Glucocorticoid-regulated microRNAs and mirtrons in acute lymphoblastic leukemia

Abstract

Glucocorticoids (GCs) induce apoptosis in lymphoid lineage cells and are therefore used in the therapy of acute lymphoblastic leukemia (ALL) and related malignancies. MicroRNAs (miRNAs) and the related mirtrons are ∼22 nucleotide RNAs derived from polymerase-II transcripts and implicated in the control of essential biological functions, including apoptosis. Whether GCs regulate miRNA-encoding transcription units is unknown. We investigated miRNA/mirtron expression and GC regulation in 8 leukemia/lymphoma in vitro models and 13 ALL children undergoing systemic GC monotherapy using a combination of expression profiling techniques, real time reverse transcription (RT)-PCR and northern blotting to detect mature miRNAs and/or their precursors. We found that mature miRNA regulations can be inferred from expression data of their host genes. Although a simple miRNA-initiated canonical pathway to GC-induced apoptosis or cell cycle arrest did not emerge, we identified several miRNAs/mirtrons that were regulated by GC in patients and cell lines, including the myeloid-specific miR-223 and the apoptosis and cell cycle arrest-inducing miR15∼16 clusters. In an in vitro model, overexpression of miR15b∼16 mimics increased and silencing by miR15b∼16 inhibitors decreased GC sensitivity. Thus, the observed complex changes in miRNA/mirtron expression during GC treatment might contribute to the anti-leukemic GC effects in a cell context-dependent manner.

Introduction

Glucocorticoids (GCs) have pronounced effects on metabolism, differentiation, proliferation and cell survival in many tissues. In the lymphoid system, they affect cell cycle progression, influence immunoglobulin and lymphokine production and, most notably, induce apoptosis in immature lymphoblasts. The latter has been implicated in the generation of the immune repertoire and the regulation of immune responses,^{1, 2, 3} and is used clinically in the treatment of childhood acute lymphoblastic leukemia (ALL) and other lymphoid malignancies.⁴ GCs mediate their effects through the GC receptor (GR), a ligand-activated transcription factor of the nuclear receptor super-family that resides in the cytoplasm and, upon ligand binding, translocates into the nucleus, where it modulates gene expression by binding to specific DNA response elements or by protein–protein interactions with other transcription factors.⁵ A large number of protein-encoding genes have been identified that are regulated by GCs in lymphoid lineage cells in experimental systems⁶ and related clinical samples,^{7, 8} but the genes responsible for cell death induction are not well understood (refer recent reviews by Schaaf and Cidlowski,² Schmidt et al.,⁶ Distelhorst,⁹ Haarman et al.¹⁰).

MicroRNAs (miRNAs) are tiny non-coding RNAs that induce post-transcriptional gene silencing through base pairing with their target mRNAs (for recent reviews and a mammalian miRNA expression atlas refer Kim and Nam,¹¹ Filipowicz et al.,¹² Grosshans and Filipowicz,¹³ Cullen,¹⁴ Bartel,¹⁵ Landgraf¹⁶). They are transcribed by RNA polymerase II as long primary transcripts referred to as ‘pri-miRNAs’. miRNAs are encoded by one arm of a stem-loop structure embedded in introns or, less frequently, exons of protein-coding or non-coding transcripts. In many cases, the corresponding host genes and transcripts have been defined.¹⁷ Subsequent to transcription, the pri-miRNAs stem-loop is cleaved by Drosha, an RNase III family member, to generate ∼70 nucleotide (nt) precursors called pre-miRNAs. In some instances, an entire intron consists of such a stem-loop structure, which is released by the splicing machinery in a Drosha-independent manner. miRNAs generated by this mechanism are referred to as ‘mirtrons’.^{18, 19, 20} Pre-miRNAs are exported by Exportin-5 to the cytoplasm, where they are further processed by Dicer, another RNase III enzyme, to generate ∼22 base pair intermediates that enter effector complexes called miRISC (miRNA-containing RNA-induced silencing complex). Here, they are rapidly converted into single-stranded ‘mature miRNAs’ that target mRNAs and thereby affect their translation and/or stability.^{21, 22}

