Deregulation of TWIST-1 in the CD34⁺ compartment represents a novel prognostic factor in chronic myeloid leukemia

Chronic myeloid leukemia (CML) is a model for cancer stem cell and tyrosine kinase inhibitor (TKI) resistance. If chronic phase (CP) patients are discriminated by Sokal score¹  at diagnosis, they remain heterogeneous and lack a reliable molecular prognosis marker. Imatinib has become the standard of CP-CML care,²  but some rare Philadelphia-positive (Ph⁺) stem cells may be insensitive to TKIs,^3-5  maintain detectable disease over years, and represent a reservoir in which secondary molecular events occur and initiate disease progression. The BMI-1 gene, involved in stem cell self-renewal and maintenance, has been proposed to predict CML disease progression.⁶  We hypothesized that TWIST-1, a transcription factor involved in embryogenesis, undifferentiated cell proliferation, and cell survival,^7-9  might be involved in CML. We analyzed its expression in CML cells from patients and in cell lines and identified it as a unique predictive marker for TKI resistance, especially in patients with no other detectable resistance mechanism.

After informed consent in accordance with the Declaration of Helsinki and local ethics committee bylaws (from the Délégation à la recherche clinique des Hospices Civils de Lyon, Lyon, France), samples were obtained from CML patients at diagnosis or during follow-up and from healthy allogeneic donors. CML cell lines were cultured as described.¹⁰  CD34-purified cells (StemCell Technologies) were placed in Tri-Reagent (Sigma-Aldrich). Total RNAs were reverse-transcribed, and quantitative polymerase chain reaction (PCR) was performed using sequence-specific primers (supplemental Table 1, available on the Blood Web site; see the Supplemental Materials link at the top of the online article; LightCycler480II system, Roche Diagnostics and QuantiFAST SYBR Green I-kit, QIAGEN) using a fast protocol (activation, 5 minutes at 95°C; amplification,10 seconds at 95°C, 30 seconds at 60°C; melting, 10 seconds at 95°C, 10 seconds at 50°C). Specific TWIST-small interfering RNA (siRNA; sequence coding regions 55-75) or a Scramble-negative control were transfected as duplexes (RNAiMax, Invitrogen).^7-9  Cells were counted (trypan blue staining), controlled for transfer efficiency (flow cytometry BDFACSCalibur), or analyzed for TWIST-1 silencing (RNA and protein). Western blot was performed as described¹¹  using monoclonal antibodies to ABL (BD Biosciences PharMingen), TWIST-1 (Abnova), and Ku80 (Roche Diagnostics). Statistically, the Wilcoxon Mann-Whitney test, with R Version 2.9.2 software, assessed differences between CML phases, responding and nonresponding patients or CD34⁺ and CD34⁻ fractions, and the Student paired t test for differences between cell lines groups. In addition, we performed an analysis of variance applying a bivariate analysis. Significant P values (P < .05) are given in the text, supplemental Table 2, and figures.

Approximately 30% of CP-CML patients develop resistance to TKIs, the mechanism(s) of which remain(s) obscure. Genome-wide microarray analysis of CML cell lines has shown that TWIST-1, elevated in other cancers, immature cells, and patients with drug resistance,¹²  is deregulated in a CML-resistant cell line.¹³  We analyzed TWIST-1 expression by quantitative PCR in purified immature (CD34⁺) and mature (CD34⁻) cells from patients at different disease phases and after emergence of TKI resistance. Compared with normal samples, the overall TWIST-1, BMI-1, and BCR-ABL expressions increased (Figure 1A) in CD34⁺ and CD34⁻ cells of, respectively 66-fold (P < .001) and 74-fold (P < .001) for TWIST-1 and 2.71-fold (P < .001) and 2.25-fold (P < .001) for BMI-1 as reported.⁶  Along disease progression, TWIST-1 expression significantly increased in CD34⁺ cells (11-fold and up to 133-fold in chronic and advanced phases, respectively; Figure 1B) and CD34⁻ cells (65-fold and 97-fold). Comparing TKI responder and nonresponder samples, we observed an approximately 100-fold increase in TWIST-1 expression, regardless of CD34 expression (P = .88; Figure 1C), strongly supporting its expression as a predictive factor for CML response to TKIs, unlike BMI-1 (supplemental Figures 1A, 2B). In accordance with previous data, we observed an increased expression of BCR-ABL between CD34⁺ and CD34⁻ cells (supplemental Figure 1A-B). We analyzed TWIST-1 expression in CML samples at diagnosis in a series of samples from 32 patients and observed an increased mean value at diagnosis for nonresponders (24.65-fold and 107.44-fold) compared with responders (1.28-fold and 1.32-fold), respectively, for CD34⁺ and CD34⁻ cells. Indeed, sensitive patients, before treatment (Figure 1D mean responder) or during treatment (Figure 1C responder), always presented a level of TWIST-1 similar to that of normal donors. The individual values of all patients (P1-P11) who developed resistance (European LeukemiaNet criteria¹⁴ ) later in their disease course had higher TWIST-1 expression than the mean value of responding patients (Figure 1D; supplemental Table 2), whereas only 2 had increased BMI-1 expression (supplemental Figure 1C). Subsequently, we think that TWIST-1 represents a unique and reliable predictive marker of TKI resistance in CML regardless of the resistance origin (Figure 1D; supplemental Table 2). Compared with the responder mean values, all nonresponder patient samples displayed a significant increased expression of TWIST-1 in CD34⁺ cells (100% of nonresponder patients), whereas only 8 of 11 tested patients (72% of nonresponder patients) had an increased Twist-1 expression in their CD34⁻ compartment (supplemental Table 2). Nevertheless, the difference between responder and nonresponder values of Twist-1 in CD34⁻ remained statistically significant (supplemental Table 2). This was confirmed by the detection of TWIST-1 expression within mononuclear cells from those patients plus one in accelerated phase and one in myeloid blast crisis (supplemental Table 3). In addition, sorted CD19⁺ cells from the patient 10 sample at diagnosis were negative for TWIST-1 expression despite a BCR-ABL positivity (not shown). In addition, from the 21 TKI-responsive tested patients at diagnosis, only 3 and 4 slightly expressed TWIST-1 more than 2-fold, in their CD34⁺ and CD34⁻ compartments, respectively (supplemental Table 2). Therefore, TWIST-1 measurement in CD34⁺ cells appears to be the most reliable parameter to discriminate since diagnosis of imatinib responders or nonresponders.

