Influence of wild-type MLL on glucocorticoid sensitivity and response to DNA-damage in pediatric acute lymphoblastic leukemia

The cell line panel has been previously described and comprised nine T-ALL lines derived in our own laboratory from pediatric ALL bone marrow specimens (PER cell lines), plus six additional T-ALL cell lines obtained from external sources [12, 13]. Cell lines were grown in RPMI-1640 supplemented with 2 mM L-glutamine, 10 nM 2-mercaptoethanol and 10-20% heat-inactivated fetal calf serum. The media for PER-cell lines contained additional non-essential amino acids and pyruvate, whilst 300 units/ml interleukin-2 is required for growth of PER-427 and PER-487. The sensitivity of the T-ALL cell lines to methylprednisolone (MPRED) and dexamethasone (DEX) has been previously published [12] and was measured using the MTT assay with drugs incubated over four days. The IC50 (drug concentration that inhibits cell growth by 50%) was used as the measure of drug resistance.

Briefly, RNA was extracted from cell lines in exponential growth phase and hybridized to Affymetrix HG-U133A microarrays [11, 14]. Microarray data were normalized using robust multi-array analysis (RMA) and all passed quality control criteria for noise, background, absent/present calls, and 3'/5' signal ratios for ACTB and GAPDH. To interrogate the biological pathways represented by MLL expression profiles we used Gene Set Enrichment Analysis (GSEA) [15]. The median value of the five MLL probe sets present on the HG-U133A was calculated for each cell line, and correlated across the panel against all other probe sets on the array using Pearson's correlation as metric (GSEA v2.0, May 2006, 10,000 permutations). GSEA examines ranked lists of genes for enrichment of biological pathways contained within four different databases: C1 (genomic loci), C2 (curated biological pathways), C3 (genes with common regulatory motifs), and C4 (computational gene networks). Since not all genes within a given biological pathway are expected to be regulated in the same direction, rankings were performed using absolute correlation values as previously described [11]. Published microarray data used for in silico analysis [14, 16‐18] was downloaded from publicly available depositories or authors' websites.

Real-time quantitative RT-PCR (qRT-PCR) was performed on total RNA from cell lines in accordance with standard Applied Biosystems protocols (Foster City, CA) and in accordance with our published methods [19]. All experiments were run in duplicates on an ABI 7700 sequence detector and data normalized to expression of beta-actin (ACTB). Primers and probe for MLL and GILZ qRT-PCR were purchased from Applied Biosystems (ABI Assays on Demand); the MLL assay targeted exons 30-31 (Refseq NM_005933).

Three pSM2 retroviral RNAi vectors for MLL (V2HS_196843, V2HS_198375, V2HS_214961) and a non-silencing (NS) control vector were obtained from Open Biosystems (Huntsville, USA). For optimal mammalian expression, shRNA inserts were subcloned with EcoRI and XhoI into MSCV-LMP (MSCV/LTRmiR30-PIGΔRI, a generous gift from Prof. Scott Lowe, Cold Spring Harbour Laboratory [20]), which contains GFP and puromycin selection cassettes and drives miR30-shRNA expression using the retroviral 5'LTR. V2HS_198375 (MLL198) was found to suppress MLL expression most efficiently in transient transfection experiments and was used for subsequent experiments. The retroviral packaging cell line PA317 (selected in HAT medium) was transfected with linearised miR30-shRNA plasmid DNA (for both NS control and MLL198) using Lipofectamine, and GFP-positive cells were selected with puromycin. Stably transfected retroviral-producing PA317 cell lines were γ-irradiated (30 Gy) and incubated at 37°C overnight before co-culture with PER-117 cells for 48 hours. Retrovirally infected PER-117 cells were subsequently removed and selected with puromycin to generate cell lines stably expressing shRNA for MLL (MLL-KD) or the NS control (MLL-Scr). Efficiency of RNAi knockdown for MLL was assessed both by qRT-PCR as described above, and by immunoblot of nuclear protein extracted from cell lines in log-phase growth using standard methods. Antibodies used were mouse anti-MLL^C/HRX, clone 9-12 (Upstate Cell Signaling Solutions, Temecula, CA), which detects the C-terminal proteolytic fragment of MLL (~180 kDa), and mouse anti-human β-actin as loading control (Pan Actin Ab-5 (ACTN05) NeoMarkers, Fremont CA). Densitometric quantitation of protein bands from multiple extractions taken at independent time points and from different cell-line stocks was performed using ImageJ software http://​rsbweb.​nih.​gov/​ij/​, with MLL expression normalized to β-actin loading.

