We used a degenerate RT-PCR screen and subsequent real-time quantitative RT-PCR assays to examine the expression of HOX and TALE-family genes in 34 cases of chromosomally defined AML for which outcome data were available. AMLs with favorable cytogenetic features were associated with low overall HOX gene expression whereas poor prognostic cases had high levels. Characteristically, multiple HOXA family members including HOXA3–HOXA10 were jointly overexpressed in conjunction with HOXB3, HOXB6, MEIS1 and PBX3. Higher levels of expression were also observed in the FAB subtype, AML-M1. Spearmann correlation coefficients indicated that the expression levels for many of these genes were highly inter-related. While we did not detect any significant correlations between HOX expression and complete response rates or age in this limited set of patients, there was a significant correlation between event-free survival and HOXA7 with a trend toward significance for HoxA9, HoxA4 and HoxA5. While patients with elevated HOX expression did worse, there were notable exceptions. Thus, although HOX overexpression and clinical resistance to chemotherapy often coincide, they are not inextricably linked. Our results indicate that quantitative HOX analysis has the potential to add new information to the management of patients with AML, especially where characteristic chromosomal alterations are lacking.
Homeodomain containing genes encode a set of master transcription factors that act during development to control pattern formation, differentiation and proliferation. Mammals possess at least 39 class I HOX genes grouped into four major clusters (HOXA–D) on four different chromosomes. The homeodomain, a highly conserved 61-amino acid helix–turn–helix DNA binding motif, is the common element defining this gene family.1 During development, HOX expression occurs temporally in accordance with both the position of a gene within its cluster and in a rostral–caudal manner.2 This same pattern can be recapitulated during retinoic acid-induced differentiation of embryonal carcinoma cell lines.3 HOX genes in equivalent positions of different clusters (paralogs) are more closely related than adjacent genes in the same cluster. Moreover, evidence from chimeric gene experiments indicates that, at least in some cases, paralogous HOX genes are functionally equivalent and that it is the expression level of the paralogous group as a whole which is critical for correct development.4
In cancer, deregulation of HOX gene expression and HOX alterations have been most convincingly demonstrated in leukemia. For example, various AML translocations fuse the nucleoporin domain of NUP98 to the homeobox of a major or divergent HOX protein.5,6,7 Similarly, the T cell ALL t(10;14) chromosomal translocation results in overexpression of the homeodomain containing gene, HOX11,8 and the t(1;19) translocation fuses E2A with the homeodomain and other regions of PBX1.9,10 In an experimental setting, a high proportion of mice transplanted with bone marrow cells overexpressing either Hoxa9, a10, b3 or b8 develop AML.11,12,13,14 In humans, the Trithorax or MLL/ALL-1 gene, which normally functions to maintain Hox gene expression, is a frequent target of chromosomal rearrangements. Joh et al15 demonstrated that a chimeric MLL-LTG9 protein led to the inhibition of Hoxa7, Hoxb7 and Hoxc9 expression in mouse 32Dcl3 myeloid cells. Recently, Rozovskaia et al16 found that the t(4;11) translocation was associated with increased expression of HOXA9 and MEIS1. Importantly, HOXA9 has been identified in a gene expression array-based screen as the single gene whose expression most correlated with treatment failure in AML.17
Recently, we examined HOX gene expression in human lung cancer.18 Interestingly, in both cell lines and direct tumors, the HOXA9 and B9 paralogs and HOXA10 were frequently overexpressed. In contrast, HOXA1 was often downregulated in concert with WNT7a. Moreover, both in our initial report and from more recent studies, we observed that HOX gene expression in lung tumors tends to fall into two groups; one subset expressing multiple HOX genes at high levels and the other with overall low HOX gene expression. We were intrigued by the potential parallels involving HOX expression patterns in two seemingly diverse types of cancer and by the opportunity to correlate HOX expression with chromosomal alterations. In patients with AML, the presence of a t(8;21), t(15;17) or inv(16) is widely recognized as a favorable cytogenetic subset.19 In contrast, patients with deletions of chromosome 5 or 7, abnormalities involving 3q, and complex karyotypes define a poor prognostic group.19 Similarly, alterations of the MLL gene are often associated with a poor prognosis although it appears that t(9;11) translocations may do better than other MLL rearrangements.20,21 However, for a substantial number of patients, characteristic chromosomal abnormalities are lacking. These patients and their physicians are left with considerable uncertainty regarding the need for additional therapy.
In this report, we demonstrate that favorable and unfavorable cytogenetic abnormalities are closely associated with distinct HOX expression patterns. When extended to a group of patients with either normal cytogenetics or those with chromosomal alterations of intermediate or unclear significance, the HOX expression patterns were similarly well demarcated and appear to be predictive of outcome, especially for the subset of patients with low levels of expression. Also, the levels of gene expression were clearly not random as there were highly significant correlations among various clustered HOX genes, PBX and MEIS family members.
Materials and methods
Homeobox amplification using degenerate primers
Two blocks of completely conserved amino acids, ELEKEF and KIWFQN, within the homeodomain were chosen for degenerate primers. These were ELEKEF: 5′-GCT CTA GA(A/G) (C/T)T(A/C/G/T) GA(A/G) AA(A/G) GA(A/G) TT and KIWFQN: 5′-GGA ATT C(A/G)T T(C/T)T G(A/G)A ACC A(A/G/T)A T(C/T)T T. PCR conditions consisted of 32–35 cycles of amplification (94°C, 1 min; 40°C, 1 min; 72°C, 1 min) with 4.5 mM MgCl2 and 2 μM primers. A band of ∼120 bp was gel isolated and cloned into a T vector. Individual clones were sequenced and analyzed using BLAST.