Our data obtained in primary CML samples confirmed TWIST-1 deregulation in the KCL22 CML cell line¹³  (resistant through unknown mechanisms). Quantitative PCR showed a 23- and 40-fold TWIST-1 expression increase (Figure 2A), with resistance in Lama-84 and KCL22, respectively, and confirmed at protein level (Figure 2B). Interestingly, TWIST-1 overexpression seems to be independent from BCR-ABL protein levels as observed in HEL, KCL22R, and Lama-84R cells¹⁰  (Figure 2B). TWIST-1 silencing by RNAi in Lama-84R and KCL22R (supplemental Figure 2) cells had no effect by itself but decreased cell numbers and viability in response to imatinib (Figure 2C-D). Conversely, overexpression of TWIST-1 in KCL22-sensitive cells (supplemental Figure 2B) significantly improved their resistance to increasing doses of imatinib (Figure 2D). Altogether, these data indicate that TWIST-1 expression is involved in imatinib resistance. Finally, we analyzed TWIST-1 expression in CD34⁻ cells during disease progression; and although it was highly expressed at diagnosis, its expression significantly decreased after complete cytogenetic remission (Figure 2E) and increased again at relapse, unlike BMI-1 (supplemental Figure 3).

In conclusion, this study demonstrates that TWIST-1 is deregulated in CML patients at diagnosis and during progression and in cells innately imatinib-resistant through a yet unknown mechanism. Our results indicate that, with BCR-ABL monitoring, TWIST-1 could represent a powerful biomarker for early detection of TKI resistance at diagnosis and during residual disease follow-up. TWIST-1 might act as a survival factor providing a selective growth advantage contributing to drug resistance, as reported in solid tumors.^12,15-17  TKI-resistant TWIST-1–overexpressing cells might emerge from leukemic Ph⁺ stem cells, but further work is necessary to evaluate the link with BCR-ABL.

The authors thank S. Joly (U590), K. Chabane (Centre Hospitalier Lyon Sud), and E. Gadolet (Hematology department, Hôpital Edouard Herriot) for technical assistance and S. Morisset for statistical analysis (Association ARCHE, Hôpital Edouard Herriot).

This work was supported in part by Inserm (V.M.-S., A.P.) and Novartis Pharma France (K.S.).

Contribution: E.C., G.H., S.J., and K.S. performed experiments and analyzed data; S.H. provided routine laboratory clinical data following international guidelines; T.V. contributed to research design and analyzed data; F.-X.M. provided cell lines and proofread the manuscript; C.D. contributed to discussions and proofread the manuscript; A.P. provided TWIST-1-vectors and proofread the manuscript; F.E.N. followed up the patients, provided primary cells, analyzed the data, and wrote the paper; and V.M.-S. performed research design, analyzed data, and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Véronique Maguer-Satta, Inserm Unité U590, Centre Léon Bérard, 28 Rue Laennec, 69373 Lyon Cedex 08, France; e-mail: maguer@lyon.fnclcc.fr; and Franck Nicolini, Hospices Civils de Lyon, Hematology department, E. Herriot Hospital, 5 place d'Arsonval, 69437, Lyon, France; e-mail: franck-emmanuel.nicolini@chu-lyon.fr.