Introduction
The clustered homeobox (
HOX) genes encode a large family of transcription factors characterized by the presence of a highly conserved nucleotide sequence called the homeodomain. This 61-amino-acid helix-turn-helix domain is responsible for the binding of HOX proteins to their target DNA sequences. In humans, the 39
HOX genes are organized into four genomic regions (the
HOXA, B, C and
D clusters) located on four chromosomes (chromosomes 7, 17, 12 and 2, respectively). Each cluster consists of 9 – 11 genes arranged in the same orientation and in paralogous groups [
1],[
2].
HOX genes play essential roles during embryogenesis by controlling cell fate along the anterior-posterior axis and specifying segment identity [
3]-[
5]. The characteristic expression of
HOX genes can also be detected in various adult tissues [
6],[
7]. During hematopoiesis, the highest expression of
HOX genes occurs in the stem and early hematopoietic progenitor cells. During maturation,
HOX expression gradually decreases, and it is minimal in differentiated hematopoietic cells [
8],[
9]. The expression of
HOX genes throughout the maturation of hematopoietic cells is tightly regulated, suggesting that disruption of this regulation contributes to the process of malignant transformation.
The oncogenic potential of
HOX genes in leukemia has been intensively studied for more than two decades. Several chromosomal translocations in leukemia involve
HOX genes either directly (e.g.,
NUP98-HOX fusion) or via their upstream regulators (e.g., MLL rearrangements) [
10]-[
13]. Moreover, the overexpression of certain
HOX genes and their cofactors are known as poor prognostic markers in leukemia patients [
14]-[
16]. The overexpression of
HOX genes is believed to induce myeloproliferation, which together with additional aberrations, may lead to leukemia.
The regulation of gene expression during hematopoiesis is controlled by the cooperation of transcription factors and the dynamic architecture of chromatin. The specific epigenetic landscape influences target gene accessibility. As major executors of epigenetic regulation, chromatin-modifying enzymes mediate DNA and histone modifications responsible for the unique dynamics of chromatin observed throughout hematopoiesis. The deregulation of this process likely contributes to the malignant transformation of hematopoietic cells.
In embryogenesis, spatio-temporal expression of
HOX genes is regulated by the trithorax-group (TrxG) and polycomb-group (PcG) proteins.
PcG genes maintain
HOX gene silencing through methylation of histone 3 lysine 27 (H3K27). In contrast,
TrxG genes are responsible for maintaining previously established
HOX gene expression through trimethylation of histone 3 lysine 4 (H3K4) [
8],[
17]. A similar effect of
PcG and
TrxG genes has been proposed in the regulation of
HOX gene expression in hematopoiesis as suggested by the severe defects of hematopoietic cells that have been reported in
PcG and
TrxG knock-out models [
18],[
19]. In addition, the H3K4 demethylase
LSD1 and JmjC-domain-containing H3K27 demethylases
JMJD3 (KDM6B) and
UTX (KDM6A) have been shown to contribute to
HOX gene regulation in embryonic development [
20],[
21]. LSD1 establishes an inactive chromatin configuration by H3K4 demethylation, whereas JMJD3 and UTX activate chromatin by demethylation of H3K27. Finally, DNA methylation has been shown to participate in the establishment of
HOX gene expression patterns, further supporting the role of epigenetics in the regulation of these genes [
22].
In this paper, we sought to determine whether the pattern of leukemic HOX gene expression was primarily driven by the differentiation stage of hematopoietic cells or determined de novo during the process of malignant transformation. To approach this question, the expression patterns of the HOX genes were correlated with the molecular genetics and morphological characteristics of the leukemic cells of patients with childhood acute myeloid leukemia (AML). To further study the regulation of HOX gene expression, we also examined the relationships of chromatin modifiers and HOX genes in normal and malignant myelopoiesis.
Discussion
Several reports have demonstrated that
HOX genes are not only potent regulators of embryonic development but also play significant roles in the regulation of many processes in adult organisms, including hematopoiesis [
37]-[
39]. The overall role of
HOX clusters in addition to that of particular
HOX genes in hematopoiesis have been revealed by various knock-out and overexpression studies of human hematopoietic cells or by studies using mouse models [
40]-[
43]. However, the degree to which
HOX genes contribute to the process of leukemogenesis has not yet been elucidated.
