Page 71 - Haematologica-April 2018

Gfi1b in AML and MDS
Furthermore, reduced expression of Gfi1b in murine models of human leukemia leads to a higher number of leukemic stem cells (LSCs). On a molecular level, aberrant regulation of the ROS/p38/Akt/FoXO pathway as a consequence of reduced Gfi1b level might contribute to these phenotypic changes.
Methods
Study samples
Characteristics of different patient cohorts have been described previously.19-25
Boundaries of GFI1B expression
To set boundaries for GFI1B expression levels in AML and MDS
patients, we correlated expression levels with the survival out- come of patients.
Mice
Gfi1bfl/fl and Gfi1bEGFP/WT, MxCre, NUP98/HOXD13 and Kras mice have been described previously.6,26-28 Mice were housed in specific pathogen-free conditions in the animal facility of University Hospital Essen. All mouse experiments were performed with the approval of the local ethics committee for animal use (authoriza- tion n. G1196/11).
Poly(I:C) treatment
MxCretg mice harboring the poly(I:C) inducible Cre recombinase gene under the control of the Mx1 promoter were crossed to Gfi1bfl/fl mice. To conditionally delete the Gfi1b alleles in the NUP98/HOXD13 MDS mouse model, Gfi1bfl/flMxCretg NUP98/HOXD13tg mice were injected intraperitoneally (i.p.), as shown previously.6 For Gfi1bfl/flMxCretgKras+/fl mice, two poly(I:C) injections were sufficient to activate the Kras oncogene and delete the Gfi1b alleles. As a control, Gfi1bfl/fl or Gfi1bwt/wt mice not carrying the MxCretg were injected with poly(I:C). Three weeks after trans- plantation of MLL-AF9-transduced lineage negative (Lin-) BM cells from Gfi1bfl/flMxCretg or Gfi1bfl/flMxCrewt mice, primary recipient mice were injected with poly(I:C) 4 times every second day.
Isolation, retroviral transduction, and transplantation of murine hematopoietic progenitor cells
Mouse leukemia was induced by transplanting Lin- BM cells that were retrovirally transduced with the MLL-AF9 oncofusion gene as well as the GFP-encoding gene, as previously described.4,27 For the limiting dilution assay, different numbers of leukemic cells were retransplanted into sublethally irradiated (3 Gy) secondary recipient mice (3-4 mice/group). The frequency of functional LSCs was determined using ELDA software.29
ChIP and ChIP–Seq analyses
Chromatin Immunoprecipitation (ChIP) and ChIP-Seq analyses were performed as previously described.4,27 Data are available from: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE88934
Online Supplementary Appendix
Details on the experimental procedures and figures are available in the Online Supplementary Appendix.
Results
Low level of GFI1B are indicative of an inferior prognosis of MDS and AML patients
To obtain a first insight into the role of GFI1B in AML prognosis, we analyzed two well-annotated published
data sets.19-21,25 In these sets, CD34+ leukemic cells and CD34+ control HSCs were used. CD34+ leukemic cells rep- resent a fraction in which LSCs are enriched, whereas CD34+ cells from healthy donors represent a fraction of cells in which HSCs are enriched.21,30 GFI1B showed lower expression in CD34+ AML blasts compared to CD34+ con- trol HSCs (Figure 1A). MDS can progress to AML, and therefore, we wanted to elucidate how GFI1B expression changes during the progression of MDS to AML. Again, GFI1B showed a lower expression in AML blasts com- pared to GFI1B expression in CD34+ cells from the BM of patients with MDS (Figure 1B).
We also analyzed an independent data set, which pro- vided whole genome expression data for LSCs in different types of AML as well as different human hematopoietic progenitor cells.20,25 GFI1B showed a lower expression in human LSCs of different AML subtypes compared to its expression in normal human myeloid progenitors (GMPs) or HSCs (Figure 1C).20 GMPs and HSCs are two fractions from which LSCs arise in mice and humans.31
We analyzed whether GFI1B level might also be inform- ative regarding the prognosis of MDS and AML patients. Based on available expression data of GFI1B and the asso- ciated survival data, we could distinguish two distinct populations with regard to GFI1B expression (Figure 2A). A low level of GFI1B (see Methods section and Online Supplementary Appendix for details) in leukemic blast cells was associated with inferior outcome with regard to over- all survival (OS) of all AML patients (Figure 2B) as well as OS and event-free survival (EFS) in the group of patients with no overt cytogenetic aberrations (Figure 2C and D). We also performed a multivariate analysis, including addi- tional factors such as age, sex and cytogenetic status, as well as mutational status of certain genes. There was a tendency for a very low GFI1B level to be an independent prognostic marker (P=0.12), but this did not reach a level of significance (data not shown). Low GFI1B expression might be associated with an inferior prognosis, but other confounding factors contribute to this association. Finally, we examined whether low GFI1B expression (the lowest 5% compared with the highest 20% of expression levels) was associated with a certain gene expression signature to obtain a first insight into how GFI1B might influence prognosis. We performed Signaling Pathway Enrichment using Experimental Datasets (SPEED) analysis (see Online Supplementary Appendix) on two separate studies,21,22 for which expression data of the full length GFI1B and associ- ated clinical data were available. Low level of GFI1B expression was associated with a reactive oxygen species (ROS)-mediated signature pathway as well as activation of mitogen-activated protein kinase (MAPK), JAK, TGFB and TLR signaling pathways (Figure 2E).
We also examined whether GFI1B expression level influences survival and disease progression from MDS to AML using a separate set of data.23,27 Again, we could dis- tinguish two different populations with regard to GFI1B expression (low and high) (Figure 2F): low expression of GFI1B correlated with poor EFS (Figure 2G).
Anguita et al.32 observed a positive correlation between the expression of a mutated form of GFI1B, which acts in a dominant-negative manner, and the expression of MLLT3 and a negative correlation with regard to SPI1. In addition, Chowdhury et al. described a negative correla- tion between GFI1B expression and MEIS1.33 In our patient cohorts, we also found an inverse correlation
haematologica | 2018; 103(4)
615