Cell growth and viability were measured using the Vi-CELL XR Viable Cell Analyzer (Beckman Coulter). Cells in exponential growth phase were seeded at 5 × 10⁵ ml^-1 in a 96-well plate in the presence or absence of dexamethasone (10 μg/ml - 258 μg/ml, Mayne Pharma Pty Ltd, VIC, Australia), 0.025 μg/ml cytarabine (ARAC; Pharmacia Pty Ltd, NSW, Australia), 0.01 μg/ml methotrexate (MTX; David Bull Laboratories), or 1 Gy gamma-irradiation, and incubated for two days at 37°C before measuring cell survival. Each drug concentration or condition was tested in triplicate and data were normalised to values obtained from untreated cells. For metabolic assays, cells in exponential growth were seeded at 5 × 10⁵ ml^-1 in fresh media and incubated for two days at 37°C before harvesting supernatants. Glucose and lactate supernatant concentrations were measured using the Amplex Red kit (Invitrogen, Australia), substituting lactate oxidase (Sigma, Australia) as required. For assessment of GILZ induction, MLL-KD and MLL-Scr cells in exponential growth were incubated with 1 μM dexamethasone (Mayne Pharma Pty Ltd, VIC, Australia) for four hours prior to RNA extraction and measurement by qRT-PCR.

Our laboratory has developed an authenticated panel of pediatric T-ALL cell lines that have been grown in the absence of drug selection. These cultures retain critical features of the primary disease and their drug resistance profile parallels the spectrum of resistance that has been observed in primary patient specimens [12]. We recently examined the baseline resistance of these 15 T-ALL cell lines to the GCs DEX and MPRED [12] and correlated the data with gene expression profiles as determined by HG-U133A microarray [11]. Although these lines have been cultured without prior exposure to in vitro drug selection pressure they demonstrate a natural spectrum of GC resistance, with IC50 values across the panel varying by 4-5 orders of magnitude (Figure 1A). This resistance profile is not explained by mutations in the glucocorticoid receptor (GR) or variations in its level of expression [21], indicating that defects downstream of the GR are primarily responsible for GC resistant phenotypes in these cell lines.

Our analysis of the microarray data revealed that GC resistance was significantly correlated with reduced expression of MLL [11]. To confirm this correlation we used qRT-PCR to measure MLL mRNA expression across the panel, using a probe targeting the 3' end of the MLL coding region. Expression levels measured by qRT-PCR were highly correlated with resistance to both GCs (Figure 1A; correlation vs. dexamethasone IC50 -0.849 (p < 0.0001), methylprednisolone IC50 -0.851 (p < 0.0001)). Whilst translocations of the MLL gene are prevalent in infant ALL they are infrequent in T-ALL [8, 9, 22], suggesting that the observed correlation reflected expression of the wild-type gene. Indeed, T-ALL cell line karyotypes indicated no abnormalities at the 11q23 MLL-locus [12], a conclusion confirmed by Southern Blot for all 15 cell lines (data not shown). On the HG-U133A microarray there are five independent probes for MLL, and these span the entire length of the gene, encompassing both sides of the major break region (MBR) that is involved in almost all translocation events (Figure 1B). Across the 15 T-ALL cell lines correlation of MLL mRNA expression and GC resistance was significant for all five probe sets (median probe significance DEX p = 0.0025, MPRED p < 0.0001) indicating no discrepancy in expression between the 5' and 3' regions of the gene. Based on these data we conclude that the observed correlation with GC sensitivity in T-ALL cell lines is related to expression levels of wild-type MLL rather than MLL-translocation products.