Real-time quantitative RT-PCR
Quantitative RT-PCR assays were performed using the ABI 5700 real-time system with SYBR-green fluorescence as previously described including the use of normalization to G3PDH.18 Briefly, RNAs were prepared from approximately 1–5 × 106 frozen bone marrow cells and extensively treated with DNase using RNeasy columns and protocols recommended by the manufacturer (Qiagen, Valencia, CA, USA). First-strand cDNA was prepared using 1–2 μg of total RNA, SuperScript II (Invitrogen Life Technologies, Carlsbad, CA, USA) and conditions recommended by the manufacturer. For the quantitative RT-PCR, 20 μl reactions were utilized which included 0.5 μl cDNA, 1.6 mM Mg++, 200 μM dNTPs, 0.1 μM primers and 0.1 μl AmpliTaq Gold (Applied Biosystems, Wellesley, MA, USA). A ‘master mix’ containing all components, except the cDNA template, was aliquoted to each reaction tube. The cDNA template was then added utilizing a small volume pipet to increase accuracy. Following an initial 95°C × 10 min to activate the DNA polymerase, 35 cycles of a two-stage PCR were used consisting of 95°C × 15 s and 60°C × 1 min. With this system, we initially began by performing the reactions in triplicate. However, the reproducibility is such that duplicate reactions are entirely sufficient. In the unusual case that duplicates differ by one PCR cycle, reactions are repeated. The PCR products were verified by gel analysis, were shown to consist of only a single band and in most cases were confirmed by DNA sequencing. The absence of contamination was verified by ‘no template’ controls and ‘no reverse-transcriptase’ controls were used to confirm the absence of DNA contamination.
The PCR primers were designed to work under a standard set of conditions and to produce products less than 200 bp. Primers were designed using PrimerExpress 1.0 (Applied Biosystems). The actual human gene sequences were obtained from GenBank. Primer sequences were as follows : HOXA1for 5′-ACC CCT CGG ACC ATA GGA TTA C; HOXA1rev 5′-AAG GCG CAC TGA AGT TCT GTG; HOXA3for 5′-TGC TTT GTG TTT TGT CGA GAC TC; HOXA3rev 5′-CAA CCC TAC CCC TGC CAA C; HOXA4for 5′-GCT CTG TTT GTC TGA GCG CC; HOXA4 rev 5′-AAT TGG AGG ATC GCA TCT TGG; HOXA5 for 5′-AGA TCT ACC CCT GGA TGC GC; HOXA5rev 5′-CCT TCT CCA GCT CCA GGG TC; HOXA7for 5′-ATC ACT CTA CCT CGT AAA ACC GAC AC; HOXA7rev 5′-ACA TAA TAC GAA GAA CTC ATA ATT TTG ACC; HOXA9 for 5′-GAG TGG AGC GCG CAT GAA G; HOXA9rev 5′-GGT GAC TGT CCC ACG CTT GAC; HOXA10for 5′-GAG AGC AGC AAA GCC TCG C; HoxA10rev 5′-CCA GTG TCT GGT GCT TCG TG; HOXB3 for 5′-CCG TGT AGA GAT GGC CAA CC; HOXB3rev 5′-CAA TCC CTT GGC CTT CTG G; HOXB6for 5′-AGG ACA AGA GCG TGT TCG GC; HOXB6rev 5′-GGC CCA AAG GAG GAA CTG TTG; HOXB7for 5′-CCA GCC TCA AGT TCG GTT TTC; HOXB7rev 5′-CCC GAA CCC GCT CCA TAG; HOXB9for 5′-AAA AAG CGC TGT CCC TAC ACC; HoxB9rev 5′-ACA CCA AAT ACC AGA CGC TGG; HOXC8for 5′-CAG AAC TCG TCT CCC AGC CTC; HoxC8rev 5′-GAC TTC AAT CCG GCG CTT TC; HOXC9for 5′-GCC TGC TGC CTC AGC ACA G; HOXC9rev 5′-GAA CCC TCC CAA ATC GCA AG; MEIS1for 5′-ATG TGA CAA TTT CTG CCA CCG; Meis1rev 5′-CCT GAA CGA GTA GAT GCC GTG; MEIS2for 5′-GAT GAA AGA GAC GGC AGC TCC; Meis2rev 5′-GGG TTG AGG TTG CAT CAT CG; PBX1for 5′-CCC ATC TCA GCA ACC CTT ACC; PBX1rev 5′-CCA TGG GCT GAC ACA TTG G; PBX2for 5′-CGA ACA CTC GGA CTA TCG CAG; PBX2rev 5′-GCT CCC TCA GCA GGT TCA TG; PBX3for 5′-GAG CTG GCC AAG AAA TGC AG; PBX3rev 5′-GGG CGA ATT GGT CTG GTT G; PRHfor 5′-CTC CAA CGA CCA GAC CAT CG; PRHrev 5′-CCT GTC TCT CGC TGA GCT GC.
