Implication of IRF4 Aberrant Gene Expression in the Acute Leukemias of Childhood

The most frequent targets of genetic alterations in human leukemias are transcription factor genes with essential functions in normal blood cell development. The Interferon Regulatory Factor 4 (IRF4) gene encodes a transcription factor important for key developmental stages of hematopoiesis, with known oncogenic implications in multiple myeloma, adult leukemias and lymphomas. Very few studies have reported an association of IRF4 with childhood malignancy, whereas high transcript levels have been observed in the more mature immunophenotype of ALL. Our aim was to investigate the expression levels of IRF4 in the diagnostic samples of pediatric leukemias and compare them to those of healthy controls, in order to determine aberrant gene expression and whether it extends to leukemic subtypes other than the relatively mature ALL subpopulation. Quantitative real-time RT-PCR methodology was used to investigate IRF4 expression in 58 children with acute leukemias, 4 leukemic cell lines and 20 healthy children. We show that aberrant IRF4 gene expression is implicated in a variety of leukemic subtypes; higher transcript levels appear in the more immature B-common ALL subtype and in T-cell than in B-cell leukemias, with the highest expression levels appearing in the AML group. Interestingly, we show that childhood leukemia, irrespective of subtype or cell maturation stage, is characterised by a minimum of approximately twice the amount of IRF4 gene expression encountered in healthy children. A statistically significant correlation also appeared to exist between high IRF4 expression and relapse. Our results show that ectopic expression of IRF4 follows the reverse expression pattern of what is encountered in normal B-cell development and that there might be a dose-dependency of childhood leukemia for aberrantly expressed IRF4, a characteristic that could be explored therapeutically. It is also suggested that high IRF4 expression might be used as an additional prognostic marker of relapse at diagnosis.


Introduction
Acute leukemia (AL) is regarded the most common type of malignancy in children. Despite the good overall response of childhood patients to current chemotherapeutic treatments and the long-term event-free survival rates exceeding 80% [1,2], a significant percentage of the population show resistance to therapy and/or relapse, often with devastating consequences. This has highlighted the importance of devising new therapies that will target the resistant clones and increase overall survival. Of significant interest are the genetic alterations that might affect the characteristics of hematopoietic stem cells (HSCs). For example, it has already been demonstrated that a subset of genes expressed in normal HSCs are highly reactivated in leukemic stem cells (LSCs), thus indicating that the LSCs express the self-renewal-associated programme normally characterising the HSCs [3]. Such findings suggest that leukemic transformation of progenitor cells is associated with ectopic reactivation of the genes responsible for self-renewal. Indeed, the presence of MLL rearrangements in both acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML) seems to relate with high level expression of genes normally expressed in HSCs during the early stages of hematopoiesis, such as FLT3 and HOXA9 [4][5][6][7][8][9][10]. Consequently, the identification of specific genetic aberrations that characterise leukemic cells, such as abnormal gene expression, can lead to a better understanding of what constitutes their unlimited self-renewing properties and subsequently reveal potential molecular targets for successful therapy.
The most frequent targets of genetic alterations in human leukemias are transcription factor genes with essential functions in normal blood cell development. The Interferon Regulatory Factor 4 (IRF4) gene encodes a transcription factor important for hematopoietic development and immune responses. It is essentially involved in all developmental stages within the B-cell lineage (except during the germinal center (GC) reaction), with known critical functions in at least 3 key developmental processes: the termination of the GC B-cell transcriptional program, immunoglobulin class-switch recombination, and plasma cell development [11][12][13]. IRF4 is induced by mitogenic stimuli, such as antigen receptor engagement, lipopolysaccharides, and CD40 signaling [14,15], which activate the Nuclear Factor-kappa B (NF-kB) pathway, ultimately leading to IRF4 promoter activation [16][17][18]. Despite functioning as a tumor suppressor gene in early B-cell development [11,19], IRF4 is a well-established oncogene in multiple myeloma (MM) [20], with oncogenic implications extending to certain adult lymphomas and leukemias [13,[21][22][23]. High expression levels have been detected in chronic lymphocytic leukemia (CLL), where they have been linked to an unfavourable prognosis [24][25][26][27]. Very few studies have reported an association of IRF4 with childhood malignancy. In childhood ALL, IRF4 has been found significantly up-regulated in the more mature CD34-CD19 + LSC population [28], whereas others have reported that an intronic polymorphism of the gene relates to a greater risk association of males to the disease [29].
Considering the importance of IRF4 in key developmental stages of hematopoiesis, as well as its oncogenic implication in malignancies such as lymphomas and chronic leukemias, investigating its expression patterns in childhood leukemias could help in the understanding of how deregulation can contribute to leukemogenesis. In this study we investigate the transcript levels of IRF4 in the diagnostic samples of a cohort of 58 pediatric acute leukemia patients and 4 cell lines, including both ALL and AML, and compare them to those of healthy controls. Our aim is to determine whether aberrant IRF4 expression extends to childhood AML or to leukemic subtypes other than the more mature CD34-CD19 + ALL population. In addition, we examine whether there are any correlations between gene expression and specific cytogenetic characteristics, as well as other diagnostic variables, such as age, gender, and karyotype. Finally, we investigate the prognostic impact of high IRF4 expression in the clinical course of childhood leukemia, through possible correlations with minimal residual disease (MRD) and survival outcome.  Table S1. BM samples were also collected from 20 healthy children to be used as negative controls for the qRT-PCR studies. Four childhood leukemia cell lines were also included in the study: REH (pre-B ALL), CCRF-CEM (T-cell ALL), CCRF-SB (T-cell ALL), and THP-1 (acute monocytic leukemia). All cell lines were obtained from the European Collection of Cell Cultures (ECACC, UK).

