TLE4 regulation of wnt-mediated inflammation underlies its role as a tumor suppressor in myeloid leukemia

2.1. Cell culture, shRNA construction, and lentiviral infection

Kasumi-1 cells (ATCC, Manassas, VA) were cultured in RPMI-1640 (Lonza, Walkersville, MD) supplemented with 10% FBS (Sigma, St Louis, MO) and 1% Penicillin/Streptomycin (Invitrogen, Carlsbad, CA). HL60 cells (ATCC) were cultured in MEMα (Invitrogen) supplemented with 20% FBS and 1% Penicillin/Streptomycin. All cells were maintained at 37 °C and 5% CO₂ at concentrations of 10⁶ cells/mL. When indicated, cells were also cultured with the following: 10 uM all-trans retinoic acid (Sigma), 50 uM indomethacin (Sigma), 100 uM cytarabine (Abcam, Cambridge, MA), 0.1 uM vitamin D3 (Sigma), 10 nM recombinant human Wnt3a (R&D Systems, Minneapolis, MN), or 10 nM ICG-001 (Selleck Chemical, Houston, TX). Non-targeting scramble control and TLE4-specific shRNA constructs were developed and delivered to cells via lentiviral delivery as previously described [10]. The shRNA used and their target sequences were: shTLE4 1 (AGTGATGACAACTTGGTGG) and shTLE4 2 (GGCATTATGTCATGTATTA). Data in figures were obtained using shTLE4 2 unless otherwise indicated. Infected cells were identified by GFP fluorescence detected using FACS LSRII or selected for via cell sorting with FACS Aria (BD, San Jose, CA). Full-length TLE4 (a.k.a.KIAA1261 kind gift of Dr. Ohara [18]) cDNAs were cloned into the MSCV-IRES-GFP retroviral vector.

2.2. Cell cycle, annexin V, and flow cytometry analysis

Cell cycling and death in Kasumi-1 cell populations were determined using DAPI cell cycle and Annexin V assays as previously described [10]. When indicated, the following fluorescent anti-human antibodies were used: CD11b-APC (ICRF44; eBiosciences, San Diego, CA), CD14-APC-Cy7 (M5E2; Biolegend, San Diego, CA), APC-Annexin V (BD, San Jose, CA), DAPI (Invitrogen, Carlsbad, CA). All flow cytometry data was analyzed using FlowJo X (Treestar, Ashland, OR) and ModFit LT 3.0 (Verity Software House, Topsham, ME). Cells were analyzed using FACS Aria or LSRII (BD).

2.3. Detection of Wnt signaling

Wnt activity was also measured in 293T cells (ATCC) using a TOPFLASH/FOPFLASH assay. Briefly, 293T cells were transfected with the following reporter constructs using TransIT reagent (Mirus): TOPFLASH construct containing 8x promoter binding sites followed by firefly luciferase, FOPFLASH firefly luciferase control, and pRL renilla luciferase transfection control. 293T cells were additionally transfected with abovementioned shRNA constructs and/or expression vectors containing AML1-ETO, CBFb, and TLE4. Treated cells were subsequently analyzed using MicroLumat PLUS LB luminometer (Berthold Technologies, Bad Wildbad, Germany) 24 h post-transfection.

2.4. Western blotting

Kasumi-1 and 293T cells treated with TLE4-specific or scramble control shRNA were lysed for protein. Western blots were run and probed as previously described [17] using the following antibodies: anti-human TLE4 (sc13377x; Santa Cruz Biotechnology, Dallas, TX), anti-Actin (sc10731; Santa Cruz Biotechnology), donkey anti-goat IgG-HRP (SC2020; Santa Cruz Biotechnology), and goat anti-rabbit IgG-HRP (ab97051; Abcam). TLE4 band intensities quantified using ImageJ and standardized to Actin band intensity (NIH, Bethesda, MD).