Cytogenetic and fluorescent in situ hybridization analyses
Chromosome analysis was performed at diagnosis using in vitro blood and bone marrow cell cultures (24 and 48 h). R-banding techniques were employed and the karyotype was assigned according to ISCN (1995).44 FISH studies using a MLL-specific DNA probe (Appligene Oncor, Ventana Medical Systems, Tucson, AZ, USA) were performed in three cases with 11q23 rearrangement. In every case, FISH revealed a translocation of the MLL gene.
Degenerate RT-PCR screening
We began our analysis of expressed HOX genes using degenerate RT-PCR amplifications in a subset of patient samples and cell lines. This was done primarily to identify HOX genes expressed at high levels relative to other HOX genes and genes which appeared to vary substantially among samples in order to ensure that these loci were included in subsequent real-time quantitative assays. Low-stringency degenerate RT-PCR amplifications, followed by subcloning and DNA sequence analysis, were successfully performed with 10 patient samples and four cell lines. These results are shown in Table 1 for HOXA and HOXB loci as well as the proline-rich homeodomain gene, PRH. We found that most patient samples expressed high relative levels of HOXA9 and HOXA10, although some cases predominantly expressed HOXA5 (patient 17) or HOXB3 (patient 14). Conversely, no HOXA1 or HOXB1 clones were identified suggesting that these genes were underexpressed. The degenerate RT-PCR amplifications used primers corresponding to two blocks of highly conserved amino acids, ELEKEF and KIWFQN, within the homeodomain. Since this region in the clustered HOX loci is uninterrupted by an intron, amplification of genomic DNA can be used as a control to assess amplification bias. We previously found that over 30 loci could be amplified from genomic DNA using these primers and conditions.18 Values for selected loci are shown in Table 1. Thus, the over-representation and under-representation of the HOX genes noted above is not due to PCR amplification biases.
Interestingly, each of the 10 patient samples that generated sufficient degenerate RT-PCR products for subcloning and sequence analysis was associated with poor prognostic cytogenetic features or was subsequently found to have high overall levels of HOX gene expression (see below). Identical experiments using samples from AMLs having favorable cytogenetics resulted in absent PCR products or weak products that had a high background of non-HOX loci after DNA sequencing (not shown). These included samples with the t(8;21) and t(15;17) translocations as well as the inversion 16. The same phenomenon occurred in cell lines. For example, degenerate RT-PCR products were readily obtained from leukemic cell lines such as HEL (erythroid/multipotent AML), LAMA84 (CML blast-crisis) and RS4;11 or MV4;11 (ALL/bi-phenotypic) and subclone isolates from these cell lines showed a predominance of HOXA9 or HOXB9 (Table 1). In contrast, weak or no degenerate RT-PCR products were obtained from the AML cell lines Kasumi-1 and NB4 which contain the t(8;21) and t(15;17) translocations, respectively (not shown). These results suggested that genes from the HOXA group, particularly HOXA3–HOXA10, as well as HOXB3, HOXB6 and, less frequently, HOXB9 are among the most abundantly expressed HOX genes in poor prognostic leukemias. These results also indicated that AMLs with favorable cytogenetic features are associated with lower levels of overall HOX gene expression.
Quantitative real-time RT-PCR assays
Initially, we utilized samples from 17 AML patients for which complete cytogenetic analyses were available (Table 2A, non-consecutive patient numbers 1–21). Subsequently, to verify and expand our findings we analyzed an additional 17 patients (non-consecutive numbers 23–39). The cytogenetic features, age, initial WBC counts/blast percentages and outcome are shown in Table 2A along with the quantitative HOX results (discussed below).
Quantitative real-time RT-PCR assays were established for 19 homeodomain containing genes which included 13 clustered HOX loci, five PBX and MEIS family members and PRH/HEX, a proline-rich homeotic gene which correlates inversely with differentiation in hematopoietic cells.22,23,24 The raw data were obtained in terms of Ct values which refers to the PCR cycle number during exponential amplification at which the product (measured in real-time by SYBR green fluorescence) crosses an arbitrary threshold. To adjust for variations in the amount of RNA, the Ct values for each gene were normalized against the Ct values for the housekeeping gene, G3PDH (ie ΔCt = Ctspecific gene − CtG3PDH). While the resulting ΔCt values are experimentally convenient, they are not readily intuitive (ie they reflect exponential amplification and higher ΔCts represent lower expression). Instead, the results for a subset of the most informative HOX loci are shown in Table 2A in terms of their relative expression compared to G3PDH, ie expression = (1/2ΔCt) × 1000. In this format, a value of 100 is equivalent to 10% of the G3PDH expression level. If more than 34 PCR cycles were required to identify a product, the values are listed as ND (not detected). One technical note is that the identification of late-appearing RT-PCR products resulting from low-abundance transcripts is affected by the amount of input RNA/cDNA. Thus, some products listed as ‘ND’ might have been detectable with the use of more template. Therefore, the Ct values for G3PDH reflecting template input are indicated in Table 2A. Other loci which were tested but not shown include HOXA1, HOXB7, HOXC8, HOXC9 and PBX1 which were either absent or expressed at low levels in most samples. In contrast, PRH was abundantly expressed (not shown) but did not appear to correlate with outcome (not shown).
