Background
Therapy with anthracycline and cytarabine results in modulation of a wide range of proteins including massive p53 protein activation followed by cell cycle arrest and apoptosis [
1‐
3]. The p53 transcription factor transactivates a wide range of pro-apoptotic genes [
4,
5] involved in cancer cell elimination, and an intact p53 gene seems essential for therapeutic response in AML [
6]. Most of the molecular mechanisms behind chemotherapy are elucidated in experimental systems and do not reflect tissue responses and the complex cell-cell interactions that are present
in vivo [
7]. As increasing evidence is proposing tumor-host mechanisms as important for effective chemotherapy [
8], there is an immediate need to investigate these issues
in vivo in human cancer.
Clinical response to chemotherapy and karyotype analysis of AML cells provide prognostic information about risk for relapse [
9]. Gene expression analysis may provide important prognostic information in the 50% of patients with standard risk for relapse due to normal karyotype [
10,
11]. Recent studies of mutations or signaling response in AML have also indicated potential for risk stratification [
12‐
14]. All these studies are based on bulk cell analysis, and propose that analysis of patient cells under DNA damaging therapy may provide biological important information about the therapy response. Common for most previous studies of chemotherapy induced gene expression is the time of sample collection at 24 h and later after start of chemotherapy [
15,
16]. We hypothesize that earlier sampling and analysis of gene expression could provide us with information about therapy responses and early resistance mechanisms against intensive chemotherapy.
In the present work we used high-density oligonucleotide microarrays to monitor therapy-induced changes of gene expression of AML blasts in seven de novo AML patients before, at 2–4 h and at 18–24 h after start of intensive chemotherapy infusion. There was no detectable decline in viability in the sampled cells. Early gene expression was dominated by p53-associated genes, while the later gene expression was dominated by genes involved in cytarabine detoxification, chemoresistance and cell-cell interactions.
Discussion
Differentially overexpressed genes at 2–4 h included p53-induced genes related to oxidative stress, cell cycle arrest, DNA repair, autophagy and apoptosis (Figure
1C, Additional File
2). Although moderate, the early upregulation of the p53 target gene DRAM (damage-regulated autophagy modulator), encoding a lysosomal protein, stands out as special [
33,
34]. It has been reported that autophagy may delay the DNA damage response and apoptotic death in breast cancer cells [
35] and this duality indicates a highly complex regulation of cell fate post therapy.
In addition to the p53-induced genes with putative anti- and pro-survival function (Additional File
2), several novel features were observed. Upregulation of cytidine deaminases (cyclic amidines) represents a significant mechanism for inactivation of cytarabine and thereby confers therapy resistance in AML (Figure
3 and Additional File
3) [
36]. Another class of genes involved in cisplatin resistance in solid cancers is the myotubularin family, some of them encoding proteins with dual phosphatase activity [
30,
37]. Several genes indicate a tumor-host modulation early during chemotherapy. This includes the upregulation of chemokine receptors like CXCR4, whose expression levels have been proposed of major prognostic impact in acute myeloid leukemia [
31].
The BCL2/BAX ratio as well as BCL2/BBC3 (PUMA) ratio decreased at early time points in all samples analyzed (Figure
2A, B). Interestingly, the baseline ratios were particularly high for two of the patients with poor outcome as has also been reported for other patients with adverse prognosis [
38], while the two patients with the lowest baseline ratios responded best to first course induction therapy (Figure
2A and Additional File
1). Induction of BAX was accompanied by increase of BBC3 mRNA, a proposed link between p53 and BAX [
39], which in turn connects TRAIL receptors with mitochondrial apoptosis [
40]. The pro-apoptotic genes, along with p53-inducible death domain (PIDD/LRDD) [
41] were expressed at 18–24 h, but was not reflected in cleaved procaspase-3 as verified by Western blotting (Figure
1A).
