Expression of CD25 independently predicts early treatment failure of acute myeloid leukaemia (AML)


CD25+ CD34+ CD38− leukaemic cells have been shown to be chemotherapy-resistant and initiate acute myeloid leukaemia (AML) in xenograft models, suggesting a leukaemic stem cells (LSC) biology (Saito et al, 2010). High CD25 (also known as interleukin 2 receptor alpha, IL2RA) expression (>10%) at diagnosis in young (<60 years) AML patients in retrospective analysis correlated with a significantly shorter overall survival (OS, P = 0·0005) and relapse-free survival (RFS, P = 0·005). CD25 expression was also associated with FLT3-internal tandem duplication (ITD) mutation, and double positive patients had the poorest OS and RFS (P = 0·001 and P = 0·003, respectively; Terwijn et al, 2009).

We examined the impact of CD25 expression among 45 patients with newly-diagnosed AML treated at our institution between February 2008 and May 2011 (see Table 1). CD25 was detected in 14 patients (17%) at diagnosis. When comparing CD25+ and CD25− patients, there was no statistical difference in distribution of the following characteristics: sex, age, cytogenetic risk, genetic risk group according to European LeukaemiaNet (ELN); presence of NPM1mut, type of induction therapy, or stem cell transplantation (SCT). 57% patients with CD25+ AML had FLT3-ITD (P = 0·0739; see Table 1). Twenty-eight (62%) patients underwent SCT, of which 17 (61%) were allogeneic and 11 (39%) autologous transplants. Expression of CD25 at diagnosis was a strong predictor of treatment failure (induction failure and relapse); 12 (86%) CD25+ patients experienced relapse compared with 14 (45%) CD25− patients (P = 0·0076). Interestingly, 3 (21%) CD25+ patients displayed an increase in the percentage of CD25+ blasts at first treatment failure (induction failure or first relapse). Five (16%) of the initially CD25− patients experienced recurrence with CD25+ clone. Three of these patients who had subsequent disease progression also showed further increase in percentage of CD25+ (Fig. 1H).

Table 1. Clinical and molecular characteristics of patients based on CD25 expression (unadjusted)
Age at diagnosisCD25 negativeCD25 positive
Median: 60 (29–83) years
Median: 61 (29–83) yearsMedian: 60 (36–75) years
N % N %
  1. CD25 expression was assessed in each patient by flow cytometry and immunohistochemistry, and correlated with clinical outcome. Positive CD25 expression was defined by percentage of CD25-positive cells within the blast population ≥1%. The median age was 60 years (29–83), while 15 (33%) patients were older than 65 years. Cytogenetic analysis indicated that 4 (9%) patients had good risk (core binding factor leukaemia), 27 (60%) had intermediate risk (diploid karyotype with no good or high risk features), and 13 (29%) patients had high risk (complex karyotype, abnormality of 3q26, or monosomy 5 or 7) features, 6 (13%) patients had mutated nucleophosmin (NPM1mut), but wild-type (unmutated) fms-like tyrosine kinase 3 (FLT3wt); 8 (18%) patients had wild-type NPM1 (NPM1wt), but did have FLT3− internal tandem duplication (ITD) (NPM1wt/FLT3−); and 9 (20%) patients had both NPM1mut and FLT3− (Table 1). Induction chemotherapy regimens were: high dose cytarabine (HiDAC) and high dose mitoxantrone (cytarabine at dose 3 g/m2 intravenously over 3 h daily on days 1–5; mitoxantrone 80 mg/m2 by intravenous infusion over 30 min once on day 2), or HiDAC with duanorubicin (cytarabine 3 g/m2 given over 3 hs every 12 h for total of eight doses followed by daunorubicin 60 mg/m2 IV push daily for 2 d, first dose 12 h after cytarabine completion; n = 16; or FLAG; standard 7 + 3 (cytarabine and idarubicin). The induction therapies were distributed equally between CD25+ and CD25− patients.