We first compared the HOX expression patterns for patient samples having favorable and unfavorable chromosomal alterations (Table 2A). There were three patients with favorable cytogenetics (patients 5, 18 and 21). These included the acute promyelocytic leukemia t(15;17) translocation, as well as the t(8;21) and inv(16) rearrangements affecting the core binding factor alpha and beta components, respectively. Twelve patients had poor prognostic cytogenetic features which consisted of MLL gene rearrangements excluding the t(9;11) translocation, deletions of chromosome 5 or 7, rearrangements involving 3q and complex karyotypes defined as four or more abnormalities. As can be seen, patients with favorable cytogenetics have considerably lower levels of HOX expression than do patients with unfavorable cytogenetic features based on chromosomal deletions or rearrangements. This was especially evident for the HOXA loci, HOXA3–A10. However, several samples with unfavorable cytogenetics also had substantially higher levels of HOXB3, HOXB6 and the TALE (three amino acid loop extension) family members, MEIS1, PBX2 and PBX3. In contrast, each of the three patients with a complex karyotype had low HOX expression.
There were 19 additional patients with either normal cytogenetics (n = 14) or alterations of unclear significance. These included additional 7p material (patient 35), trisomy 8 in a single metaphase (patient 27), a deletion of 11q24 (patient 07) and one patient (No. 08) with a t(9;11) translocation which has been reported to convey a less adverse prognosis than other MLL gene rearrangements.21 In addition, one patient (No. 26) had no cytogenetic results. Of these 19 patients, two (Nos 03 and 27) had low overall levels of HOX expression. Two other patients (Nos 33 and 34) had low overall HOX levels except for HOXA9 and PBX2.
In our initial studies, we observed that HOXB9 was overexpressed in three of five leukemic cell lines which also generally overexpressed other HOX genes (eg HEL, LAMA84 and K562). However, far fewer patient samples had elevated levels of HOXB9. Interestingly, the four highest levels of HOXB9 were all obtained from samples having an AML-M1 FAB subtype (patients 39, 30, 14 and 36) although, conversely, not all AML-M1 phenotypes had elevated HOXB9 expression. These same cases also tended to have the highest levels of expression for several other of the HOXA and HOXB genes as well as for MEIS1 (Table 2A). Using the Wilcoxon sum-rank test, we compared the quantitative HOX expression results in FAB subtype M1 vs non-M1 subtypes in all 34 patients. Indeed, statistically significant higher levels of expression were observed for HoxA3 (P = 0.0168), HoxA4 (P = 0.0391), HoxA5 (P = 0.0041), MEIS1 (P = 0.0105) and PBX3 (P = 0.0376) in AML-M1. However, the correlation between HOXB9 and the M1 phenotype was not statistically significant as several of the patients with AML-M1 lacked elevated expression. We also noted that several patient samples with normal cytogenetics had higher levels of multiple HOX genes than cases with poor prognostic cytogenetics. We have not examined this in more detail but these findings were observed from two sets of samples (non-consecutive numbers 1–21 and 23–39) prepared and studied at different times indicating that this is a reproducible observation.
Finally, there was good correlation between the degenerate RT-PCR data and the specific quantitative analyses. For example, the majority of degenerate RT-PCR products from patient 1 were derived from HOXA4, A9, A10 and B3. High expression of these genes was confirmed by specific RT-PCR assays. Similarly, by degenerate RT-PCR, HOXB3 transcripts comprised 48% of the total clones isolated from patient 14. By real-time quantitative RT-PCR, this same sample was found to express HOXB3 at levels which were approximately 11.7% of G3PDH expression. Conversely, no HOXA1 subclones were identified by degenerate RT-PCR and the subsequent quantitative assays confirmed it was generally expressed at low levels (not shown). Although specific PCR primers can accurately assess the relative expression of individual HOX loci among samples, individual primer pairs may amplify their targets with different efficiencies making a direct comparison between genes less precise. These degenerate RT-PCR amplifications provided a useful complementary approach and a degree of confirmation for the real-time assays.
Analysis of patient outcome
Our study utilized patient samples that had been cytogenetically characterized for the purpose of analyzing quantitative HOX expression in different types of AML. While we were able to obtain some degree of clinical follow-up, the number of samples and the degree of follow-up were too limited for an optimal analysis of response and survival correlations. Of the 34 patient samples, 26 were evaluable for outcome and 25 of these were available for initial response to therapy. Patients were excluded if they died within 1 month of admission (patients 9, 13, 28, 30, 36, 37), were not treated (patients 14, 17), or treated only with palliative intent (patient 34) as indicated in Table 2A. For each of the genes, we determined cut-off values using ROC curves25,26 to optimize sensitivity and specificity. These cut-off values are indicated in Table 2B.
With this data set and Fisher's exact test, no statistically significant differences were observed in the complete response (CR) rate for high vs low expression of HOXA9 or any of the other HOX genes. However, the overall CR rate was high in this group of patients. While 7/7 patients with low HOXA9 expression (<1.69, Table 2B) achieved a CR, so did 15/18 patients with higher levels of expression. Next, we used the Spearman correlation coefficients to examine the relationship between age and gene expression. None of the genes was significantly correlated with age. In addition, the Wilcoxon sum-rank test was applied to compare the gene expression values between patients below and above 60 years of age. No significant correlations were identified. Finally, we used the Cox regression model and log-rank test to look at event-free survival with an event defined as either no response to induction chemotherapy, relapse or death after 1 month. We found that HOXA7 values less than 0.02 were significantly correlated with an improved event-free survival (P = 0.0474). Similarly, low expression of other HOX genes was marginally significant for improved event-free survival, ie HoxA9 (P = 0.0795), HoxA4 (P = 0.0744) and HoxA5 (P = 0.0925) reflecting the limited number of patients studied.