Expression and regulation of miRNAs/mirtrons and their host genes can be studied at the level of the primary transcripts (pri-miRNAs), as polyadenylated and spliced mRNAs, as ∼70 nucleotide pre-miRNAs or as ∼22 nucleotide mature miRNAs using different techniques (Figure 1). A strong correlation between the tissue-specific expression patterns of the flanking mRNA and the embedded miRNA has been observed for some miRNAs,^{14, 17} suggesting that reliable mature miRNA expression patterns might be derived from host gene expression data. This raises the attractive possibility of exploiting conventional whole genome expression profiles not only for defining expression and regulation of miRNA-containing host genes, but also for obtaining relevant information regarding the corresponding mature miRNAs (provided miRNA and host gene are transcribed in the same orientation). Such a strategy is particularly relevant in the case of existing expression profiling data derived from clinical samples that are often unique and/or difficult to obtain.

Figure 1

miRNA biogenesis and detection. Three types of miRNA-containing host genes (‘exonic’, ‘intronic’ and ‘mirtrons’) are shown at the level of the corresponding primary transcripts (pri-miRNAs), where they are first processed by the Drosha and the conventional RNA processing pathway (‘exonic’ and ‘intronic’ pri-miRNAs) or the conventional RNA processing pathway only (mirtrons) to generate pre-miRNAs that are exported into the cytoplasm and further processed by the Dicer complex to generate mature miRNAs. The large box on top of the Figure includes all miRNA-related sequences that can be detected on the Exon 1.0 ST array (that is, pri-miRNAs, pre-miRNAs and miRNA-related mRNAs). Only the latter can be detected on the U133 Plus 2.0 (shaded box within the large box). The box at the bottom indicates techniques to identify and quantify mature miRNAs (that is, northerns, RT-PCR, mirVANA Bioarrays).

Despite advances in our understanding of miRNA biology and their growing role in human cancer and other diseases,^{23, 24} little is known about the transcriptional regulation of miRNAs. More specifically, whether expression of miRNAs is regulated by the GR and, if so, whether this might occur in ALL lymphoblasts exposed to GC treatment, has not been addressed. Using a childhood T-ALL in vitro model for GC-induced apoptosis, we identified several GC-regulated mature miRNAs. By using a number of techniques, we found that expression and regulation of these fully processed miRNAs could frequently be inferred from analyses of transcripts from their host genes (that is, pri-miRNAs, pre-miRNAs and other miRNA-related transcripts, Figure 1). Encouraged by these findings, we searched for GC-regulated miRNAs in whole genome expression profiles from malignant lymphoblasts of ALL children during systemic GC monotherapy, thereby defining the GC response of the majority of the currently known miRNA-containing genes in these patients.

Materials and methods

The details of this section are available online as Supplementary Information.

Results

Detection of GC-regulated mature miRNAs in CCRF-CEM T-ALL cells

To investigate whether GCs regulate the expression of mature miRNAs in human ALL cells, small RNAs from GC-treated and vehicle control-treated CCRF-CEM-C7H2 cells were investigated for expression of 320 human miRNAs using the mirVana miRNA Bioarrays version 1 (see Supplementary Table S3 for complete data). Five miRNAs were regulated more than twofold on both replicated spots after dye swap normalization. miR-19b and -181a were regulated only at the early time points, induction of miR-223 was only detectable after 24 h, and miR-15b and -16 showed increasing induction reaching significant levels after 24 h (Table 1a).

Table 1 Detection of GC-regulated miRNAs and their precursors by microarry-based expression profiling^a

We next attempted to validate the regulations seen on the Ambion arrays by two additional approaches, that is, quantitative real time reverse transcription (RT)-PCR and northern blotting. Induction of miR-19b and -181a could not be reconfirmed by either method even though we used two independent RT-PCR systems (data not shown). In contrast, the inductions of miR-223, -15b and -16 were clearly seen by real time RT-PCR and northern blotting (Figure 2). Therefore, we concluded that the expression of the mature forms of miR-223, -15b and -16 is induced by GC in this human T-ALL cell line.

