Fax: (713) 745-4000
Reverse phase protein array profiling reveals distinct proteomic signatures associated with chronic myeloid leukemia progression and with chronic phase in the CD34-positive compartment
Version of Record online: 19 APR 2012
Copyright © 2012 American Cancer Society
Volume 118, Issue 21, pages 5283–5292, 1 November 2012
How to Cite
Quintás-Cardama, A., Qiu, Y. H., Post, S. M., Zhang, Y., Creighton, C. J., Cortes, J. and Kornblau, S. M. (2012), Reverse phase protein array profiling reveals distinct proteomic signatures associated with chronic myeloid leukemia progression and with chronic phase in the CD34-positive compartment. Cancer, 118: 5283–5292. doi: 10.1002/cncr.27568
- Issue online: 19 OCT 2012
- Version of Record online: 19 APR 2012
- Manuscript Accepted: 29 FEB 2012
- Manuscript Revised: 23 FEB 2012
- Manuscript Received: 28 DEC 2011
- chronic myeloid leukemia;
- protein expression;
- reverse phase protein array;
- blastic phase;
- chronic phase;
Chronic myeloid leukemia (CML) is a clonal stem cell malignancy whose pathogenesis is driven by constitutive activation of the breakpoint cluster region–v-abl Abelson murine leukemia viral oncogene homolog 1 (BCR-ABL1) kinase. Although BCR-ABL1 activation is present in all patients with CML, patients can present in 3 different phases characterized by an increasingly worse prognosis and diminished responsiveness to tyrosine kinase inhibitors: chronic phase, accelerated phase, or blastic phase. The biologic basis for progression from chronic phase to blastic phase and for regulating the homeostasis of tyrosine kinase inhibitor-resistant CML stem cells is not entirely understood.
To shed some light into these aspects of CML biology, the authors used reverse phase protein arrays probed with 112 individual monoclonal antibodies to compare protein expression patterns in 40 samples of leukemia-enriched fractions from patients with CML (25 in chronic phase, 5 in accelerated phase, and 10 in phase).
An analysis of variance (significance cutoff, P < .01) unveiled a set of proteins that were overexpressed in blastic phase, including heat-shock protein 90 (hsp90); retinoblastoma (Rb); apoptosis-inducing factor (AIF); serine/threonine-protein phosphatase 2A (PP2A); B-cell leukemia 2 (Bcl-2); X-linked inhibitor of apoptosis protein (Xiap); human homolog of Drosophila Mad (mothers against decapentaplegic) and related Caenorhabditis elegans gene Sma, family member 1 (Smad1); single-stranded DNA binding protein 2 alpha (SSBP2α); poly(adenosine diphosphate-ribose) polymerase (PARP); GRB2-associated binding protein 2 (Gab2); and tripartite motif containing 24 (Trim24). It is noteworthy that several of these proteins also were overexpressed in the CD34-positive compartment, which putatively contains the CML stem cell population.
The results from this study indicated that reverse phase protein array analysis can unveil differentially expressed proteins in advanced phase CML that can be exploited therapeutically with targeted approaches. Cancer 2012. © 2012 American Cancer Society.
Constitutively active breakpoint cluster region–v-abl Abelson murine leukemia viral oncogene homolog 1 (BCR-ABL1) kinase activity drives the pathogenesis of chronic myeloid leukemia (CML) by promoting survival and proliferation and by inhibiting apoptosis.1 BCR-ABL1 interacts with and activates a wide array of downstream protein substrates involved in various signaling pathways.1 Imatinib is a tyrosine kinase inhibitor (TKI) that targets the enzymatic activity of the BCR-ABL1 kinase and, thus, suppresses the activation of important oncogenic downstream pathways.2 Frontline therapy with the TKI imatinib rendered projected rates of complete cytogenetic response of 87% in patients with CML in chronic phase (CML-CP) who were followed for 5 years.3 At the 7-year landmark, the event-free survival rate was 81%, and the overall survival rate was 86%.4 It is noteworthy that, after 12 months of imatinib therapy, only 39% of patients achieved a major molecular response (ie, BCR-ABL1/ABL1 ≤0.1%), but most still harbored detectable transcripts.5, 6 Failure to achieve a major molecular response resulted in worse event-free and progression-free survival.