HOX gene interactions
We used the quantitative data from the real-time RT-PCR measurements of all 34 AML samples to look for significant inter-HOX gene correlations. These results are shown in Table 3. There was a high correlation between each of the HOXA3–HOXA10 family members and each other. The same relationship existed for those members of the HOXB group that were tested. Similarly, expression levels of nearly all the HOXA and B family members examined were significantly correlated with MEIS1 and PBX3 expression, but generally not with MEIS2 and PBX2. In some cases, based on the specific loci studied, we were able to analyze the expression levels between paralogous genes. For instance, HOXA3 expression was significantly correlated with HOXB3 and, likewise, HOXA9 levels were correlated with HOXB9. However, HOXA3 was also correlated with the expression of HOXB6 and HOXB9 and similar relationships can be seen for other HOXA and B loci. In contrast, MEIS1 and MEIS2 levels were unrelated whereas PBX2 and PBX3 expression was significantly correlated. MEIS1 levels were best correlated with PBX3 although a significant relationship was also seen with PBX2. Thus, the picture that emerges from these quantitative data is one of multiple HOX gene underexpression or overexpression and also one in which MEIS1 and PBX3 are the most likely cooperating TALE-family members in adult AML.
In this report, we have been able to take advantage of technological improvements in quantitative gene expression using real-time RT-PCR to study chromosomally defined cases of AML with clinical follow-up. Strikingly, we identified substantially different patterns of HOX expression between leukemias associated with favorable vs unfavorable cytogenetic features. Importantly, these differences appear to extend to cases lacking characteristic chromosomal alterations. While the number of patients was too small to show a difference in initial complete response rates, we could demonstrate an improved event-free survival for low expression of HOXA7 and there was marginal significance for other HOXA genes. This is not surprising given the very high correlation among the expression levels of various HOXA and B loci (Table 3).
While these results must be confirmed in larger studies, they suggest that patients with low HOX expression have a favorable outcome. Since each of the good prognostic cytogenetics cases had low HOX expression and each of the poor prognostic group based on chromosomal deletions or rearrangements had high expression (Table 2), we anticipate that quantitative HOX analysis may be most useful prognostically for patients with normal karyotypes, complex karyotypes and other rearrangements of intermediate or unclear significance. While speculative, our initial results suggest that low HOX expression may be similar to good prognostic cytogenetics. In contrast, while patients with high HOX expression do less well overall, there are clear exceptions (eg patients 20 and 26). Our results are consistent with the report of Golub et al17 using gene expression microarrays, who noted that HOXA9 was the single best predictor of outcome in patients with AML. Our findings extend those observations and demonstrate that multiple genes of the HOXA and B class, especially HOXA3 to HOXA10, are jointly overexpressed in poor prognostic AMLs along with certain TALE family members. Previously, Magli et al27 reported that blocks of HOX genes were coordinately expressed in AML cell lines, which supports our findings in direct patient samples. We also found that overexpression of several HOX genes was significantly more frequent in the FAB-M1 subtype although our series lacked cases with FAB-M6 and M7 subtypes.
A more basic question concerns the mechanism of HOX gene overexpression. Using normal CD34+ subsets and degenerative RT-PCR, Sauvageau et al28 found that approximately 90% of isolated subclones were from the HOXA cluster and most were derived from HOXA4 to HOXA9 transcripts. Kawagoe et al,29 as well as Lawrence et al,30 found that HOX expression was normally downregulated in bone marrow cells as they progressed from CD34+ to CD34− and that this downregulation was lost in AMLs. Therefore it appears, at least in part, that the spectrum of HOX gene expression we observed in AML may not be qualitatively different from corresponding normal cells. Rather the difference might be either quantitative in nature or involve deregulation such that HOX expression is constitutive. For instance in the mouse, constitutive expression of HOX genes including Hoxa7, a9, a10, b3 and b8 results in acute leukemia.11,12,13,14 Similarly, recurrent chromosomal translocations in humans involving HOXA9,7 PBX110 and the homeodomain containing gene, HOX11,8 result in deregulated expression and leukemia. Our results demonstrate that there are major differences between good and poor prognostic AMLs in terms of quantitative HOX gene expression. These quantitative alterations may be extremely important. For example, Greer et al4 demonstrated that paralogous HOX genes can be functionally equivalent and that it is the expression level of the paralogous group as a whole that is critical for correct development.