2.5. Expression analysis via RNAseq and qRT-PCR

RNA was harvested from Kasumi-1 cells using TRIzol (Invitrogen) 7 days after lentiviral spinoculation with scramble control or TLE4-specific shRNA. Library construction was performed after polyA-tail selection and used for 50 bp paired-end reads that were aligned to human genome GRCh37.75 using STAR and quantified using HTSeq. DESeq was used for normalization and identification of differentially expressed genes. Gene Set Enrichment Analysis (GSEA; Broad Institute, Cambridge, MA) was performed via Java application by ranking genes using t-scores. Expression levels of select differentially expressed genes and others of interest were performed using qRT-PCR as previously described [17]. Primer sequences for SYBR Green and Taqman assays are listed in supplemental table.

2.6. Primary human AML patient samples

Diagnostic bone marrow specimens from AML patients with del(9q) or t(8;21) del(9q) and normal CD34+ samples were obtained from patients at Massachusetts General Hospital or from the Children’s Oncology Group as described previously with informed consent and assent from parents and patients in accordance with protocols approved by Massachusetts General Hospital Institutional Review Board and Children’s Oncology Group [10].

2.7. Statistics

Unless otherwise specified, analyses used student’s unpaired t-test or two-way ANOVA with Graphpad Prism (Graphpad Software, La Jolla, CA). For RNAseq analyses, significantly differentially expressed genes were identified by false discovery rate < 0.2.

3.1. TLE4 knockdown increases proliferation and inhibits induced differentiation of leukemia cells

To determine the effects of TLE4 knockdown (T4KD), Kasumi-1 cells were treated with scramble control or two different TLE4-specific shRNA via lentivirus. qRT-PCR and Western blotting confirmed 60–72% TLE4 message reduction and 75–82% TLE4 protein expression reduction, respectively, in T4KD cells (Fig. 1A). shRNA treated cells were tracked over a period of 18 days to monitor cell growth. Consistent with previous studies [10], T4KD cells had significantly faster growth than their control counterparts (Fig. 1B). DAPI cell cycle and Annexin V analysis with T4KD Kasumi-1 cells seven days post-infection showed increased S and G2/M-phase cells concomitant with decreased dead cell populations (Fig. 1C and D). qRT-PCR further shows significant reduction of AML1-ETO repressive targets CEBPa and GATA1, suggesting T4KD enhances AML1-ETO activity [19,20]. (Fig. 1E).

Flow cytometric analysis revealed T4KD was able to repress ATRA induction of CD14+ populations compared to control (Fig. 2A). The differences in CD14+ populations suggests TLE4 may regulate the differentiation potential of t(8;21) leukemia. To better understand the effect of TLE4 on differentiation of myeloid leukemia cells, T4KD using lentiviral-based shRNA was also achieved in HL60 cells, a human promyelocytic leukemia cell line harboring the PML-RARα oncoprotein. Treatment with vitamin D3 induces HL60 differentiation with increased surface marker CD14-positive cells [21,22]. T4KD was able to blunt vitamin D3 differentiation of HL60 cells, as indicated by reduced induction of CD14+ populations (Fig. 2B). T4KD was also able to reduce expression levels of PU.1 and MYB compared to controls in HL60 cells, quantified by qRT-PCR (Fig. 2C). Our results illustrate the ability of TLE4 to regulate the differentiation potential of leukemia cells independent of AML1-ETO function.

3.2. Low expression of TLE4 and TLE1 is not unique to t(8;21) AMLs

Del(9q) has a relatively unique association with t(8;21) and is not seen to any appreciable extent with any other AML translocation. As expression may be affected by other means such as methylation we sought to determine if low TLE1 or TLE4 expression was characteristic of other AML subtypes. We examined the curated data sets of over 2000 AML samples using the Blood-Spot web interface, (http://servers.binf.ku.dk/bloodspot), which includes over 2000 AML samples including those from the TCGA Research Network (http://cancergenome.nih.gov) as well as from the International Microarray Innovations in Leukemia Study Group [23]. As compared to AML samples with normal karyotype, which have a level of expression of TLE1 and TLE4 that is about average among AMLs, expression levels of TLE4 are significantly lower in t(8;21), t(11q23)/MLL, and del(9q), while expression of TLE1 was significantly lower in t(8:21), Trisomy 8, Trisomy 13, and HSC (see Supplemental Fig. S1 in the online version at DOI: http://dx.doi.org/10.1016/j.leukres.2016.07.002). The overall levels of TLE1 expression were lower in AML samples compared to TLE4.