To gain further insight into the transcriptional programs associated with MLL, the expression profile of this gene across the T-ALL cell line panel was correlated to the expression of all other genes on the microarray. This output was analyzed with GSEA to identify the biological networks associated with variations in MLL expression. The strongest signatures were returned from the C2 (curated pathway) and C4 (computational gene network) databases, with 17 and 83 enriched gene sets respectively falling within the significant GSEA false discovery rate (FDR). Very few significantly enriched gene sets were identified from genomic loci and regulatory-motif databases (C1 and C3). The top ranked significant gene sets from the C2 and C4 databases are listed in Tables 1 & 2. The majority of these pathways are involved with the control of cell growth and metabolism (e.g. MYC-regulated pathways, RNA transcription, oxidative phosphorylation, the TCA cycle, proteasomal regulation, nucleotide synthesis, translation initiation and antioxidant defense). The overwhelming direction of expression of these pathways was a negative correlation with expression of MLL. Thus lower expression of MLL in these cell lines is associated with signatures consistent with a proliferative phenotype. In addition the expression levels measured by each of the five MLL probe sets were found to correlate significantly with cell line doubling times [12] (median correlation 0.67, p < 0.01). These findings are in keeping with our previous observation that reduced expression of MLL is part of a proliferative metabolism signature that is associated with GC resistance in T-ALL cell lines [11]. Importantly, several gene sets were involved with the regulation of apoptosis (MORF_AATF, MORF_MAPK2), p53 response (MORF_EI24, GNF2_NS) and DNA damage repair (MORF_UNG), with the direction of association linking reduced MLL-expression with the activation of anti-apoptotic transcriptional networks (Table 2).

Recent evidence suggests that the genes most commonly translocated with MLL are not selected at random but may in fact be functionally related as part of an 'MLL-web' [5, 23, 24]. If this is true, then in the context of the observed relationship between MLL expression and GC resistance in the present study (Figure 1A) one might predict that the expression of these genes would similarly be correlated with GC resistance in our T-ALL cell lines. Of the >50 known translocation partner genes of MLL, 43 are represented on the HG-U133A microarray (corresponding to a total of 93 probe sets). Despite the absence of MLL-translocations in the T-ALL cell lines we observed that a large number of these (18 genes, 26 probe sets) were significantly correlated to MPRED and DEX resistance (Table 3). This association is much greater than would be predicted by chance alone (exact binomial test, p < 0.001). It is relevant that the majority of the genes listed in Table 3 are involved in transcriptional regulation (GMPS, DCPS, ELL, LPP, AF10, CREBBP, EP300, AF4), proliferation (GAS7) or metabolism (CBL, GPHN and ACACA, the latter being the rate limiting enzyme for conversion of acetyl-coA into cholesterol). The correlation of these genes with GC resistance may therefore be reflective of the metabolic and proliferative changes driving this phenotype in T-ALL cell lines of which MLL appears to be a part [11].

Since our data indicated an association between GC sensitivity and expression levels of MLL in T-ALL in vitro we looked for further evidence in the literature for such an association. Holleman et al previously examined the ex vivo sensitivity of diagnostic pediatric ALL patient specimens to individual induction therapy agents and correlated the findings with gene expression data measured in the same samples using HG-U133A Affymetrix microarrays [17]. We examined this data for the expression level of MLL in T-ALL patient specimens from this cohort that were determined to be sensitive or resistant to prednisolone. Importantly, three of the five MLL probe sets on the array showed a significantly lower expression of MLL in resistant samples confirming the association we observed in T-ALL cell lines. Figure 2A shows the data for the probe set with the strongest association (212079_s_at, p < 0.001 unpaired t-test), and for the summary of the five probe sets calculated using median expression values (p < 0.05, unpaired t-test). For further evidence of a link between MLL expression and GC resistance we examined a dataset we have previously published comparing gene expression patterns in pediatric ALL specimens taken at the time of diagnosis and relapse [14]. Although we were not able to directly measure the GC sensitivity of these specimens it is known that almost all patients initially respond to induction therapy and achieve first remission, whilst GC resistance is a well-documented feature of relapse [25, 26]. It is therefore reasonable to expect that many of the relapse specimens in this cohort would have elevated GC resistance compared to their diagnostic counterparts. Examining the same MLL probe sets as above, we observed a decrease in MLL expression in T-ALL relapse specimens vs. diagnosis specimens (Figure 2B) comparable to that measurable in GC resistant vs. sensitive specimens [17] (Figure 2A). This differential was only statistically significant for probe set 212079_s_at (p < 0.001, unpaired t-test), but the same trend was visible for the other four probe sets and is reflected in the summary of the median expression values for all five probes (Figure 2B). Since both of these studies involve T-ALL patients it is likely that the majority of patients within these cohorts do not have rearrangements affecting MLL. Taken together, this data provides clear support from two independent data sets that the correlation we have observed between wild-type MLL expression and GC sensitivity in T-ALL in vitro appears to also be relevant in vivo.