One genetic mechanism associated with HOX gene overexpression involves rearrangements of MLL. Each of our patient samples which included t(4;11), t(9;11) and t(10;19;11) translocations, as well as two ALL/biphenotypic cell lines with t(4;11) translocations, had HOX gene overexpression. Rozovskaia et al16 recently reported that HOXA9 and MEIS1 were upregulated in ALL with the 4;11 translocation. In AML, Lawrence et al31 reported that both HOXA9 and MEIS1 were frequently co-expressed, which our results confirm. However, in our AML patient samples, only one of the three MLL rearrangements was associated with MEIS1 overexpression, whereas most of the patients with normal cytogenetics or other chromosomal features had MEIS1 overexpression (Table 2). While MLL rearrangements have been described in the absence of chromosomal rearrangements by standard metaphase analysis,32 the reported frequency (11%) is far too low to account for our observations. Interestingly, we identified the same dichotomy of high and low HOX expression in lung cancer18 where MLL rearrangements have not been reported. Thus, it is possible that deregulation of other components in a common pathway leads to similar HOX expression phenotypes in different malignant diseases.
Most of the clustered HOX genes that have been directly implicated by retroviral activation in murine leukemia (eg HOXA7, A9, A10, B3 and B8) were identified as overexpressed in our degenerate RT-PCR survey and subsequently confirmed by real-time quantitative assays. An exception was HOXB8 which we did not observe. The effects of constitutively expressing these genes in mouse hematopoietic cells are profound. For example, HOXA9 infected bone marrow cells develop a typical appearing AML with a high percentage of immature cells and blasts.33 HOXA9 also immortalizes and confers a proliferative advantage to primary murine myeloid cells.34 Similarly, HOXA10 overexpression causes leukemia in mice12 and, in human cord blood CD34+ progenitors, leads to increased myeloid proliferation and the appearance of blasts.35 Expression of HOXB3 in mice results in a marked proliferation of myeloid cells with tissue infiltration and splenomegaly, although the cells retain some capacity to differentiate.33 In addition to the above genes, we found that HOXA4 and HOXA5 were overexpressed. A paralogous gene in mouse, HOXB4, causes a 50-fold increase in transplantable murine stem cells36 and overexpression of human HOXA5 causes a shift in the differentiation of CD34+ cells from erythroid to myeloid especially when more primitive CD34+ subsets were infected.37
Similar results have been obtained with deregulation of TALE homeodomain genes which were initially found to be activated in murine leukemias via retroviral integrations.38,39 We observed that MEIS1 and PBX3 were frequently overexpressed, especially in conjunction with other HOX loci. PBX and HOX proteins form hetero-dimers which together recognize longer and more specific DNA target sequences,40 and at least part of the HOXA9 effects on myeloid cells require its Pbx interaction domain.34 Recently, PBX1 has been shown to bind transcriptional repressor elements (N-CoR) and co-precipitate the SIN3a repressor complex.40 Thus, transcriptional repression by HOX-PBX heterodimers is likely to be an important mechanism in human leukemogenesis. In this regard, Ferrara et al41 demonstrated that 100% of AML blasts underwent in vitro differentiation in response to combined treatment with retinoic acid and trichostatin A, an inhibitor of histone deacetylase proteins which are contained in repressor complexes. MEIS proteins regulate the nuclear accumulation of PBX and together form a trimeric complex with HOX proteins.42,43 Thus, mechanistically the overexpression of HOX, PBX and MEIS family members very likely leads to a greater effects on HOX target genes.
In conclusion, the quantitative analysis of HOX gene expression has the potential to add new information to the clinical management of patients especially where characteristic chromosomal alterations are lacking. Moreover, it seems likely that the overexpression of multiple HOX genes contributes to the pathogenesis of poor prognostic AML. While there are many remaining scientific questions to be answered, the mechanism(s) of HOX overexpression and its molecular consequences are particularly relevant.
Scott MP, Tamkun JW, Hartzell GW III . The structure and function of the homeodomain Biochim Biophys Acta 1989 989: 25–48
Graham A, Papalopulu N, Krumlauf R . The murine and Drosophila homeobox gene complexes have common features of organization and expression Cell 1989 57: 367–378
Flagiello D, Gibaud A, Dutrillaux B, Poupon MF, Malfoy B . Distinct patterns of all-trans retinoic acid dependent expression of HOXB and HOXC homeogenes in human embryonal and small-cell lung carcinoma cell lines FEBS Lett 1997 415: 263–267