3.3. TLE4 knockdown leads to upregulation of inflammatory and immune response pathways, a hallmark of t(8;21) leukemia

To better understand the mechanism behind the oncogenic effects of T4KD, we performed RNAseq analysis of Kasumi-1 cells seven days after lentiviral delivery of control or TLE4-specific shRNA. Results are summarized in a heatmap (Fig. 3A) and were verified by qRT-PCR query for select genes with log₂ fold change greater than 0.5 and false discovery rate (FDR) of less than 0.2 in an independent experiment (Fig. 3B). After filtering differentially expressed genes for FDR < 0.2, GSEA analysis using upregulated geneset showed enrichment of immune and inflammation-related pathways (Fig. 3C). Additionally, the same analysis identified enrichment of genes found to be upregulated in AML1-ETO expressing monocytes, further suggesting the contribution of TLE4 knockdown enhancing AML1-ETO function [24]. Interestingly, qRT-PCR using RNA harvested from primary human AML diagnostic bone marrow samples show increased levels of FOS, CEBPb, and PTGER4 expression in del(9q) samples, higher in those with t(8;21) del(9q), and even higher levels in non-del(9q) t(8;21) as compared to CD34+ controls (Fig. 4A). These findings emphasize the importance of inflammatory pathways in t(8;21) myeloid leukemia and the ability of del(9q), and apparently other cooperating mutations, to augment these pathways.

3.4. Increased inflammation signature is mediated through interplay between TLE4 and a COX-Wnt signaling axis

Previous work by Zhang et al. describes AML1-ETO-driven upregulation of COX that contributes to increased b-Catenin stability and activity, possibly through inhibition of GSK3β [25]. Given the known repressive role of TLE4 on Wnt signaling [12], T4KD is expected to upregulate and intensify Wnt signaling due to AML1-ETO. qRT-PCR using samples from shTLE4- treated Kasumi-1 cells seven days after knockdown revealed significantly increased COX1 and COX2 expression as well as those genes associated with prostaglandin metabolism and inflammation identified from above RNAseq, including FOS, PTGER4, and CEBPb (Fig. 4B). T4KD-mediated changes in myeloid-associated gene expression in HL60 cells were concomitant with similar increases in inflammatory genes FOS, CEBPb, PTGER4, and IL1b (Fig. 4C).

To confirm the induced inflammatory signature is a Wnt-responsive effect, naïve Kasumi-1 cells were serum starved for 24 h and cultured in media with rhWnt3a for 48 h qRT- PCR analysis of these cells identified various Wnt-responsive genes, including those previously identified as Wnt targets [26–28], such as LEF1 and CCND3, as well as those previously described in the T4KD-associated inflammatory gene signature, such as COX1, COX2, FOS, CEBPb, and PTGER4 (Fig. 5A). To further determine the role of Wnt signaling with T4KD, 293T cells were transduced with AML1-ETO to better characterize the relationship between TLE4 and the AML1-ETO-COX-Wnt signaling axis. Ectopic expression of AML1-ETO in 293T cells was able to increase COX1 and COX2 expression by at least seven-fold as determined by qRT-PCR (Fig. 5B). TOPFlash/FOPFlash Wnt reporter assay shows AML1-ETO was able to increase Wnt signaling activity in 293T cells by two-fold. This was abrogated with ectopic overexpression of full length TLE4, verifying 293T cells as an alternative system for modeling the AML1-ETO-COX-Wnt signaling axis (Fig. 5C). Subsequently, 293T cells were transduced with AML1-ETO and either control or TLE4-specific shRNA. T4KD was able to increase Wnt signaling in 293T cells with AML1-ETO expression (Fig. 6A). qPCR using RNA harvested from these cells confirm upregulation of COX1 and COX2 expression in AML1-ETO-expressing 293T cells treated with TLE4-specific shRNA compared to control (Fig. 6B).