Figure 2

Verification of glucocorticoids (GC) regulation of miR-15b, -16 and -223 by real time RT-PCR and northern blotting. (a) CEM-C7H2 cells were treated for 2, 6, 12 and 24 h with 100 nM dexamethasone or 0.1% ethanol as control, and expression of miR-223, -15b and -16 was quantified by real time RT-PCR. Shown are M-values (log2 fold change values of regulations between dexamethasone and control samples) ±s.d. from three independent experiments performed in triplicate and normalized to miR-320, which served as unregulated control. P-values <0.05 are indicated. (b) Twenty μg pooled total RNA from CEM-C7H2 cells treated in biological triplicates for 12 and 24 h with 100 nM dexamethasone or 0.1% ethanol as control were size-separated and probed with [³²P] labeled probes for the indicated miRNAs. Probes for 5.8S rRNA served as loading control.

Detection of GC-regulated miRNA-related transcripts on expression profiling microarrays

In the next set of experiments, we investigated the extent to which regulations observed on the level of mature miRNAs could be detected on the level of transcripts derived from their host genes (that is, pri-miRNAs, pre-miRNAs and miRNA-related transcripts as shown in Figure 1). To this end, we determined expression profiles of CEM-C7H2 cells treated with GC in biological triplicates for 6 h and 24 h on two platforms, that is, the Affymetrix U133 Plus 2.0 and the Exon 1.0 microarrays. Neither platform detects mature miRNAs, as they are lost during target preparation due to their small size. However, as explained in Figure 1, various precursors and related transcripts can be detected depending on the probes (25 nucleotide-long oligomers) present on the array and the route by which the RNA is converted into labeled ‘targets.’ For U133-type arrays (also referred to as 3′ arrays), the RNA is reversely transcribed using oligo(dT) primers, hence only polyadenylated transcripts from miRNA-encoding genes can be detected (such transcripts may or may not contain miRNA sequences, as shown in Figure 1). Targets for Exon arrays are generated by random priming, ensuring that all RNA species are labeled. Depending on the probes used to generate the signal (a process called ‘summarization’), Exon arrays may detect primary transcripts from miRNA-containing genes, their polyadenylated transcripts and, in addition, Drosha-processed pre-miRNAs (Figure 1).

We identified all probe sets on Exon 1.0 and U133 Plus 2.0 arrays that detect miRNA- and mirtron-related transcripts (Supplementary Table 1) and determined whether their expression was regulated after 6 h and/or 24 h exposure to GC (see Supplementary Table S4 and S5 for complete exon and U133 data, respectively). In agreement with the real time RT-PCR and northern blotting experiments, the induction of miR-19b and -181a could not be confirmed on exon or U133 arrays (Tables S4 and S5). In contrast, GC induction of the gene encoding miR-223 (DQ680071) and SMC4, which contains miR-15b and -16-2 in one of its introns,¹⁷ was readily detected on both platforms (Table 1b and c).

In addition to miR-223 and the miR-15b∼16-2 cluster, three other GC regulations of miRNA-related transcripts were detected both on the Exon and U133 2.0 arrays (Table 1b and c), that is, induction of ATAD2 (miR-548d) and PRKCA (miR-634), and repression of MIRH1/C13orf25 encoding the miR17∼18a∼19a∼20a∼19b-1∼92-1 cluster. The miRNAs encoded by the induced transcripts (that is, miR-548d, miR-634) were not present on the Ambion array, which explains why they were not detected on this platform. In contrast, the miRNAs encoded by the repressed MIRH1 transcript (miR-17, -18a, -19a, -20a, -19b and -92) were present on the Ambion array, yet no repression was recorded.

To address this discrepancy, we performed additional real time RT-PCRs for mature miR-548d (ATAD2), miR-634 (PRKCA), and miRNAs 19a, 19b and 92 as representatives for MIRH1. Although miR-634, a novel miRNA candidate only found in humans,²⁵ was not detectable in CEM cells nor in seven other cell lines (see further below), the miR-548d induction seen on the Exon 1.0 und U133 Plus 2.0 arrays was reconfirmed (see Supplementary Table S7). Thus, with the exception of a non-conserved candidate miRNA (miR-634), all inductions detected by microarray technology were confirmed on the level of mature miRNAs. In contrast to this, but in agreement with the Ambion array, repression of the investigated members of the MIRH1 cluster, that is, miR-19a, -19b and -92, was not observed. One explanation of why the repressions observed on microarrays were not seen at the level of mature miRNAs might be that the latter are relatively stable, and hence their repression becomes detectable only after a longer time period. In the CCRF-CEM system, however, such investigations are precluded by the fact that these cells start undergoing apoptosis shortly after 24 h.