Greer JM, Puetz J, Thomas KR, Capecchi MR . Maintenance of functional equivalence during paralogous Hox gene evolution Nature 2000 403: 661–665
Raza-Egilmez SZ, Jani-Sait SN, Grossi M, Higgins MJ, Shows TB, Aplan PD . NUP98-HOXD13 gene fusion in therapy-related acute myelogenous leukemia Cancer Res 1998 58: 4269–4273
Nakamura T, Yamazaki Y, Hatano Y, Miura I . NUP98 is fused to PMX1 homeobox gene in human acute myelogenous leukemia with chromosome translocation t(1;11)(q23;p15) Blood 1999 94: 741–747
Borrow J, Shearman AM, Stanton VP Jr, Becher R, Collins T, Williams AJ, Dube I, Katz F, Kwong YL, Morris C, Ohyashiki K, Toyama K, Rowley J, Housman DE . The t(7;11)(p15;p15) translocation in acute myeloid leukaemia fuses the genes for nucleoporin NUP98 and class I homeoprotein HOXA9 Nat Genet 1996 12: 159–167
Hatano M, Roberts CW, Minden M, Crist WM, Korsmeyer SJ . Deregulation of a homeobox gene, HOX11, by the t(10;14) in T cell leukemia Science 1991 253: 79–82
Nourse J, Mellentin JD, Galili N, Wilkinson J, Stanbridge E, Smith SD, Cleary ML . Chromosomal translocation t(1;19) results in synthesis of a homeobox fusion mRNA that codes for a potential chimeric transcription factor Cell 1990 60: 535–545
Kamps MP, Murre C, Sun XH, Baltimore D . A new homeobox gene contributes the DNA binding domain of the t(1;19) translocation protein in pre-B ALL Cell 1990 60: 547–555
Sauvageau G, Thorsteinsdottir U, Hough MR, Hugo P, Lawrence HJ, Largman C, Humphries RK . Overexpression of HOXB3 in hematopoietic cells causes defective lymphoid development and progressive myeloproliferation Immunity 1997 6: 13–22
Thorsteinsdottir U, Sauvageau G, Hough MR, Dragowska W, Lansdorp PM, Lawrence HJ, Largman C, Humphries RK . Overexpression of HOXA10 in murine hematopoietic cells perturbs both myeloid and lymphoid differentiation and leads to acute myeloid leukemia Mol Cell Biol 1997 17: 495–505
Kroon E, Krosl J, Thorsteinsdottir U, Baban S, Buchberg AM, Sauvageau G . Hoxa9 transforms primary bone marrow cells through specific collaboration with Meis1a but not Pbx1b EMBO J 1998 17: 3714–3725
Perkins A, Kongsuwan K, Visvader J, Adams JM, Cory S . Homeobox gene expression plus autocrine growth factor productionelicits myeloid leukemia Proc Natl Acad Sci USA 1990 87: 8398–8402
Joh T, Hosokawa Y, Suzuki R, Takahashi T, Seto M . Establishment of an inducible expression system of chimeric MLL-LTG9 protein and inhibition of Hox a7, Hox b7 and Hox c9 expression by MLL-LTG9 in 32Dcl3 cells Oncogene 1999 18: 1125–1130
Rozovskaia T, Feinstein O, Foa R, Blechman J, Nakamura T, Croce CM, Cimino G, Canaani E . Upregulation of Meis1 and HoxA9 in acute lymphocytic leukemias with the t(4:11) abnormality Oncogene 2001 20: 874–878
Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP, Coller H, Loh ML, Downing JR, Caligiuri MA, Bloomfield CD, Lander ES . Molecular classification of cancer: class discovery and class prediction by gene expression monitoring Science 1999 286: 531–537
Calvo R, West J, Franklin W, Erickson P, Bemis L, Li E, Helfrich B, Bunn P, Roche J, Brambilla E, Rosell R, Gemmill RM, Drabkin HA . Altered HOX and WNT7A expression in human lung cancer Proc Natl Acad Sci USA 2000 97: 12776–12781
Wheatley K, Burnett AK, Goldstone AH, Gray RG, Hann IM, Harrison CJ, Rees JK, Stevens RF, Walker H . A simple, robust, validated and highly predictive index for the determination of risk-directed therapy in acute myeloid leukaemia derived from the MRC AML 10 trial. United Kingdom Medical Research Council's Adult and Childhood Leukaemia Working Parties Br J Haematol 1999 107: 69–79
Thirman MJ, Gill HJ, Burnett RC, Mbangkollo D, McCabe NR, Kobayashi H, Ziemin-van der Poel S, Kaneko Y, Morgan R, Sandberg AA, Rowley JD . Rearrangement of the MLL gene in acute lymphoblastic and acute myeloid leukemias with 11q23 chromosomal translocations N Engl J Med 1993 329: 909–914
Baer MR, Stewart CC, Lawrence D, Arthur DC, Mrozek K, Strout MP, Davey FR, Schiffer CA, Bloomfield CD . Acute myeloid leukemia with 11q23 translocations: myelomonocytic immunophenotype by multiparameter flow cytometry Leukemia 1998 12: 317–325
Bedford FK, Ashworth A, Enver T, Wiedemann LM . HEX: a novel homeobox gene expressed during haematopoiesis and conserved between mouse and human Nucleic Acids Res 1993 21: 1245–1249
Jayaraman PS, Frampton J, Goodwin G . The homeodomain protein PRH influences the differentiation of haematopoietic cells Leuk Res 2000 24: 1023–1031
Manfioletti G, Gattei V, Buratti E, Rustighi A, De Iuliis A, Aldinucci D, Goodwin GH, Pinto A . Differential expression of a novel proline-rich homeobox gene (Prh) in human hematolymphopoietic cells Blood 1995 85: 1237–1245
Murphy JM, Berwick DM, Weinstein MC, Borus JF, Budman SH, Klerman GL . Performance of screening and diagnostic tests. Application of receiver operating characteristic analysis Arch Gen Psychiatry 1987 44: 550–555