Sample handling
Red Blood Cell Lysis Buffer (RBC Lysis Buffer) was added to every BM sample (2:1 ratio), in preparation for RNA extraction. Every sample was vortexed and left at room temperature for 3-5 minutes to ensure sufficient lysis of red blood cells, then centrifuged at 3,000 rpm for 7 minutes, so that the white blood cells, intact, collected at the bottom of the tube, while the lysed red blood cells remained in the liquid supernatant. For immunophenotyping, samples were collected in EDTA tubes containing 1 ml of Hank's Balanced Salt Solution for every 1 ml of BM specimen. Specimens were analyzed within 3 hours post collection.

RNA Extraction
Total RNA was extracted from BM leukocytes/myelocytes, using TRIzol reagent (Invitrogen, Inc.) according to the manufacturer's protocol. Extraction of RNA from the 4 cell lines was performed with the same protocol, only without the addition of RBC Lysis Buffer, for obvious reasons. RNA was treated with amplification grade DNase I to eliminate any residual genomic DNA from the sample and further purified with the RNeasyH Micro Kit (Qiagen Inc). RNA quantification measurements were performed on the NanoDropH ND-1000 Spectrophotometer (Nanogen Inc.).

Flow cytometry and Immunophenotype
Blast cell immunophenotyping was performed by direct immunofluoresence and 5-colour flow cytometry, on a FC500 instrument (Beckman Coulter, Brea CA). Apoptosis and necrosis measurements were performed as described previously [30].