Studies by others have suggested that apoptotic AML blasts will not be detected in peripheral blood samples during chemotherapy [
27,
28]. Dedicated phagocytes or neighboring cells are presumably clearing these apoptotic AML cells from circulation through receptors and adaptor molecules that can link apoptotic cells to phagocytes (reviewed in [
42,
43]). Our patients experienced no clinical symptoms of tumor lysis, reflecting an intact absorbance of apoptotic cells in the patient undergoing chemotherapy. This observation was consistent with absence of apoptosis by annexin V/propidium iodide analysis or procaspase-3 cleavage (Figure
1). Routinely, nuclear morphology was examined after density gradient centrifugation of mononuclear cells, and nuclei were not observed fragmented or condensed (data not shown). Our previous reports of pre-apoptotic BAX induction in vitro [
44], and the lack of cleaved caspase substrate proteins in AML patient samples support the conclusion that the early phase after chemotherapy represents a physiological window to examine pre-apoptotic gene modulation
in vivo. We observed no induction of genes that are involved in the classical clearance of apoptotic cells in the circulating AML blasts [
42,
43] except for an induction of the extracellular molecule MFGE8 (milk fat globule-EGF factor 8 protein) (data not shown). MFGE8 is like annexin V capable of binding to phosphatidylserine (PS) and facilitates engulfment by bridging PS on the apoptotic cell with macrophages [
45]. Another indication of apoptotic cell clearance was a 9-fold increase in mRNA expression of GalNAc4S-6ST (N-acetylgalactosamine 4-sulfate 6-O-sulfotransferase), and a high level of this molecule has been detected on apoptotic human peripheral blood lymphocytes [
43]. Furthermore, several receptors that are expressed on monocytic/macrophage lineage cells were upregulated, probably related to chemotherapy induced differentiation of the leukemic cells [
46]. We hypothesize that lack of classical "eat-me" signals of the pre-apoptotic cells in this study is caused by the nature of chemotherapy induced cell death and cell phenotype.
We observed increased expression of T-cell receptor complex components after 18–24 h of therapy (Figure
3). AML cell number in peripheral blood was rapidly declining after chemotherapy and partly followed by a decrease in lymphocytes, not sufficient to explain the increase in T-cell receptor related genes. The chemotherapy induced expression of T-cell receptor complex genes could reflect aberrant gene expression of leukemic blasts, mirroring the flexible pattern of gene expression observed in pluripotent hematopoietic stem cells [
47].
Conclusion
In conclusion, this study of gene expression of AML blasts
in vivo following start of induction chemotherapy confirms the vivid response to therapeutic DNA damage observed in experimental systems. The most prominent genes that were upregulated immediately after chemotherapy are pivotal determinants of apoptosis regulation
in vitro (Additional File
4). More striking is the upregulation of genes potentially involved in interaction between AML blasts and the host microenvironment, supporting the hypothesis that the host response in chemotherapy is crucial for persistent remission [
48]. The observations presented here provide us with a more nuanced picture of leukemic cell demise after intensive chemotherapy
in vivo, and motivate for an expanded patient study to search for possible therapy response biomarkers.
Acknowledgements
Hua My Hoang is acknowledged for excellent technical assistance with nucleic acid labeling and hybridization, and Beth Johannessen is acknowledged for excellent work with RNA purification. The work was funded by the Haukeland University Hospital Research Fund, The Norwegian Cancer Society (HS01-2006-0468), The Research Council of Norway (101538 and 160320/V50), the European Commission 6th Framework Program Contract 504743 and The Research Council's program in Functional Genomics (FUGE).
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AMØ designed the study, performed experiments, analyzed data, and wrote the manuscript. NA performed experiments, analyzed data, wrote the manuscript. THB and LS analyzed data and reviewed the manuscript. IJ designed the bioinformatics study and reviewed the manuscript. ØB provided biological material, designed the study and reviewed the manuscript. KHK designed the study and wrote the manuscript. BTG designed the study, provided biological samples, analyzed data, and wrote the manuscript.