To address the above hypothesis, we determined the half-life time of two exemplary miRNAs. We treated two stably transfected CCRF-CEM derivatives with tetracycline-regulated expression of a GFP-miR15b∼16-2 fusion transcript for 3d with doxycycline to induce the fusion transcript and compared the stability of the entire transcript and the mature miRNA after doxycycline withdrawal. Although the recombinant precursor RNA, as detected by real time RT-PCR for eGFP, disappeared rapidly after washing out doxycycline, the expression levels of mature miR-15b and -16 declined much more slowly (Figure 3). If this observation can be extended to other miRNAs, it would explain why we were unable to reconfirm repressions seen on the microarrays on the level of mature miRNAs.

Figure 3

Analysis of miRNA half-life. Two CCRF-CEM derivatives with tetracycline-regulated expression of an eGFP-miR15b∼16-2 fusion transcript (C7H2-2C8-eGFP-miR15b∼16-2 nos. 6 and 23) were treated for 3d with 500 ng/ml doxycycline to induce the fusion transcript, washed, re-cultured in the absence of doxycycline and the expression of eGFP, and mature miR-15b and miR-16 determined by real time RT-PCR after the times indicated. TBP and miR-320 served as housekeeping controls, respectively. The values obtained after doxycycline exposure were set at 100% and all other values expressed as percentage thereof. Mean values ±s.d. for two independent experiments with both cell lines analyzed in triplicate are shown.

The combined data show that conventional expression profiling platforms can provide important information concerning expression and regulation of a large number of miRNA-containing genes. Moreover, regulations of the miRNA-containing host genes appear to be transmitted to the level of the corresponding mature miRNAs, although possibly with considerable delay in the case of gene repressions. Altogether, these findings encouraged us to extend our analyses to expression profiles from children with ALL.

GC-regulated miRNA host gene transcripts in lymphoblasts from ALL children undergoing systemic GC monotherapy

As the above data further supported the reported good correlation between expression of mature miRNAs and their flanking mRNAs,^{14, 17} we exploited our previously determined expression profiles derived from 13 children (3 T-ALLs, 10 precursor B-ALLs) during 6–8 h and 24 h of systemic GC monotherapy⁷ to search for GC-regulated miRNAs. On the U133 Plus 2.0 array, 491 probe sets detecting mRNAs derived from 182 miRNA-encoding genes were identified (Supplementary Table S1). When the 13 children were considered as a group of biological replicates, the best regulated miRNA-encoding transcripts were the induced DQ680071 mRNA encoding miR-223 and the repressed ATAD2 and MCM7 transcripts encoding miR-548d-1 and miR-106b∼93, respectively (Table 1d; complete data in Supplementary Table S6a).

As this type of analysis may obscure events in subgroups of patients, we also analyzed the response of individual children (Supplementary Table S6b and c). Using >2-fold regulation in four or more individuals as an arbitrary cutoff, three additional probe sets were identified detecting the mirtron miR-1233 containing GOLGA8A (induced in four children after 24 h), the miR-128b encoding ARPP21 (repressed in four children after 24 h) and a transcript from the DLEU2 locus (induced in four children after 6–8 h). The latter locus is particularly interesting because it encodes the miR15a∼16-1 cluster, which has been implicated in the pathogenesis of B-CLL^{26, 27} and, if overexpressed, results in apoptosis²⁸ and cell cycle arrest.²⁹ Interestingly, although this cluster was not induced in CCRF-CEM cells, the highly related miR15b∼16-2 cluster encoded by the SMC4 gene was induced.