Real Time Quantitative RT-PCR
Gene expression was investigated with the Real-Time Reverse Transcription PCR (qRT-PCR), using the Plexor TM One-Step qRT-PCR System kit (Promega Inc.) and the LightCycler 2 Instrument (Roche Diagnostics GmbH). Both the reverse transcriptase and DNA polymerase were added during the initial reaction setup, and the thermal cycler was programmed to perform the reverse transcription step, immediately followed by the thermal cycling program.
For each 25 ml PCR reaction, a 12.5 ml aliquot from the PlexorH Master Mix (26) was added to a LightCyclerH glass capillary (Roche Diagnostics GmbH), followed by 0.5 ml of RNasinH Plus RNase Inhibitor, 0.2 ml of ImProm-II TM Reverse Transcriptase, 2 ml of primers (1 ml Forward primer and 1 ml Reverse primer), 1 ml of RNA (20 ng/ml) and 8.8 ml of nucleasefree water. For the housekeeping gene reaction only 1 ml of primer was added, as both the Reverse and Forward primers were provided in the same tube by the manufacturer (Eurogentec S.A.).
The thermal cycling programme was as follows: 1 cycle of 45uC for 5 minutes (reverse transcription), 1 cycle of 95uC for 2 minutes (initial denaturation and inactivation of the reverse transcriptase), 40 cycles of 95uC for 5 seconds (denaturation) and of 63uC for 35 seconds (annealing and extension), and 60uC to 95uC, ramp 0.4uC/second intervals (melt temperature curve). The specificity of the PCR products was verified by gel analysis and products were shown to consist of only a single band.
The sequences of the primers in the 59 to 39 direction were as follows: IRF4F: 59-AG-CGC-ATT-TCA-GTA-AAT-GTA-AAC-ACA-T-39 and IRF4R: 59-TCT-TGT-GTT-CTG-TAG-ACT-GCC-ATC-A-39. Housekeeping genes GAPDH and b-actin were used for normalization purposes. In their majority experiments were performed with the GAPDH housekeeping gene.
Initially, method sensitivity was tested by creating a standard curve with kanamycin control RNA as provided by the manufacturer. The standard curve included six points at a final RNA concentration ranging from 50 ng to 0.0005 ng (ie. 0.5 pg) in 10-fold increments, therefore it was decided that for the patient and control samples a total of 20 ng would be sufficient for each reaction.
During the first 6 experiments (6 patients) reactions were performed in duplicates and experiments were performed in triplicates to assess intra-and inter-assay variability (no significant variability found). In all other experiments the patient and control samples were tested in triplicate reactions and experiments were performed once. The average value for all three reactions was used for quantification. Non-template negative control reactions were included with each run. All PCR products were electrophoresed on a 2% agarose gel to confirm successful amplification of the desired products.

Real-Time Data pre-processing and Analysis
Real-time data were collected and pre-processed using the LightCycler Software Version 3.5 (Roche Diagnostics GmbH). Following this, data were exported to the Plexor TM Analysis Software (www.promega.com/plexorresources) (Promega, Inc.) and gene expression was obtained in the form of Ct values; this refers to the PCR cycle number during exponential amplification at which the product (measured in real-time by fluorescence emission) crosses an arbitrary threshold. Data pre-processing was performed with MicrosoftH Excel.)

Statistical Analysis
First of all an independent t-test was performed between patients and controls to assess whether there were any differences in the mean GAPDH expression levels (Ct values) between the two groups. The t-test failed to reveal a statistically reliable difference (p = 0.068), so we accepted all GAPDH Ct values as reliable normalization constants. A second t-test was then performed to compare the mean IRF4 expression levels between the two groups, and a statistically significant difference was revealed (p = 0.003). The Ct value of IRF4 (for every patient and healthy control) was then normalized against the Ct value obtained for the housekeeping gene, GAPDH (ie. DCt = Ct of IRF4 -Ct of GAPDH) and the corrected gene expression levels were defined as the expression level for each gene divided by the GAPDH expression level. Taking the corrected expression levels (DCt) into account, each patient was tested for significance in differential expression using a z-test. Patients were regarded as significantly differentially expressing the gene if their z-score had a corresponding p-value of #0.05. T-tests were performed using the IBM SPSS Statistics program (

Immunophenotypic and cytogenetic subtypes
Immunophenotypes, karyotypes and cytogenetics are also presented in Table S1. Our results are in agreement with other studies regarding the frequencies of the immunophenotypic groups [31,32]. Similarly, our calculated cytogenetic frequencies also seem to be in agreement with those of other western countries [33].