SexFemales 21 (46·7%)Males 24 (53·3%)
Sex (P = 0·3437)
Cytogenetics (P = 0·1906)
Induction (P = 0·1557)
3 + 726·5321·4
Type of transplant (P = 0·8881)
NPM1mut (P = 0·3671)929·0642·9
FLT3-ITD (P = 0·0739)929·0857·1
Treatment failure (= 0·0076)1445·21285·7
Induction failure (P = 0·2590)825·8642·9
Induction death (P = 0·5695)13·217·1
Death (= 0·9538)1341·9642·9
Table 2. Patient and disease characteristics evaluated in univariate (unadjusted) and multivariate (adjusted) analysis
FactorsTreatment failure (induction failure or relapse)Overall survival
HR95% confidence interval P HR95% confidence interval P
  1. NPM1, nucleophosmin; FLT3-ITD, fms-like tyrosine kinase-3 internal tandem duplication; ELN, European LeukaemiaNet; HiDAC, high dose cytarabine-based regimen.

Univariate Analysis
Sex (Male versus Female)1·050·49–2·280·900·950·38–2·330·90
Age (<60 versus 60+ years)1·520·70–3·320·290·810·33–2·000·64
Favourable versus adverse0·200·03–1·640·320·270·03–2·220·33
Intermediate versus adverse0·830·35–1·96 0·560·21–1·48 
Genetic risk group ELN
Favourable versus adverse0·090·01–0·780·050·270·06–1·360·39
Intermediate-I versus adverse1·580·63–3·96 0·840·28–2·51 
Intermediate-II versus adverse1·180·37–3·77 1·110·31–3·94 
 FLT3-ITD (mutated versus wild type)2·481·14–5·390·021·140·45–2·910·78
 NPM1 (mutated versus wild type)0·970·42–2·230·941·250·49–3·170·64
CD25 expression (positive versus negative)3·521·56–7·950·00251·070·40–2·810·90
Induction therapy 3 + 7 versus HiDAC-based2·390·89–6·430·091·730·50–5·970·38
Stem cell transplant (yes versus no)0·380·17–0·830·010·270·11–0·690·006
Type of transplant
Autologous versus no0·110·03–0·510·020·220·06–0·800·02
Allogeneic versus no0·630·28–1·41 0·320·11–0·88 
Sorafenib yes versus no2·430·95–6·210·061·430·47–4·300·53
Multivariate Analysis
CD25 (positive versus negative)2·971·28–6·890·01
 FLT3-ITD (mutated versus wild type)1·960·88–4·380·11
Figure 1.

(A) Cumulative incidence of relapse (CIR) based on CD25 expression at diagnosis, (B) overall survival (OS) based on CD25 expression at diagnosis; (C) CIR based on presence of CD25 and/or FLT3-ITD, (D) OS based on presence of CD25 and/or FLT3-ITD; (E) CIR based on CD25 expression at any time (diagnosis or relapse), (F) OS based on CD25 expression at any time (diagnosis or relapse); (G) OS based on stem cell transplantation, and (H) Increase in % of CD25+ blasts at first treatment failure (induction failure or first relapse) and second treatment failure (2+ relapse). Note: Median times to event are shown in the legend on each figure. Median times that could not be estimated are indicated as ‘Undefined’.

CD25 expression and FLT3-ITD were predictors for treatment failure in univariate analysis (all results given as Hazard Ratio [95% Confidence Interval]: CD25+: 3·52 [1·56–7·95], P = 0·0025; FLT3-ITD: 2·48 [1·14–5·39], P = 0·02). CD25 expression remained predictive in multivariate analysis (CD25+: 2·97 [1·28–6·89], P = 0·01; FLT3−: HR 1·96 [0·88–4·38], P = 0·11; Table 2). The median cumulative incidence of relapse (CIR) for all patients was 9·9 months and the median OS was not reached. The estimated 6-month CIR was significantly worse in CD25+ patients compared to CD25− patients (76% vs. 36%, P = 0·0012; Fig. 1A). This however, did not translate into a difference in OS (1-year OS: CD25+ 64% vs. CD25− 67%, P = 0·9505; Fig. 1B). The estimated median CIR for patients with both CD25+ and FLT3-ITD+ was 2·4 months compared with 4·1 months for patients with either CD25+ or FLT3-ITD+, and 19·1 months in patients with CD25− and FLT3wt (P = 0·0229; Fig. 1C). There was no difference in OS among these groups (1-year OS: 70% vs. 46 vs. 72%, P = 0·262; Fig. 1D). When comparing the outcome of patients who were CD25+ at diagnosis or at relapse with patients who remained CD25−, the estimated 6-month CIR was significantly worse in CD25+ patients compared to CD25− patients (76% vs. 28%, P = 0·001; Fig. 1E). This again did not translate into a difference in OS (1-year OS: CD25+ 55% vs. CD25− 68%, P = 0·2868; Fig. 1F). In exploratory analyses realizing the potential bias of patient selection (and timing) of SCT, patients undergoing SCT had significantly higher 1-year OS (73%) compared with patients without SCT (34%; P = 0·0034; Fig. 1G). Additionally, SCT portended superior OS in multivariate Cox regression analysis (0·27 [0·11–0·69], P = 0·0034; Table 2). These findings need to be further investigated in a controlled situation to eliminate bias.