In our T-ALL cell lines we observed a 35-fold variation in MLL expression across the panel that correlated with GC resistance (Figure 1A). To assess the degree with which endogenous MLL expression levels vary in primary ALL patient specimens we analyzed data published by Ross et al who performed gene expression profiling of pediatric ALL subtypes [18]. Figure 3A shows that of all the pediatric ALL subtypes, the widest variations in MLL expression levels are found in patients with T-lineage ALL and those with MLL-rearrangements. To examine the prognostic relevance of MLL expression variation in patients with MLL-disease we examined a publication describing the use of Affymetrix HG-U95v2 microarrays to examine gene expression patterns in ALL patients with MLL-rearrangements [16]. These authors reported that such patients could be clustered on the basis of their genome-wide transcriptional profile into two distinct subgroups (called A and B) that demonstrated dramatically different survival rates (Figure 3B, box, p = 0.0005). By analyzing the data from their study we have ascertained that the expression of MLL was significantly lower in poor-outcome patients (Group A) compared to those with good outcome (Figure 3B, bar chart, p = 0.008). The HG-U95v2 probe for MLL targets the 3' UTR of the gene, meaning that it would either detect expression of the full-length (non-translocated) MLL allele remaining in these patients, or the expression of any reciprocal fusion that was transcribed as far as this 3' probe. Certainly it would not detect signal from primary MLL-translocation products. While the authors did not experimentally determine GC sensitivity in their study [16], the data are consistent with the hypothesis that the level of wild-type MLL expression is linked to therapeutic outcome even in patients that have an MLL-translocation on the alternate allele.

To assess the role of wild-type MLL in GC resistance phenotypes we used a retroviral RNAi expression system in the PER-117 T-ALL cell line to generate cell lines stably expressing shRNA for MLL (MLL-KD) or a non-silencing shRNA scrambled control (MLL-Scr). MLL mRNA expression in MLL-KD cells was 69% lower on average than in MLL-Scr control cells as assessed by qRT-PCR (Figure 4A, p < 0.0001). This correlated to a ~20% reduction in MLL nuclear protein as assessed by immunoblot (Figures 4B and 4C, p < 0.05). Proliferation assays demonstrated that MLL-KD cells grew approximately 10% faster on average than MLL-Scr cells (Figure 4D, p < 0.05 ANOVA). To assess GC sensitivity in the two lines, cell viability was assessed after a two-day incubation with dexamethasone (Figure 4E). MLL-KD demonstrated increased viability compared to MLL-Scr cells at all doses tested (p < 0.05, two-way ANOVA) indicating GC resistance. To assess the specificity of this protective effect we examined the sensitivity of the cells to gamma-irradiation, and incubation with cytarabine (ARAC) and methotrexate (MTX). Interestingly, MLL-KD cells showed greater survival following gamma-irradiation indicating resistance to DNA damage (Figure 5A, p < 0.05 unpaired t-test). Resistance to ARAC and MTX however was not significantly different between the two cell lines. The proportion of dying (necrotic) cells after two days was significantly reduced in MLL-KD cells in response to both dexamethasone and gamma-irradiation, indicating a cytoprotective effect of MLL knockdown (Figure 5B). Baseline viability in untreated cells was not significantly different between the cell lines.

To assess the effects of MLL knockdown on cell metabolism we compared rates of glucose consumption and lactate production between the two cell lines. Consistent with an increased rate of proliferation MLL-KD cells demonstrated an increased rate of glucose consumption compared to control cells. This was accompanied by a decreased rate of lactate production, resulting in a significant drop in the lactate production:glucose consumption ratio in MLL-KD cells (Figure 5C). Finally, since MLL is known to be a master transcriptional regulator we assessed whether the GC resistant phenotype of MLL-KD cells might represent transcriptional suppression of GC response elements by measuring the induction of GILZ, a well-characterized GC-response gene, following incubation with dexamethasone. There was no significant difference in the induction of GILZ mRNA between MLL-KD and MLL-Scr cell lines following a 4 hour incubation with dexamethasone (Figure 5D), indicating that GC-transcriptional responses in MLL-KD cells appeared to be intact.