Quantitative Real Time RT-PCR analysis
A quantitative real-time RT-PCR assay was established for the gene of interest. IRF4 was found to be differentially overexpressed, as compared to the healthy controls, in 30 patients  (Figure 1). The patients over-expressing IRF4 are also highlighted in bold in Table S1. Five (5) of these 30 patients and the 3 cell lines mentioned above revealed a much higher expression ratio (R.11.6) than the rest (p = 0.000). In addition, the gene was found to be differentially downregulated in 6 patients (,10%) ( Table 1). Therefore, patients were divided into 4 groups: subnormal ratio (ratio ,2.4), normal (healthy) ratio (2.4# ratio ,7), high ratio (7, ratio ,11.6) and very high ratio (11.6, ratio). In the subnormal ratio group, the mean IRF4 expression was significantly lower (p = 0.000) than in the healthy controls group (File S3). Overall, the gene was found up-regulated in a broader spectrum of leukemic subtypes and maturation stages than previously thought, with the highest transcript levels appearing in the AML group ( Figure 1).

Correlation Analysis Between Gene Expression and Cytogenetic Abnormalities
From our group of pediatric patients found to over-express IRF4, only 8 (26.7%) appeared to have one of the commonly found cytogenetic translocations, and that was either TEL-AML1 or MLL rearrangement, with the exception of one patient who was BCR-ABL+ve (Table S1). The majority of patients with high transcript IRF4 levels (n = 22) did not have a detectable cytogenetic translocation at diagnosis or had additional cytogenetic abnormalities, such as extra gene copy numbers, deletions, etc. No apparent correlation was found between high gene transcript levels and specific cytogenetic abnormalities.

Correlation Analysis Between Gene Expression and other diagnostic variables
No apparent correlation was found between abnormal IRF4 expression and age, gender, or karyotype (data not shown).

Correlation Analysis Between Gene Expression and MRD levels
No apparent correlation was found between abnormal IRF4 expression and resistance to therapy (as measured in MRD level on day 30). Only 3 patients had detectable MRD on day 30 since initiation of treatment (NYII protocol for ALL and AML-BFM 2004 protocol for AML), one of whom relapsed, along with another patient who relapsed but who had no detectable MRD on day 30 (Table S2).

Correlation Analysis Between Gene Expression and Overall Survival
No significant correlation was found between abnormal IRF4 expression and overall survival (OS) (as measured in years of survival from the diagnosis) on comparing patients with normal and patients with high IRF4 expression (Figure 2, A-H). Please note that Patients 56 and 57 were diagnosed at a much earlier date (2003) and so were regarded as outliers and were not included in the analysis. Even though we have observed twice as many deaths (4 versus 2) in the patient group over-expressing IRF4 (Table S2), the Kaplan Meier curves have not produced a statistically significant difference (p.0.05). The results for estimated leukemia-free time (i.e. time to relapse) were marginally significant, as it can be seen in Figure 2 (E & G) (p = 0.08, CI = 90% and