Hanley JA, McNeil BJ . A method of comparing the areas under receiver operating characteristic curves derived from the same cases Radiology 1983 148: 839–843
Magli MC, Barba P, Celetti A, De Vita G, Cillo C, Boncinelli E . Coordinate regulation of HOX genes in human hematopoietic cells Proc Natl Acad Sci USA 1991 88: 6348–6352
Sauvageau G, Lansdorp PM, Eaves CJ, Hogge DE, Dragowska WH, Reid DS, Largman C, Lawrence HJ, Humphries RK . Differential expression of homeobox genes in functionally distinct CD34+ subpopulations of human bone marrow cells Proc Natl Acad Sci USA 1994 91: 12223–12227
Kawagoe H, Humphries RK, Blair A, Sutherland HJ, Hogge DE . Expression of HOX genes, HOX cofactors, and MLL in phenotypically and functionally defined subpopulations of leukemic and normal human hematopoietic cells Leukemia 1999 13: 687–698
Lawrence HJ, Sauvageau G, Ahmadi N, Lopez AR, LeBeau MM, Link M, Humphries K, Largman C . Stage- and lineage-specific expression of the HOXA10 homeobox gene in normal and leukemic hematopoietic cells Exp Hematol 1995 23: 1160–1166
Lawrence HJ, Rozenfeld S, Cruz C, Matsukuma K, Kwong A, Komuves L, Buchberg AM, Largman C . Frequent co-expression of the HOXA9 and MEIS1 homeobox genes in human myeloid leukemias Leukemia 1999 13: 1993–1999
Caligiuri MA, Strout MP, Lawrence D, Arthur DC, Baer MR, Yu F, Knuutila S, Mrozek K, Oberkircher AR, Marcucci G, de la Chapelle A, Elonen E, Block AW, Rao PN, Herzig GP, Powell BL, Ruutu T, Schiffer CA, Bloomfield CD . Rearrangement of ALL1 (MLL) in acute myeloid leukemia with normal cytogenetics Cancer Res 1998 58: 55–59
Thorsteinsdottir U, Kroon E, Jerome L, Blasi F, Sauvageau G . Defining roles for HOX and MEIS1 genes in induction of acute myeloid leukemia Mol Cell Biol 2001 21: 224–234
Schnabel CA, Jacobs Y, Cleary ML . HoxA9-mediated immortalization of myeloid progenitors requires functional interactions with TALE cofactors Pbx and Meis Oncogene 2000 19: 608–616
Buske C, Feuring-Buske M, Antonchuk J, Rosten P, Hogge DE, Eaves CJ, Humphries RK . Overexpression of HOXA10 perturbs human lymphomyelopoiesis in vitro and in vivo Blood 2001 97: 2286–2292
Sauvageau G, Thorsteinsdottir U, Eaves CJ, Lawrence HJ, Largman C, Lansdorp PM, Humphries RK . Overexpression of HOXB4 in hematopoietic cells causes the selective expansion of more primitive populations in vitro and in vivo Genes Dev 1995 9: 1753–1765
Crooks GM, Fuller J, Petersen D, Izadi P, Malik P, Pattengale PK, Kohn DB, Gasson JC . Constitutive HOXA5 expression inhibits erythropoiesis and increases myelopoiesis from human hematopoietic progenitors Blood 1999 94: 519–528
Nakamura T, Largaespada DA, Shaughnessy JD Jr, Jenkins NA, Copeland NG . Cooperative activation of Hoxa and Pbx1-related genes in murine myeloid leukaemias Nat Genet 1996 12: 149–153
Moskow JJ, Bullrich F, Huebner K, Daar IO, Buchberg AM . Meis1, a PBX1-related homeobox gene involved in myeloid leukemia in BXH-2 mice Mol Cell Biol 1995 15: 5434–5443
Saleh M, Rambaldi I, Yang XJ, Featherstone MS . Cell signaling switches HOX-PBX complexes from repressors to activators of transcription mediated by histone deacetylases and histone acetyltransferases Mol Cell Biol 2000 20: 8623–8633
Ferrara FF, Fazi F, Bianchini A, Padula F, Gelmetti V, Minucci S, Mancini M, Pelicci PG, Lo Coco F, Nervi C . Histone deacetylase-targeted treatment restores retinoic acid signaling and differentiation in acute myeloid leukemia Cancer Res 2001 61: 2–7
Shen WF, Rozenfeld S, Kwong A, Kom ves LG, Lawrence HJ, Largman C . HOXA9 forms triple complexes with PBX2 and MEIS1 in myeloid cells Mol Cell Biol 1999 19: 3051–3061
Shanmugam K, Green NC, Rambaldi I, Saragovi HU, Featherstone MS . PBX and MEIS as non-DNA-binding partners in trimeric complexes with HOX proteins Mol Cell Biol 1999 19: 7577–7588
Mitelman F (ed). ISCN (1995): An International System for Human Cytogenetic Nomenclature S Karger: Basel 1995 Reference added in proof
We gratefully acknowledge the generous support from the Lulu Frankel Foundation. DNA sequencing support was provided by the University of Colorado Cancer Center DNA Sequencing Core. We also wish to acknowledge the generous support from L'Association pour la Recherche sur le Cancer (JR, AB, FB) and La Ligue Nationale Contre le Cancer, Comité de la Vienne et de Charente-Maritime (JR, AB, FB).
About this article
Cite this article
Drabkin, H., Parsy, C., Ferguson, K. et al. Quantitative HOX expression in chromosomally defined subsets of acute myelogenous leukemia. Leukemia 16, 186–195 (2002). https://doi.org/10.1038/sj.leu.2402354
- chromosome translocation
BMC Cancer (2020)
Identification of the hub genes and pathways involved in acute myeloid leukemia using bioinformatics analysis
Investigating the role of the innate immune response in relapse or blast crisis in chronic myeloid leukemia
International Journal of Laboratory Hematology (2020)