CML is typically diagnosed in chronic phase (CP); however, in the absence of effective therapy, all patients in CP inexorably will transform to blastic phase (BP), usually through an accelerated phase (AP). Even in the presence of imatinib therapy, a fraction of patients will transform to BP, which is extremely resistant to therapy that, at best, responds very briefly to TKIs.7 Limited information exists regarding predictive markers of transformation to AP or BP,8, 9 thus limiting the ability to design individual, risk-adapted therapeutic strategies for patients with CML. In recent years, analyses of large gene expression microarray databases have identified gene signatures that appear to stratify patients at high risk for transformation to advanced phase CML.10, 11 However, mRNA expression analyses obviate post-transcriptional modifications (eg, phosphorylation, ubiquitination) that play crucial roles in the function of proteins involved in signaling pathways activated by BCR-ABL1 kinase. Furthermore, the correlation between transcriptomics and protein expression is very changeable,12 suggesting that delineating protein profiles may be more relevant than investigating the level of mRNA that is produced from specific genes.
Having previously demonstrated that protein expression signatures, based on the activation state of the cell cycle, apoptosis, and signal transduction-regulating proteins, existed and were prognostic in acute myeloid leukemia and acute lymphoblastic leukemia,13-15 we extended those observations to evaluate protein expression patterns in a collection of CML samples obtained from patients who received treatment with TKIs in an attempt to identify differentially expressed genes and/or signaling pathways that predicted progression to advanced-phase CML, thus potentially providing a means to discover novel therapeutic targets and perhaps enabling the development of risk-adapted therapeutic strategies.
MATERIALS AND METHODS
We generated a reverse phase protein array (RPPA) using protein derived from the leukemia-enriched fraction from 40 primary CML samples with the objective of defining comprehensive proteomic expression patterns in CML (Table 1). Samples were collected between April 2005 and May 2008 after informed consent obtained according to the regulations of and sanctioned by the Investigational Review Board of The University of Texas M. D. Anderson Cancer Center. Of the 40 patient samples that were included in this analysis, 25 samples were in CP, 5 samples were in AP, and 10 samples were in BP. Of the 10 BP patient samples, 6 were in lymphoid BP, and 4 were in myeloid BP.
|Parameter||No. (%)||Median [Range]|
|Age, y||52 [39-80]|
|WBC, 109/L||10.7 [3.5-211]|
|Platelet, 109/L||178 [39-681]|
|Bone marrow blasts, %||3 [0-97]|
|Peripheral blood blasts, %||2 [0-82]|
|Peripheral blood basophils, %||2 [0-10]|
|BCR-ABL1/ABL1 ratio at sample collection, %||68.3 [0.07-100]|
|Clonal evolution||3 (7.5)|
|No. tested||19 (47.5)|
|No. of previous therapies at sample collection||2 [0-6]|
|Median follow-up, mo||23 [1-94]|
Generation of the Reverse Phase Protein Array and Immunostaining
All protein lysates were prepared from fresh cells on the day of collection. We included as controls 16 normal, CD34-positive bone marrow samples and 9 normal peripheral blood lymphocyte samples. Samples were printed as 5 serial 1:2 dilutions in duplicate using an Aushon 2470 Arrayer (Aushon Biosystems, Inc., Billerica, Mass). The construction of the RPPA has been previously described.16 Each CML array included a total of 6912 dots printed per slide. The slides were probed with 112 antibodies against apoptosis, cell cycle, signaling regulating proteins, integrins, and phosphatases among other functional protein groups, including 85 antibodies directed against total protein, 22 directed against phosphorylation-specific sites, and 5 directed against caspase or poly(adenosine diphosphate-ribose) polymerase (PARP) cleavage sites (Fig. 1). Each array was incubated with a specific primary antibody detected by using a biotinyl-linked catalyzed signal amplification system (DAKO, Carpenteria, Calif). The protein content of each array spot was detected using 3,3′-diaminobenzidine tetrachloride chromogen stain. The signal was scanned with a 256-shade gray scale at 600 dots per inch. Total pixel intensities were calculated with background correction using Microvigene software (version 2.0; VigeneTech Inc., Carlisle Mass). For each sample, the slope of the regression line that best fit the linear range of the dilution curve was used to determine relative protein expression. Data were analyzed using the R software package (R Foundation for Statistical Computing, Vienna, Austria) with loading control and topographic background normalization.
