of SALL4. Next, we asked how Sall4 deletion may affect relevant histone methylation status at the promoter regions of such genes. We focused on histone H3 lysine 4 (H3K4) and 79 (H3K79) since they have been linked with SALL4 regulatory functions and well studied in MLL-r leukemias. Indeed, SALL4 binding to Meis1, Hoxa9, and Hox10 promoters has been determined by ChIP-Seq assays as described below. As shown in Fig. 4b, ChIP-qPCR assays clearly confirmed the enrichment of H3K4me3 and H3K79me2/3 at the proximal regions of Meis1 and Hox9 promoters, and 4-OHT administration in these cells fundamentally decreased their levels at all the examined sites. In contrast, while the repressive mark H3K9me3 was also found highly enriched in these promoter regions, 4-OHT administration failed to decrease their levels (Fig. 4b). Thus, in MLL-AF9-induced leukemia, SALL4 modulates MLL-AF9 downstream gene expression via defined epigenetic modification process.

The altered H3K4me3 enrichment at MLL-AF9 target genes should be associated with the methyltransferase activities of at least wild type MLL and LSD1, as both factors have previously been shown to interact with SALL4, while MFPs have lost H3K4 methyltransferase activity [19, 24‐26]. We then considered the potential of a direct SALL4-DOT1L interaction. In previous studies, SALL4 overexpression caused enhanced H3K4me3 and H3K79me2/3 at Bmi1 promoter in 32D myeloid progenitor cell and at HOXA9 promoter in AML cells respectively [18, 19]. Additionally, Hox7 and Hox10 promoters also showed significantly decreased H3K79me2/3 following 4-OHT administration (Fig. 4b). In transfected 293 T cells, we detected that HA-tagged SALL4A and SALL4B isoforms were immunoprecipitated by DOT1l protein, while the C-terminal fragment was not (Fig. 5a, b). Consistently, in an accompanying enzymatic activity assay, the SALL4 isoforms, but not the C-terminus truncated mutant, purified increased tri-methylated H3K79 from the chromatins (Fig. 5c). We next sought to confirm their protein interactions in MLL-AF9-transformed BM cells. As shown in Fig. 5d, by a co-IP assay using an IP-grade anti-SALL4 antibody, SALL4-DOT1L interaction was clearly demonstrated by endogenous protein pulldown. Moreover, when the cells were subjected to 4-OHT-mediated Sall4 deletion, we observed reduced DOT1L levels that were immune-precipitated by SALL4. Notably, LSD1, the previously identified SALL4-bound epigenetic factor [17] was also found immunoprecipitated by SALL4 and showed reduced pulldown after 4-OHT administration (Fig. 5d).

We then further evaluated these co-IP findings with a subsequent ChIP-qPCR assay. Because there is no available ChIP-grade anti-DOT1L antibody, we transduced the MLL-AF9 BM cells with our newly generated DOT1L-FLAG-expressing lentiviral vector (by BCM Vector Core), and its expression has been confirmed by Western blotting using antibodies against the Flag tag and DOT1L protein (not shown). As expected, both DOT1L-FLAG and LSD1 were detected bound to the Meis1 and Hoxa9 promoters, and 4-OHT treatment in these cells largely reduced their enrichment at the same examined promoter regions (Fig. 5e). Thus, this group data supports that SALL4 dynamically recruits DOT1L and LSD1 and regulates important downstream gene expressions in MLL-AF9 leukemic cells.

Given the clear regulatory effects of SALL4 on Hoxa9 and Meis1 mRNA expressions, we initially performed rescue experiments using CFU and proliferation assays. However, following 4-OHT-mediated Sall4 deletion, combined overexpression of HOXA9 and MEIS1 (by retroviral transduction) failed to restore the MLL-AF9 transforming activity, although they could induce leukemogenesis in tested mice (not shown). Therefore, SALL4 may exert a relatively “proximal” effect in this model, or these factors control important but non-overlapping regulatory pathways. To further elucidate SALL4-regulated molecular networks, we conducted mRNA microarray assays with the MLL-AF9-transformed BM cells. We started with a relatively earlier event (where 4-OHT was treated in cells for 60 h) with the goal of detecting an “acute” effect following Sall4 deletion. Around 77% reduction of Sall4 mRNA levels has been verified by qRT-PCR, and the 106 up- and 57 downregulated genes were analyzed with FDR (false discovery rate) < 0.05. The complete results from the microarray analysis will be published elsewhere. Interestingly, the upregulated genes include tumor suppression/proapoptotic factors Glipr1, Trpm2, Trp53inp1, and Wig1 (Zmat3); cell cycle inhibitors Cdkn1a (p21), Trp53inp1, and Wig1; the retinoic acid (RA) biosynthesis inhibitor Dhrs3; HSPC colony-forming repressor Slfn2; and hematopoietic differentiation markers Col5a1, Fyb, Irf8, and Pira6. In contrast, the markedly downregulated genes include TGFβ factors Tgfβ2, Tgfβ3, and Tgfβr3, which are all critical regulators in maintaining proper proliferation of HSPCs [36, 37]. Some other downregulated genes are related to chemo-resistant or leukemia aggressiveness such as Thbs1, Tgm2, Ambp, and the AF9 regulator Sgk1, which negatively regulates the DOT1A-AF9 repressor complex [38, 39]. Further, PANTHER (Protein ANalysis THrough Evolutionary Relationships) classification revealed that inflammation mediated by chemokine and cytokine signaling pathway (Ccl3, Ccl4) and gonadal hormone receptors (Jun, Igf1, Dusp1)—the newly identified leukemic cell stimulators [40]—are all affected by SALL4 levels (see Additional file 1C). These data support that SALL4 likely regulates MLL-AF9 leukemogenesis via multiple signaling pathways involving cell cycle arrest, differentiation, apoptosis induction, chemo-resistance, and importantly, HSPC, or leukemic stem cell (LSC) essential networks such as TGFβ. Further in-depth functional assays are needed to prove this notion. Το validate the microarray results, we performed RT-PCR analysis on 14 of the differentially regulated genes, which confirmed the findings. Some of the data are shown in Fig. 3c.