Combining the responses in the ALL children and the CCRF-CEM model system highlights not only several quite dramatic differences (such as opposite regulation of ATAD2/miR-548d-1), but also remarkable commonalities. Thus, miR-223 and members of the apoptosis inducing miR-15/16 family were increased by GC both in the cell line model and in ALL patients.

miR-223 and miR-15/16 family members in ALL cell lines

We next examined whether miR-223 and miR-15/16 family members were regulated by GC in several widely used leukemia/lymphoma cell lines, and whether this might be related to GC sensitivity. We determined apoptosis induction by GC in eight cell lines and performed real time RT-PCR for mature miR-223, -15b and -16, as well as for some other miRNAs (that is, miR-181a, 19a, 19b, 320, 548d, 634, 661, 92 and let7b). As shown in Figure 4a, the T-ALL cell lines MOLT-4 and Jurkat and the Burkitt lymphoma line Daudi were resistant to cell death induction by GC, whereas the precursor B-cell leukemia lines NALM-6, RS4;11, PreB-697 and a GR-transfected subclone of the Jurkat T-ALL cell line were sensitive to GC-induced apoptosis, although to different degrees. The GC-resistant MOLT-4, untransfected Jurkat and Daudi cell lines did not regulate any of the investigated miRNAs, but this was also true for the GC-sensitive PreB-697 and NALM-6 precursor B-cell lines (Supplementary Table S7). RS4;11 cells induced miR-15b with a mean M-value of 1.3, but this was statistically below cutoff (pBH=0.06) and miR-223 showed a tendency for upregulation in these cells (M=0.6, pBH=0.08, Supplementary Table S7). Interestingly, however, GR transfected, and thus GC-sensitive, Jurkat cells induced miR-223, -15b and -16, similar to the GC-sensitive CCRF-CEM T-ALL line (Figure 4b and Supplementary Table S7). Taken together, miR-223 and miR-15/16 family members were induced by GC in a subset of children with ALL and in some ALL cell lines that, in the Jurkat model system, correlated with GC sensitivity.

Figure 4

miRNA regulation by glucocorticoids (GC) in leukemic cell lines. (a) GC-induced apoptosis in leukemia cell lines. CCRF-CEM-C7H2, Jurkat (untransfected and transfected with rat GR), MOLT-4, PreB697, NALM-6, RS4;11 and Daudi cells were cultured in the presence of 10⁻⁷ M dexamethasone or 0.1% ethanol as carrier control for the time indicated and subjected to apoptosis determination using flow cytometric analysis of propidium iodide-stained nuclei. Shown are mean values ±s.d. of specific apoptosis (apoptosis in GC-treated samples minus apoptosis in corresponding vehicle controls) of biological triplicates. Untransfected Jurkat, MOLT-4 and Daudi cells were resistant to dexamethasone and the values from these cell lines were pooled. (b) GC-regulated miRNAs in leukemia cell line. The indicated mature miRNAs were quantified in total RNA from untransfected Jurkat (GC-resistant), Jurkat transfected with rat GR (GC-sensitive) and GC-sensitive CCRF-CEM-C7H2 cultured in the presence of 10⁻⁷ M dexamethasone or 0.1% ethanol as carrier control for 24 h by real time RT-PCR using miR-320 as housekeeping miRNA for data normalization. Shown are regulations between dexamethasone-treated and control cells expressed as mean M-values (log2 fold changes) ±s.d. of three biological replicates, each determined in technical triplicates. P-values <0.05 are indicated by an asterisk. Additional miRNAs (hsa-let-7b, miR-15a, -19a, -19b and -92) and cell lines (MOLT-4, PreB697, NALM-6, RS4;11 and Daudi) analyzed as above gave no significant regulations with M-values better than ±1 (complete data in Supplementary Table S7).

Functional analyses of miR15b∼16

To address whether GC regulation of the miR15/16 family might affect the sensitivity of leukemia cells to GC-induced apoptosis, we transiently transfected CEM-C7H2 cells with a cocktail of miRNA-mimics or -inhibitors corresponding to miR15b and miR-16 to mimic or inactivate their effects, respectively. In all experiments, transfection efficiency as determined by co-transfection with a labeled oligonucleotide (BlockIT) was >85%. In the case of miRNA mimics, this led to a strong increase of the respective miRNA as determined by miRNA-specific real time RT-PCR, whereas miRNA inhibitors reduced the corresponding miRNA by two- to fourfold (Figure 5, right panels). Neither mimics nor inhibitors had any detectable effect on cell survival on their own; however, miR15b/16 overexpression increased, whereas their specific inhibition decreased GC-induced apoptosis to a weak but significant extent (Figure 5, left panels).

