In this pilot study, we analyzed gene expression changes over the first two days of standard induction chemotherapy in AML patients. We could show induction of relevant pathways related to immune response, cell proliferation and apoptosis, but also including genes of importance for drug pharmacokinetics. Some of these changes were detectable already after the first day of treatment. Overall, the NF-κB pathway was significantly activated after two days of treatment compared to baseline. This was also reflected by expression alterations in the iNOS pathway, which is directly linked to NF-κB signaling. The canonical NF-κB pathway is normally activated by pro-inflammatory cytokines, Toll-like receptor binding or T-cell receptor activation, subsequently leading to proteasomal degradation of the NF-κB inhibitor IκB. The release from IκB allows NF-κB to relocate from the cytoplasm to the nucleus where it regulates a wide variety of genes. In oncogenesis, this includes the transcription of genes promoting cell proliferation and survival as well as angiogenesis, metastasis, and suppression of the tumorsuppressors p53 and FOXO3a (21). In AML, a growing body of evidence has demonstrated dysregulated NF-κB (22) and opened up the possibility of targeting this pathway; here, proteasome inhibitors are constituting a promising group of drugs with broad application to many types of cancers. In addition, in vitro experiments have shown that treatment with cytarabine induces NF-κB expression in the HL-60 cell line (23). Our results in contrast show relative up-regulation of IκB and correspondingly down-regulation of NF-κB between baseline and Day 2. This is indicative of an overall positive effect of the chemotherapy treatment, inhibiting cell proliferation and survival; however, we cannot determine whether these transcriptional changes are present in all cell populations (malignant as well as healthy) or to which extent these relative changes are reflecting a decreasing proportion of malignant cells (with assumed dysregulated NF-κB signaling).
When analyzing the patients based on their first response evaluation (bone marrow analysis at day 15 or 28), we were able to identify pathways that were significantly affected in one group but not the other. These differences could potentially be indicative of differences in tumor cell sensitivity to drug treatment. As such, the pathways Purine de novo biosynthesis II, tRNA charging, and Cyclins and cell cycle regulation were inhibited in patients reaching CR already after the first course of induction therapy, which might demonstrate reduced tumor cell proliferation in this group. These pathways were not significantly affected in the group of patients that did not achieve complete remission; however, two patients did so after additional treatment. In contrast to the CR group, nCR patients displayed significant activation of pathways associated with cell cycle progression, anti-apoptotic signaling and malignant transformation. Persistent activation of Stat3, as well as other Stat proteins, has been implicated as an oncogene promoting malignant transformation in various experimental settings (24, 25). Stat3 target genes include Bcl-X and Bcl-2, both encoding anti-apoptotic proteins, and cooperation with c-Jun constitute yet another mechanism of interference with tumor cell apoptosis. In addition, the targets Cyclin D1 and Myc promote proliferation; in solid tumors an influence of Stat3 on VEGF expression could also further promote tumor growth and metastasis/tissue invasion through increased angiogenesis, a process of lower relevance for blood cancers. Many other target genes are likely involved in the oncogenic potential of Stat3, and activation of this protein does not necessarily lead to malignant transformation. However, suppression of Stat signaling has demonstrated tumor growth arrest, induction of apoptosis in cancer cells and inhibition of malignant transformation, indicating that it has potential as a future drug target (24, 26–28). Our results suggest that patients displaying rapid activation of the Stat3 pathway could be harboring more resistant or highly proliferative leukemic clones. Interestingly, another pathway identified to be significantly activated in patients with poor initial treatment response was Salvage pathways of pyrimidine ribonucleotides. This pathway represents an important system for the maintenance of ribonucleotide pools in all proliferating cells, but also have implications for drug response (29, 30).Notably, the induction of the enzyme cytidine deaminase (CDA) could result in a more treatment-resistant phenotype in these patients, as the pyrimidine analog Ara-C is subject to degradation into Ara-U by CDA. CDA induction as a mechanism of resistance to Ara-C has been demonstrated in vitro (31, 32), and our results support also an in vivo relevance.
Some considerations of this study need to be addressed. Even though our study population is small and the influence of interindividual variation may occlude common changes in gene expression related to the treatment, we are partly compensating for this by comparing the different time point data to that from the baseline sample for each patient. However, the bulk RNA sequencing data represent the gene expression in both healthy and cancerous blood cells, and the leukemic blast proportion in blood samples from AML patients varies. Throughout the treatment, blood leukocyte counts are rapidly decreasing in many patients, and in our approach, we are not discriminating between cancer cells and healthy cells being eradicated. Changes in the relative abundance of different cell populations are likely to be reflected in the expression data. As an example, we cannot determine whether a decreased gene expression at day two compared to baseline means that the expression has been inhibited, or if it means that a specific cell population responsible for the baseline expression was eradicated, leaving only cell populations with another pattern. In addition, our patient samples were collected over a long period, and variations in treatment protocols and sampling procedures may have influenced the result. Initial principal component analysis indicated that the data tended to partly cluster depending on inclusion period (defined as before or after the introduction of national treatment guidelines, data not shown) although there was an overlap between these clusters.
In conclusion, we show that time-course analysis of gene expression represents a feasible approach to identify relevant pathways affected by standard induction chemotherapy in AML patients. This poses as a potential method for elucidating new drug targets and biomarkers for categorizing disease aggressiveness and evaluating treatment response. However, more studies on larger cohorts are warranted to elucidate the transcriptional basis for drug response. As single-cell RNA sequencing methodologies are developed and becoming more common, this would likely represent an option to further enhance the resolution also in time course studies like this one, although single-cell approaches are associated with other practical challenges in both sample management and analysis compared to conventional bulk RNA sequencing. Preferably, future studies would include more time points as well as follow-up analysis in the event of relapse.