The aberrant expression of HOX genes has been reported in the majority of leukemia patients. However, it remains unknown whether this aberrant expression represents a genuine driver of leukemogenesis or a passenger effect reflecting a differentiation block. Another possible explanation takes into consideration an impact of the molecular aberrations present in leukemic cells with further biological consequences. Here, we attempted to shed light on the expression of HOX genes in normal hematopoietic precursor cells versus their malignant counterparts with respect to their differentiation stage arrest in AML.
The crucial prerequisite for such an analysis is the appropriate identification of subpopulations of healthy BM cells representing the stages of myelopoiesis that can be matched to their respective morphological counterparts in AML. We managed to sort these subpopulations and analyzed their
HOX gene expression patterns. The expression of
HOX genes was higher at the initial stages of hematopoiesis and gradually decreased with the maturation of the hematopoietic cells, supporting the generally accepted hypothesis that
HOX genes are strong regulators of hematopoiesis (particularly at the early stages) [
37].
A comparison of matched normal and malignant hematopoietic precursor cells at the same differentiation stage demonstrated the distinct expression patterns of the
HOX genes in the leukemic cells. This indicates that the aberrant patterns of
HOX gene expression cannot be simply explained by the differentiation statuses at which the cells have been arrested. This is similar to what we previously observed in pediatric patients with ALL, who were found to exhibit differential
HOX gene expression between the subgroups and their matched normal precursors according to differentiation stage [
15]. Altogether, our results support the hypothesis that the dysregulation of
HOX genes is involved in the process of neoplastic transformation.
The analysis of childhood AML patients revealed a different expression profile of
HOX genes among the FAB subtypes and the subgroups of patients bearing unique molecular rearrangements. The most diverse subgroup of AML was AML M3, which showed the lowest levels of
HOX gene expression. This subgroup is characterized by the presence of the
PML-RARa fusion gene, which generates an aberrant retinoic acid receptor unresponsive to the physiological levels of this molecule.
RUNX1-RUNX1T1+,
CBFb-MYH11+ and
MLL- rearranged AML patients also showed unique
HOX gene expression patterns.
MLL rearrangements have been previously shown to have a determinant role on
HOX gene expression [
44]. Moreover, we revealed that AML patients bearing the
PML-RARa fusion gene had low expression levels of the
HOX genes regardless of
FLT3/ITD status. This finding is even more interesting considering that
FLT3/ITD has been shown to be associated with the upregulated expression of
HOX genes in leukemia patients [
33]. Therefore, we performed an analysis of a larger cohort of AML patients [
34] from Erasmus MC-Sophia Children’s Hospital and replicated the results drawn from our cohort of pediatric AML patients. This analysis showed that despite the overall upregulation of the
HOX genes in
FLT3/ITD+ AML patients,
HOX gene expression in
FLT3/ITD+ PML-RARa+ patients was significantly lower compared to the
FLT3/ITD+ patients without this fusion protein. Therefore, in this case, the
PML-RARa fusion gene may be superior to
FLT3/ITD with respect to its role in the process of malignant transformation. Based on these results, we suggest that AML-specific fusion oncoproteins may impact the upstream pathways that deregulate the
HOX genes, thereby acting as the major underlying factors of their characteristic expression patterns observed in leukemic cells.
Our analysis of the AML patients also showed significantly lower expression levels of
HOXA in the SR compared with the HR patients (in accordance with a previous study [
14]). These results suggest that the assessment of
HOX gene expression patterns may allow for the prediction of aggressive cases of leukemia and may therefore be taken into consideration in risk stratification. However, we suggest that this observation is a consequence of the allocation of patients with different molecular aberrations to particular AML risk groups (i.e.,
PML-RARa+ patients with the lowest
HOXA gene expression levels being assigned to the SR group and
MLL+ cases with the highest expression levels of
HOXA genes being allocated to the HR group) and not an independent prognostic factor.
