In our new work we examine how networks determine fundamental features of blood cancer progression, with implications for patient stratification and treatment.
Cancer progression is enabled and driven by the accrual of mutations that alter the fundamental behaviours of the cell. These mutations may promote growth, prevent normal cellular function, or allow new functions necessary to support the tumour to develop. How these mutations alter cells and how they combine is an important question for understanding the origin and evolution of cancer.
Our specific interest for myeloproliferative neoplasms (MPN) diseases rose from a study that showed how MPN patients could be stratified into two clinically distinct subpopulations, determined by the order in which the TET2 and JAK2 genes were mutated (1). Patients whose JAK2 was mutated before TET2 are younger at presentation of the disease in clinics, have a higher risk of thrombosis, and respond better to Ruxolitinib, a JAK2 inhibitor drug. Predominant common myeloid progenitors at the expense of differentiated megakaryocytic and erythroid cells were specific features of TET2-first patients. This observation led us to question what underlying mechanism enabled this, and whether it was relevant to other blood cancers too.
The first question that arose was what types of gene interactions would support this bifurcation. After exploring several ideas, we found that a single gene, downstream of both mutated genes, with a positive self-feedback loop could retain the order of mutations. In Figure 1, a first mutation of one gene could render the central gene insensitive to the second mutation, and therefore the aberrant expression level of the central gene to be dictated by the first mutation. The overall phenotype of the cell for each combination of mutation orders would therefore be determined by a combination of the two known genes with a putative third gene. This mini-network, or motif was sufficient to explain the importance of ordering of mutations.
Figure 1. Both of the genes JAK2 and TET2 are mutated in myeloproliferative neoplasms, but different orders of the mutation can lead to distinct changes in the cell types that expand, drug sensitivity, and time of presentation in the clinic.
With this motif in hand, the next issue was to identify candidates for the third gene. The positive self feedback loop suggested that a key feature of the gene would be bimodal expression, representing cell populations trapped in one or other state. To explore this, we used public datasets from acute myeloid leukaemia (AML) to seek out genes with bimodal expression that were also downstream of TET2 and JAK2. As AML and MPN are closely related, and indeed MPN can evolve into AML, this dataset appeared an excellent starting place. The frequency of MPN transformation to secondary AML is related to the initial MPN disease type, and thus a better understanding of the molecular events driving the different subtypes of MPNs may help clinicians diagnose patients with higher risk of thrombosis and AML progression.
This led us to identify HOXA9 as being both downstream of JAK2 and TET2, and that had a positive self-feedback loop, with evidence of bimodal expression. We further found that in AML HOXA9 bimodal expression was associated with clinical features, such as age and WBC counts, but also patient classification into specific French- American-British (FAB) or molecular subtypes. HOXA9 already had a well-known role in AML prognosis, but our work demonstrates that this gene may act in AML as a discrete switch, whose state determines AML clinical characteristics such as classification and survival. Moreover, this gave us evidence to suggest the JAK2/TET2/HOXA9 motif could determine MPN clinical progression.
From this initial motif we then expanded it into a larger network model, connecting known cellular phenotypes to the motif and relevant downstream effectors. In addition to insights into the control of cancer cell fate through HOXA9, the MPN model recapitulates the disease symptoms using well-known hematopoietic transcription factors such as GATA1 and CEBPα but also the NOTCH pathway. Thanks to this model, we predict that patients who first had a TET2 mutation have a reduced number of erythroid cells as a result of TET2 indirect downregulation of GATA1 and KLF1, which explains the reduced number of PV diseases in TET2 first patients despite the presence of JAK2 mutation. This network model allowed us to make several other new predictions, which we tested against publicly available data and new experiments. Whilst further work is necessary to establish the roles of HOXA9 and other genes in the network to different disease characteristics our paper proposes the first model for understanding MPN evolution and offers new insights for stratification of patients.
1. A. Ortmann, D. G. Kent, J. Nangalia, Y. Silber, D. C. Wedge, J. Grinfeld, E. J. Baxter, C. E. Massie, E. Papaemmanuil, S. Menon et al., “Effect of mutation order on myeloproliferative neoplasms,” New England Journal of Medicine, vol. 372, no. 7, pp. 601–612, 2015.