Figure 5

Functional analysis of miR15b∼16 in CCRF-CEM T-ALL cells. CCRF-CEM-C7H2 cells were transfected with a mixture containing either mimics (a) or inhibitors (b) of miR-15b and miR-16 or corresponding controls along with an FITC-labeled second control oligonucleotide (‘BlockIT’, Dharmacon Inc., Lafayette, CO, USA) for assessing transfection efficieny. Twenty-four hours after transfection, the cells were treated with 10⁻⁷ M dexamethasone (Dex +) or 0.1% ethanol (Dex −) as carrier control for the time indicated, and subjected to determination of apoptosis using the AnnexinV/propidium iodide method. Mean percentages of cell death ±s.d. from three independent experiments are shown. The plots on the right show expression levels of the indicated miRNAs as determined by real time RT-PCR expressed as −ΔCT (that is, CT value of miR-15b or -16 minus CT of miR320, an unregulated control) from aliquots of the same cells shown in the plots on the left 48 h after transfection and 24 h after exposure to Dex (+) or 0.1% ethanol (−). Filled symbols represent measurements from samples treated with miRNA mimics (a) and inhibitors (b), open symbols depict the corresponding controls.

Discussion

This study shows, for the first time, GC-induced changes in the expression of human mirtrons and miRNAs in ALL cell lines and children with ALL during systemic GC monotherapy. By using a combination of techniques to detect various types of miRNA-related sequences (Figure 1) in the CCRF-CEM model, we show that mature miRNA expression and regulation can be inferred from microarray-based expression analyses of their host genes. Although post-transcriptional regulation of miRNAs has been reported for some miRNAs,^{30, 31} our data combined with those of others^{14, 17} suggest that, in many instances, expression/regulation of miRNA-containing host genes determines the fate of their mature forms. Thus, we found a good correlation between regulations of mature miRNAs and the corresponding transcripts detected on two types of microarrays when gene inductions were compared. Interestingly, repressions detected on the microarrays were not paralleled by repression of the corresponding mature miRNA. This may be explained by the increased stability of mature miRNAs compared with their precursors, as exemplified for the miR-15b∼16-2 cluster (Figure 3). With some delay, however, the expected reduction of mature miRNAs did occur. Taken together, analysis of miRNA-related transcription units in conventional expression profiling data sets might provide useful information as a starting point for further analyses.

How can our findings be incorporated into existing or novel concepts concerning the molecular basis of the anti-leukemic (or other) effects of GC, in particular GC-induced leukemia cell apoptosis? In general terms, the observed regulations were moderate (∼2-fold) and appeared to be cell context-dependent, that is, they were not observed in all patients and cell lines and, in some instances, occurred in opposite directions (such as miR-548d/ATAD2, which was induced in CCRF-CEMs and Jurkats but repressed in several children, Table 1). Such a heterogeneous response pattern in systems sensitive to GC-induced apoptosis argues against a simple canonical pathway initiated by GC-regulation of any one of the investigated miRNA-containing transcription units. This does, however, not exclude the possibility that some of them might contribute to GC-induced apoptosis or cell cycle arrest in certain cell lines or patients. One of the most interesting candidates in this respect is the miR-15∼16 family comprised of miR-15a, -15b and -16. These miRNAs are encoded in two clusters (15a∼16-1 and 15b∼16-2) embedded in the DLEU2 and SMC4 loci, respectively.¹⁷ They have been implicated in cell cycle arrest²⁹ and in cell death/survival decisions, the latter supposedly by targeting BCL2.²⁸ Interestingly, BCL2 as well as Bcl-XL, a predicted target of miR-15/16, are repressed at the protein level by GC in CCRF-CEM cells.³² Thus, the observation of GC induction of the SMC4 locus in two GC-sensitive ALL models (Figure 4b) and of transcripts from the DLEU2 locus in some ALL children (Supplementary Table S6b and c) raised the possibility that transcriptional induction of the miR-15/16 family might contribute to GC-induced apoptosis in a subset of ALL patients and cell lines. This concept was supported, at least for the CCRF-CEM T-ALL model, by functional analyses (Figure 5) showing a weak but significant effect on GC-induced cell death.