Considering the profound contribution of chromatin modifiers to the embryonic regulation of
HOX genes and the essential roles of
HOX genes in hematopoiesis, the dysregulation of chromatin modifiers may deregulate the entire process of hematopoiesis and subsequently lead to malignant transformation. However, the exact roles of epigenetic modifications in the regulation of leukemic
HOX gene expression remain to be elucidated. It has recently been shown that
HOX genes possess unique chromatin regions called bivalent domains. These domains are characterized by the presence of both repressive (methylated H3K27) and activating (methylated H3K4) histone methylation marks and are found in genes poised to be activated according to cell-specific requirements [
45]. To determine the role of chromatin modifiers in the regulation of
HOX genes in normal hematopoiesis and leukemogenesis, we analyzed the expression patterns of DNA methyltransferases, histone H3K27/H3K4 demethylases, and selected
PcG and
TrxG genes in subpopulations of healthy BM cells and BM samples of patients with AML. We found an inverse correlation of histone demethylase (Modifiers 2) and
DNMT (Modifiers 3) gene expression in normal and malignant hematopoiesis. In contrast to healthy hematopoiesis, we found an interesting correlation between chromatin modifier gene expression and that of the
HOX genes in the AML samples. The most pronounced correlation was observed with the AML M3 subtype. The specific relationship of the
HOX genes with the epigenetic modifiers in this morphological subgroup could be affected by the presence of the
PML-RARa fusion gene. In particular,
HOX gene expression was positively associated with the histone H3K27 demethylases,
JMJD3 and
UTX, and inversely correlated with
DNMT3b. Notably, both
JMJD3 and
UTX have recently been suggested to play roles in hematopoiesis [
24],[
46]. Moreover,
UTX has been shown to directly bind to the
HOXB1 locus [
21],[
24]. Taken together, the results implicate chromatin modifiers in the establishment of the aberrant leukemic expression of
HOX genes in pediatric AML patients.
Although the expression of
BMI1 was not altered during hematopoiesis, a Spearman correlation analysis showed that this gene was positively correlated with
HOX gene expression in the leukemic samples. It has been reported that
BMI1 determines the proliferating abilities of the cells by inhibiting the
p16 gene.
HOXA9 was also shown to target
p16 and impair the senescence of cells [
47]. Thus, the expression levels of the histone methyltransferase
BMI1 are likely to reflect the proliferation statuses of leukemic cells without directly impacting
HOX gene expression [
48].
Interestingly, the
PML-RARa and
RUNX1-RUNX1T1 fusion oncogenes have been shown to cooperate with repressive complexes, leading to alterations in chromatin architecture.
PML-RARa causes profound changes in the epigenetic landscape, mainly by recruiting chromatin-modifying enzymes to target sequences or by the deregulation of their functions [
49],[
50]. Furthermore, recent studies have shown that the degradation of the PML-RARa oncoprotein results in dramatic changes to the landscape of histone modifications [
51]. Similarly,
RUNX1-RUNX1T1 has also been shown to recruit epigenetic modifiers to target sequences [
52]. These findings together with our data suggest that AML-specific oncoproteins regulate
HOX gene expression through epigenetic modifications. However, further studies are needed to understand the roles of epigenetic modifiers in the regulation of normal as well as leukemic
HOX gene expression and their cooperation with AML fusion oncoproteins.
In summary, we found that the expression patterns of the HOX genes in leukemic cells are not solely determined by their particular differentiation stages. Conversely, we assume that the specific molecular aberrations that are typical of AML are the major determinants of the leukemic expression patterns of the HOX genes. Our results also demonstrate the differing contributions of epigenetic modifiers to HOX gene expression in healthy and malignant hematopoiesis.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
KSK performed majority of the sample processing and analyses and wrote the manuscript. KR was responsible for the qPCR analyses, and KF performed the statistical analyses. EM defined the crucial characteristics of the sorted subpopulations of the healthy BM cells and designed all of the FACS sorting experiments. MZ was responsible for the molecular characterization of the AML patients. HD and JT revised the manuscript and provided critical intellectual feedback. JaS was responsible for the clinical management of the patients. As a senior author, JuS coordinated all of the experiments, revised the manuscript and is the principal investigator of the whole study. CMZ, MvdHE and MF contributed by providing datasets and performing the gene expression profiling analyses of the replication sets. All authors read and approved the final manuscript.