In an attempt to further classify the direct regulatory targets of SALL4, we next performed ChIP-Seq assays. High-stringency data analysis identified 350 SALL4-enriched peaks. The 350 peaks were associated with 451 protein-coding genes, including key MLL-AF9 target genes, previously identified SALL4-bound genes, and important leukemia-related genes, such as Cebpα, Id2, Elf1, Evl, Flt3, Meis1, Nf1, Tal1, Tcf7l1, Gata6, Sox12, Bahcc1, Nkx2-3 [12, 17, 41‐44], Hox factors Hoxa9, Hoxa10, Hoxa11, Hoxa13, Notch ligand Jag2, and Wnt/β-catenin signaling regulator Wnt7b (Fig. 6a and see Additional file 2 for a full list). ChIP-qPCR has been conducted on four selected binding sites, plus one positive and an unbound negative control, which confirmed the findings (Fig. 6a). Analysis of the 451 genes using the DAVID bioinformatics resources [45] identified 12 significantly enriched pathways (p ≤ 0.05), including cancer/AML-specific pathways, signaling pathway-regulating pluripotency of stem cells, and Hedgehog (Hh) signaling pathways (Fig. 6b). These data again support an important role of Sall4 in MLL-r leukemogenesis. In comparison with the mRNA expression data, we found that not many of the ChIP-Seq-identified SALL4 targets were associated with early expression changes as those detected in mRNA microarray assays. Considering SALL4’s dynamic and complex network in gene regulation, the limited overlap between binding and differential expression could be in part related to the length of time that 4-OHT was treated, the presence of other co-regulators in play, and/or the relatively lower number of genes that were identified in present assays. Further in-depth studies would help to fully clarify these issues.

Given our long-term goal of developing SALL4-based therapeutic strategy, we have asked how SALL4 depletion may affect normal hematopoiesis in mice. As described in Fig. 2a, mice receiving tamoxifen failed to develop leukemia, and all mice were viable without apparent phenotype defects. To gain a better understanding of SALL4 requirement in normal hematopoiesis, we then injected 8-week-old Sall4 ^f/f /CreER^T2, Sall4 ^f/f /Cre-, and RosaCreER ^T2 control mice with tamoxifen and verified Sall4 excision as we did with the MLL-AF9 AML model. Consistently, in these studies, tamoxifen-mediated Sall4 deletion did not cause significant differences in peripheral myeloid, erythroid, and platelet counts, nor BM cellularity as compared with their respective littermate controls (Sall4 ^f/f /Cre-). The RosaCreER ^T2 control mice showed altered WBC and monocyte numbers, but they are still within the normal range (Table 1 and not shown). To examine the effect of Sall4 deletion in each of the BM hematopoietic stem and progenitor compartments, we then performed multi-parameter flow cytometry analyses. Surprisingly, as shown in Fig. 7a–b, no significant differences in the number of HSC, multipotent progenitor MPP1, MPP4, CMP, or MEP populations were observed between Sall4 knockout and their respective controls, and the multipotent progenitor MPP2 and MPP4 populations were even increased in TAM-treated Sall4 ^f/f /CreER^T2 mice.

To further validate above in vivo findings with a separate model, we then attempted to cross Sall4 ^f/f mice with vav-Cre mice (Jackson Laboratory) to delete Sall4 throughout the hematopoietic compartment during embryonal development, even though Sall4 homozygous knockout has previously been shown embryonic lethal [3, 28]. Interestingly, breeding of the Sall4 ^f/f /vav-Cre mice were successful from all of the five groups of breeder pairs that were examined, and Sall4 ^f/f /vav-Cre mice were born at normal Mendelian frequency with essentially normal body and organ weights. PCR typing also confirmed excision of Sall4 exon sequences from the analyzed hematopoietic samples (Fig. 7c). Again, in these hematopoietic-specific Sall4 deletion mice, no significant differences in basal BM hematological parameters and each HSPC fractions were detected. Thus, Sall4 deletion causes no adverse effect on normal hematopoiesis in mice in present models (Table 2 and Fig. 7c).