Discussion
In the present study we have found high IRF4 transcript levels in a significant number of patients (51.7%) and in 3 of the 4 cell lines, as compared to the healthy controls. This up-regulation seems to associate with a variety of leukemic subgroups, possibly reflecting the many roles of IRF4 in the immune system [12]. Considering that IRF4 is essential for several stages of normal B-and T-cell activation and differentiation, it is only logical to assume that aberrations in the IRF4 transcriptome would also reflect in lymphoid-cell associated malignancy.
A significant difference in IRF4 transcript levels was observed between the pre-B and B-common cohorts, as well as between the total B-and T-cell ALL groups (Figure 1), with higher transcript levels appearing in the B-common and T-cell subtypes, respectively. In the past, however, IRF4 mRNA expression has been found low in adult samples of T-cell ALL [34]. At this point it is difficult to validate the reasons for this difference, besides possible age-related dissimilarities in the biology of ALL, yet it points out that aberrant IRF4 transcriptional activity is developmental-stagespecific as well as cell-type-specific. It has already been demonstrated that IRF4 is expressed at varying levels throughout B-cell development, with expression peaking in plasma cells, therefore varying transcript levels in human malignancies could mirror IRF4 expression in normal lymphoid activation and differentiation [35,36]. However, in contrast to the Gene Expression Omnibus (GEO) data profiles from healthy tissues [37] and a recent study from le Viseur et al [28], in our leukemia samples we have observed that the greater the distance from a differentiated B-cell, the higher the level of IRF4 expression. In other words, significantly higher transcript levels were observed in the Bcommon subgroup, than in the more mature pre-B subtype (CD34 dim/ -CD19 + ). This suggests that, at least for the particular patient cohort studied here, ectopic expression of IRF4 does not exactly mirror the expression patterns encountered in normal Bcell development, but rather the situation seems to reverse. This observation could reflect differences in our experimental set-up, as opposed to the study from le Viseur et al [28], as we have not measured IRF4 expression in flow-sorted subpopulations from the same diagnostic sample but have rather based our analyses on the overall prevailing immunophenotype for every patient. On the other hand, the 3 cell lines, despite being phenotypically distant from differentiated cells, appeared to express even higher levels of the gene than the patients (all 3 fell into the very high ratio .11.6); this possibly implies that in childhood leukemia, IRF4 gene expression is not subject to the feedback regulation mechanisms that operate in normal tissues.
In the past, IRF4 expression was regarded an almost exclusive characteristic of the lymphoid lineage [15], but recent data have also demonstrated an involvement in myeloid cell differentiation [38][39][40]. The implication is still not entirely understood, but ectopic expression of IRF4 in myeloid progenitor cells in vitro was recently shown to inhibit myeloid cell growth and to tune the balance of lineage selection in myeloid progenitor cells by stimulating macrophage differentiation, whilst inhibiting neutrophil differentiation [41]. Subsequently, expression has been found down-regulated in chronic myeloid leukemias (CML), as well as in adult AML patient samples [34,42,43]. Here we report an association of IRF4 up-regulation with childhood AML, as it is in this patient cohort that we observe the highest transcript levels, including the AML cell line. Interestingly, 2 of the 4 AML patients over-expressing IRF4 fell into the very high ratio group (11.6, ratio) ( Table 1) and were in fact positive for M5 maturation, whereas high expression was also observed in the M1 stage (without maturation). This could also highlight the role of IRF4 in early myeloid cell differentiation and, subsequently, in myeloid leukemia without maturation but, since the number of observed cases here is small, no such conclusions can be made at this point. Analysis of a larger cohort of patients and of myeloid cell populations at various maturation stages should be able to confirm or reject the validity of this notion. The opposing expression levels between adult and childhood AML (low versus high) are most probably the result of the different pathogenetic mechanism that discriminates between adult and childhood leukemias.
Significant differentially low IRF4 expression, as compared to the healthy controls (p = 0.000) was observed in 6 of the patient samples that we analysed (,10% of total patients), all of which were B-cell ALL positive. This observation is consistent with the presence of a B-lineage-restricted repressor on the promoter of the IRF4 gene, which can become over-expressed in a cross-section of B-lineage leukemia [44,45]. Since IRF4 binds to its own promoter region, creating the potential for positive auto-regulation, its expression can reach levels that activate the repressor [17]. Other leukemic subtypes seem to lack the repressor and therefore do not exhibit abnormally low IRF4 expression. Furthermore, at least one micro-RNA species has been identified as an IRF4 repressor. In the B-cell lineage, hsa-miR-125b (GC-enriched hsa-miR-125b) has been shown to directly down-regulate the expression of IRF4 and of the repressor PRDM1/BLIMP1 [46]. While IRF4 expression in non-B-cell leukemia can be deregulated toward higher levels, in Blineage leukemia activation of the repressor may also facilitate a reduction in IRF4 transcript levels below the expected average.
Overall, as it can be seen in Figure 1, it seems that childhood leukemia, irrespective of subtype or cell maturation stage, is characterised by a minimum of approximately twice the amount of IRF4 gene expression encountered in healthy children (as measured in corrected expression levels (DCt)). This might suggest that there is a dose-dependency of childhood leukemia for ectopically expressed IRF4, a characteristic that could be explored further for its therapeutic potential. A therapy based on IRF4knockout would most likely be hindered by the potentially detrimental side effects. IRF4-deficient mice show profound defects of the immune system, such as impaired lymphocyte function, inability to generate cytotoxic, antibody and antitumor responses, accompanied by a severe reduction in serum immunoglobulin concentrations [14]. Surprisingly, mice lacking one allele of IRF4 appear phenotypically normal, whereas a 50% knockdown of IRF4 mRNA and protein has proved effective in clearing MM cell lines [14,47]. Perhaps the same should be applied to leukemic cells in order to evaluate whether modified or partial silencing of the gene appears to have an inhibitory or apoptotic effect, and thus whether it opens up a new opportunity of a therapy aimed at IRF4.
No significant correlation was found between high IRF4 expression and OS (Figure 2, A-H). This could be due to the small number of patients included in the analysis or perhaps to the fact that for most of these patients OS could not be calculated in terms of a 3-year or a 5-year survival. Whereas a marginally significant correlation appeared to exist between abnormal IRF4 expression and estimated leukemia-free time (p = 0.08 and p = 0.05) (Figure 2, E & G), a significant correlation appeared (p = 0.03) when leukemia-free time was estimated within a specific time frame of 3 years (Figure 2, I). This implies that high IRF4 expression might be used as an additional prognostic marker of relapse at the diagnosis of childhood acute leukemias. Further analyses on a larger number of patients over-expressing IRF4, over a longer period of time (i.e. 5-year survival), should be able to confirm this observation in the near future.
Additional efforts should be made in order to identify the putative targets of IRF4 and to determine the ones directly regulated by it. A number of proteins have already been identified as repressors in Bcells and found to interact with IRF4: PU.1, Spi-B, E47, NFAT, Stat6, Bcl-6 and Blimp-1 [14,38,[48][49][50]. In addition, CD40 signaling has been found to directly induce the expression of IRF4 in GC B-cells which, in turn, binds to the promoter region of the BCL-6 gene to repress its expression [51]. We are currently investigating the expression patterns of other hematopoietic progenitor genes in a larger cohort of pediatric patients in an attempt to gain a better understanding of how deregulated transcription on key signaling pathways affects the malignant phenotype and disease progression in childhood leukemia.

