Leukemia-initiating cells (LIC) are considered the source of therapy resistance and relapse, including in acute myeloid leukemia (AML). Resistance of LIC to standard chemotherapy is often attributed to the quiescent phenotype of LIC. Whether biological characteristics relevant for therapy resistance are stable or reversible traits of LIC is unknown.


Here, we asked whether long-term dormant cells might exist in AML patient-derived xenografts (PDX) and which characteristics they might have.


We analyzed 8 serially transplantable PDX models, originally established from AML patients of various groups. We genetically engineered AML PDX cells to express luciferase and a red fluorophore to enable in vivo imaging and isolation of very low numbers of human cells from mice. To track proliferation dynamics of individual cells, we labeled PDX with the proliferation sensitive dye CFSE. Cells were re-isolated from murine bone marrow at different time points after transplantation, and CFSE content, which is reduced in bisections by every cell division, was analyzed by flow cytometry. Long-term dormant cells were defined as cells that had undergone less than 3 cell divisions, while rapidly growing cells had divided more than 7 times during 10 days of in vivo growth.


A rare subpopulation of long-term dormant cells was detected at day 14 in all but one pediatric AML PDX samples, despite a general increase in tumor load by several orders of magnitude. Time course analysis of 3 samples for up to 29 days revealed that dormant tumor cells can still be detected at later time points. These data indicate that the majority of AML PDX samples contain a rare fraction of long-term dormant tumor cells.

As dormancy is often attributed to cancer stem cells, we tested whether LIC are enriched in the long-term dormant compartment. Dormant and cycling cells were isolated from the murine bone marrow following 10 days of in vivo growth and LIC frequency analyzed by limiting dilution transplantation assay. Interestingly, although only dormant cells display the typical stem cell characteristic of dormancy, both dormant and cycling cells exhibited similar LIC frequency in both samples tested. This indicates that not the entire LIC pool is dormant in AML, but that AML LIC can reside either in a dormant or a proliferating state.

To determine whether dormancy is a constant or reversible trait of the dormant subpopulation, we first asked whether dormant cells can re-induce a tumor with similar growth kinetics as cycling cells upon re-transplantation. Re-transplantation of low numbers of dormant or cycling cells revealed that all tumors showed identical growth behaviour, indicating that dormant cells are capable to switch into proliferation after re-transplantation. Second, we asked if proliferating cells are able to switch into dormancy. Therefore, high cell numbers of bulk cells and cycling cells from 1st recipient mice were re-labeled with CFSE, transplanted into secondary recipient mice, and CFSE distribution was analyzed after 10 or 15 days. Transplanted cycling cells and bulk cells gave rise to a similar fraction of dormant cells, indicating that cycling cells are capable to switch into dormancy after re-transplantation. These data reveal a remarkable functional plasticity of AML PDX cells in mice.

To determine whether dormant and cycling cells differ in their drug response, mice were injected with CFSE labeled PDX cells and treated from day 7 to 9 with Cytarabine and with DaunoXome on day 7. Treatment response was analyzed at day 10. Treatment reduced tumor burden by approximately one order of magnitude compared to control mice in all 4 PDX samples analyzed. Importantly, the relative proportion of dormant cells within all re-isolated AML cells increased after treatment. This data indicates that dormant cells preferentially survive chemotherapy.


In summary, our data demonstrate that AML PDX samples contain a rare subpopulation of dormant and treatment resistant cells. We show that dormancy is a reversible trait of AML PDX. As dormant and cycling cells contain similar LIC frequencies, our data provide evidence that dormancy and stemness are independent traits of AML. In conclusion, dormancy as well as the reversibility of the dormant phenotype are important biological characteristics of AML that need to be considered when designing treatment strategies that aim to eradicate drug resistant LIC.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.

Sign in via your Institution