Differential protein expression was assessed by either the 2-sided t test (for 2-group comparisons) or a 1-way analysis of variance (for multiple group comparisons), using log-transformed data. Expression heat maps were generated using JavaTreeview (open source software) (http://sourceforge. net/projects/jtreeview/ accessed March 31, 2012).17
Protein Expression Patterns in Chronic Phase, Accelerated Phase, and Blastic Phase
We first analyzed the differences in protein expression between samples obtained from patients with CML in CP, AP, and BP. To that end, we centered proteins on the median across all samples. An analysis of variance using a minimum statistical significance cutoff of P < .01 revealed a set of 20 proteins (from a total of 112 proteins probed) that were expressed differentially across the different phases of CML (Fig. 2). Proteins like heat-shock protein 90 (hsp90); retinoblastoma (RB); apoptosis-inducing factor (AIF); serine/threonine-protein phosphatase 2A (PP2A); B-cell leukemia 2 (Bcl-2); X-linked inhibitor of apoptosis protein (XIAP); human homolog of Drosophila Mad (mothers against decapentaplegic) and related Caenorhabditis elegans gene Sma, family member 1 (Smad1); single-stranded DNA binding protein 2 alpha (SSBP2α); PARP; GRB2-associated binding protein 2 (Gab2); and tripartite motif containing 24 (TRIM24) had low expression in patients with CML-CP but progressively increased as patients progressed to BP, and samples that were obtained from patients with CML-AP exhibited intermediate levels between CP and BP. Conversely, the expression of protein kinase C delta phosphorylated at serine 664 (PKCδ.p664); v-Akt murine thymoma viral oncogene homolog (AKT) activated by phosphoinositide-dependent kinase 1 at threonine 308 (AKTpT308); actin, phosphorylated p70 ribosomal S6 serine/threonine kinase (p70S6Kp); Ras-related C3 botulinum toxin substrates 1, 2, and 3 (Rac1.2.3); phosphorylated pyruvate dehydrogenase kinase 1 (PDK1p); mitogen-activated protein kinase kinase (MEK); and cyclin dependent kinase 4 (CDK4) decreased gradually as patients progressed from CP to BP, with samples obtained from patients with CML-AP exhibiting intermediate levels. To validate these results, next, we correlated expression levels of the proteins that were expressed differentially in our RPPA platform with the expression levels of their mRNA transcripts in CML samples (Table 2). To that end, we compared our RPPA data with the publicly available gene transcription profile data set previously reported by Radich et al.10 Of our 12 RPPA features that were high in BP versus others (P < .05), 10 features also were high in BP at the mRNA level; and, of our 8 RPPA features that were low in BP, 4 features also were low at the mRNA level (Table 2).
Protein Expression in the CD34-Positive and CD34-Negative Compartments
Next, we conducted a similar analysis to investigate potential differences in protein expression between the CD34-positive (23 samples) and CD34 negative (37 samples) compartments. This is relevant because the CD34-positive compartment putatively contains the leukemia-initiating cells, also referred to as the CML stem cell population. A t test was used for each protein, and the proteins that were significantly differentially expressed (P < .01) were selected. Forty-two proteins were identified as differentially expressed in CD34-positive CML cells (Fig. 3), including up-regulation in the CD34-positive compartment of proteins involved in the integration 1/wingless (WNT)/β-catenin pathway (transcription factor 4 [TCF4], survivin), adhesion proteins (integrin-β3, focal adhesion kinase [FAK], v-src sarcoma [Schmidt-Ruppin A-2] viral oncogene homolog [SRC]), the signal transducer and activator of transcription (STAT) pathway (STAT3, STAT3 phosphorylated at tyrosine 705 [STAT3p705], Bcl-2 extra large [BclXL]), PARP, phosphorylated phosphatase and tensin homolog (pPTEN), v-myc myelocytomatosis viral oncogene (MYC), phosphorylated PKCα, mammalian target of rapamycin (mTOR), and PP2A. Conversely, several proteins were down-regulated in CD34-positive cells: mouse double-minute 2 (p53 binding protein homolog) (Mdm2), p38 mitogen-activated protein kinase (p38), mitogen-activated protein kinase kinase (MEK), AKT activated by phosphoinositide-dependent kinase 1 at threonine 308 (AKTpT308), the nuclear factor κB (NFκB) pathway (PKCδ, NFκB subunit p65 [NFκB.p65], SH domain-containing inositol-5′-phosphatase 1 [SHIP1]), and proapoptotic proteins (BH3-interacting domain death agonist [BID], Bcl-2-interacting mediator of cell death [BIM]). When the levels of these proteins were compared with their mRNA transcripts using the gene transcription profile data set reported by Radich et al,10 we observed significant overlap. Of our 21 RPPA features that had high CD34-positive counts versus CD34-negative counts, 8 features also had high CD34-positive counts (P < .05) at the mRNA level; and, of our 21 RPPA features that had low CD34-positive counts, 9 also had low mRNA levels (Table 2).