Conclusions
In summary, our study shows that aberrant IRF4 gene expression is implicated in a variety of leukemic subtypes, possibly reflecting the many roles of the gene in the immune system. In contrast to other studies, we observe higher transcript levels in patients characterised by a more immature immunophenotype, and in T-cell leukemias than in leukemias of the B-lineage, implying that in childhood leukemia, IRF4 gene expression is not subject to the feedback regulation mechanisms that operate in normal tissues. In addition, ectopic expression of IRF4 does not seem to be a specific feature of lymphoid leukemia, but it also extends to myeloid leukemia, as it was in our AML group that the highest expression levels were observed. Again, our results are in contrast to other studies documenting a down-regulated IRF4 expression in adult myeloid leukemias [34,42,43], even though this is most probably due to age-related dissimilarities and a different pathogenetic leukemia mechanism. Interestingly, we show that childhood leukemia, irrespective of subtype or cell maturation stage, is characterised by a minimum of approximately twice the amount of IRF4 gene expression encountered in healthy children.
This implies that there might be a dose-dependency of the disease for aberrantly expressed IRF4, a characteristic that could be explored therapeutically. In addition, a statistically significant correlation was found between high IRF4 expression and relapse, suggesting a prognostic value of IRF4 in the clinical course of childhood leukemias. Data from the past few years indicate that many important regulators of HSC development are also implicated in the abnormal self-renewal capacity of the LSCs. Only through the identification of these regulators, and of their downstream targets, will we gain a better understanding of what constitutes their unlimited self-renewing properties and subsequently reveal potential molecular targets for successful therapy.

Supporting Information
File S1 Independent t-Tests for patients and cell lines over-expressing IRF4 (both high and very high expression) and for patients under-expressing IRF4. (DOC)