Additionally, T4KD and control Kasumi-1 cells were cultured in media supplemented with 10 uM ICG-001, a small molecule inhibitor that binds with CBP (CREB binding protein) to block b-Catenin/TCF-mediated Wnt signaling [12,29–32]. Activating interactions between CBP, b-Catenin, and TCF require clearance of TLE/TCF binding [12], thus ICG-001 is predicted to inhibit T4KD-mediated Wnt signaling activation. Fold change of GFP+ lentiviral treated T4KD and control Kasumi-1 cells significantly diminished over 15 days in presence of ICG-001 (Fig. 7A). qRT-PCR analysis further demonstrated that ICG-001-mediated Wnt inhibition was able to abrogate expression of not only COX1 and COX2, but also the aforementioned inflammation and Wnt target genes (Fig. 7B). These results suggest that, while COX inhibition does indeed block Wnt signaling and consequent upregulation of inflammatory genes, the effects mediated by T4KD are largely dependent on release of Wnt regulation. Therefore, aberrant increases in Wnt signaling due to T4KD may explain its contributory leukemic effects to AML1-ETO function through a Wnt-induced inflammation.

3.5. COX inhibition is able to inhibit the proliferative and chemotherapy resistance effects conferred by TLE4 knockdown

Despite many current efforts to develop Wnt inhibitors, direct blockade of Wnt signaling remains challenging to clinical therapeutics due to its ubiquitous and integral function in normal development and stem cell maintenance [33]. However, studies have shown a number of clinically approved drugs antagonize Wnt signaling indirectly, including cyclooxygenase inhibitors [34]. We found that the addition of 50 uM indomethacin (INDM), a non-selective COX inhibitor, was able to significantly reduce Wnt signaling in AML1-ETO-expressing 293T cells treated with shTLE4 compared to DMSO control, suggesting COX inhibition is able to abrogate T4KD-induced increases in Wnt signaling (Fig. 6A). qPCR analysis showed that decreased Wnt signaling in these cells was concomitant with reductions in COX1 and COX2 expression as well (Fig. 6B). Kasumi-1 cells treated with control or TLE4-specific shRNA were cultured in 50 uM INDM for 16 days (Fig. 8A). COX inhibition significantly reduced growth of T4KD cells compared to DMSO control by four days. By day 13, growth of INDM-treated T4KD Kasumi-1 cells plateaued and fell below that of control cells.

Similarly treated Kasumi-1 cells were cultured in 100 uM cytarabine (AraC), a conventional chemotherapy agent used at a level comparable to serum concentrations in patients undergoing high-dose therapy [35]. Prior studies have shown that b-Catenin and Wnt signaling activity are associated with chemotherapy resistance in various cancer models, including colon, breast, and pancreatic cancer [36–38]. We found that T4KD was able to confer relative resistance to AraC treatment, indicated by significantly reduced dead cell populations (Fig. 8B). Interestingly, the addition of INDM was able to overcome T4KD-induced AraC resistance and resulted in even higher levels of cell death compared to control, supporting the potential use of INDM as adjuvant therapy for conventional chemotherapeutics.

Furthermore, qRT-PCR assays queried changes in expression levels of the inflammatory signature and various myeloid differentiation markers in Kasumi-1 cells due to T4KD and INDM treatment. qRT-PCR assays demonstrate upregulation of multiple Wnt targets in Kasumi-1 and 293T cells treated with TLE4-specific shRNA compared control, which can be significantly inhibited, and in some instances blocked, with the addition of INDM (Fig. 9). The expression levels of ELANE, MPO, and PU.1 were repressed in T4KD Kasumi-1 cells (Fig. 10) consistent with inhibition of differentiation. INDM was able to increase expression of these genes, suggesting that COX inhibition is able to relieve repression due to T4KD. The effects of combination ATRA and INDM therapy on differentiation could not be assessed due to severe lethality (data not shown).