Together with preclinical data our observations support the hypothesis that CD25 is expressed on LSCs (Saito et al, 2010). The prognostic value of CD25 expression has been established in B-cell-acute lymphoblastic leukaemia (especially in BCR-ABL1+; Paietta et al, 1997). In AML, high CD25 expression was associated with short RFS and OS in retrospective analysis of two Dutch-Belgian-Swiss Haemato-Oncology Cooperative Groupstudies (Terwijn et al, 2009) with standard induction chemotherapy and with escalated anthracycline based induction. Escalated anthracycline improves survival of DNMT3Amut, NPM1mut and MLL+ AML (Patel et al, 2012), but does not seem to improve survival of the high risk CD25+ or FLT3-ITD+ AML (Terwijn et al, 2009). The majority of our patients received induction that contained high dose cytarabine (HiDAC) (Ramanathan et al, 2010). Many CD25+ patients also carried FLT3-ITD mutation, which, in preclinical models, confers sensitivity to cytarabine, but resistance to anthracyclines. Thus, patients with FLT3-ITD positive AML may not benefit from escalated anthracycline (Pardee et al, 2011), but rather HiDAC. We considered CD25+ patients for consolidation with allogeneic or autologous SCT early in their disease course and observed that SCT appeared to abrogate the negative impact of CD25 expression on OS. This indirectly supports the hypothesis that CD25 positivity identifies chemotherapy-resistant LSCs, and effective strategies may need to include immunotherapy in the form of allogeneic SCT.

CD25 expression on AML cells may control cell-to-cell interactions, or represents a specific stage of differentiation and maturation of myeloid progenitors, or a chemoresistant state. IL2 increases survival and chemotherapy resistance in CD25+ chronic lymphocytic leukaemia in vitro (Decker et al, 2010). IL2 therapy failed to show a benefit in first remission AML in multiple randomized trials (Buyse et al, 2011). Only one small randomized clinical trial in relapsed AML patients showed a trend toward clinical benefit (Meloni et al, 1999). Recent data suggest that IL2 may promote production of T- regulatory cells and thereby mitigate chronic graft-versus host disease (cGVHD; Koreth et al, 2011), and several clinical trials evaluating the impact of IL2 on cGVHD are currently underway. If increased rates of AML recurrence are observed, it will be important to distinguish whether IL2 increased the survival of LSCs or decreased the graft-versus-leukaemia effect via effects on T regulatory cells.

In summary, our findings indicate that CD25 can serve as a valuable marker for LSCs, which are recognized to be chemotherapy-resistant. Continued examination of CD25 as a prognostic and marker of minimal residual disease is warranted, while relevant targeted therapeutic strategies should be explored.

Author contributions

JC: designed research, performed research, analysed data, and wrote the paper; HY: performed research, analysed data, and wrote the paper; MR: performed research, analysed data, and wrote the paper; GDR: performed research, wrote the paper; WVW: performed research; NF: performed research; LS: performed research; EO: performed research; JB: performed research; BB: analysed data; AK: analysed data; SH: performed research; BW: performed research; LH: performed research; AE: wrote the paper; AGR: analysed data, and wrote the paper; RN: designed the chemotherapy protocols for AML and SCT regimens from which the data were generated, analysed data and wrote the paper.

Conflicts of interests

The authors have no competing interests.