Comparisons Not Associated With Differences in Protein Expression
A series of comparisons did not render significant differences in protein expression. For instance, when we divided the patient samples according to their BCR-ABL1 mutational status, we could not identify any significant differences in protein expression between patient samples that carried unmutated BCR-ABL1 and those that carried BCR-ABL1 proteins with mutations within the kinase domain. Then, we examined potential differences between patients who carried different BCR-ABL1 kinase domain mutations to ascertain whether specific mutations are associated with differential activation of downstream signaling pathways. We did not observe any differences in protein expression between samples with the phenylalanine to leucine mutation at codon 317 (F317L), which is known to be resistant to dasatinib therapy, versus all other mutations (excluding samples that carried unmutated BCR-ABL1). Similarly, the same analysis was done comparing samples with the threonine to isoleucine mutation at codon 315 (T315I) (known as resistant to all currently clinically available TKIs) and those with other BCR-ABL1 mutations.
We used RPPA profiling to identify specific subsets of proteins whose expression is associated with CML progression. Likewise, specific proteins appear to be differentially expressed in the CML stem cell compartment. The implications of these results are multiple, but perhaps the most important is that some of the proteins differentially expressed in advanced-phase CML samples or in the CD34-positive compartment may represent therapeutic targets. This is of importance because, although single-agent TKI therapy is highly effective when given to patients in CP, it produces suboptimal responses in patients with advanced-phase CML. This is particularly true for patients in BP, in whom responses are rare and typically short-lived, which dramatically compromises the long-term outcomes in this patient population. In CP, TKI therapy induces very high rates of complete cytogenetic response, but the number of patients who achieve complete molecular remission (ie, undetectable BCR-ABL1) is limited.3 This inability of many patients to clear BCR-ABL1 transcripts below the level of detection of currently available polymerase chain reaction techniques has been attributed to the persistence of a small population of quiescent leukemia-initiating cells within the compartment of bone marrow CD34-positive cells. Although dividing BCR-ABL1-positive cells are readily cleared by TKIs, it has been demonstrated that quiescent CD34-positive/CD38-negative CML stem cells, which are estimated to account for approximately 0.5% of the CD34-positive population in the bone marrow, resist imatinib-mediated apoptosis.18-20 The use of more potent TKIs, such as dasatinib21 and bosutinib,22 has also been associated with negligible activity against this quiescent subset of cells. Eradication of BCR-ABL1-positive leukemic stem cells likely will require the use of agents aimed at non-BCR-ABL1 kinase proteins. RPPA analyses have the potential to unveil such protein targets by identifying differentially expressed proteins in the CD34-positive compartment.
We used unsorted CML samples to establish that hsp90, Rb, AIF, PP2A, Bcl-2, Xiap, Smad1, SSBP2a, PARP, Gab2, and TRIM24 were overexpressed in patients who had advanced-phase CML compared with patients who had CML-CP. The level of expression of these proteins gradually increased from CP to AP to BP, suggesting that the altered expression of such proteins is an inherent phenomenon to the process of CML transformation. The finding that TRIM24 expression increases during transformation is novel, because it was demonstrated recently that this protein is involved in the pathogenesis of human malignancies like breast and liver cancer. PP2A, hsp90, Bcl-2, and Gab2 were linked previously to the pathogenesis of CML, and it is noteworthy that the latter 2 can be targeted by small-molecule therapeutics.23-25 The high level of expression of these proteins in CML-BP suggests that PP2A and hsp90 inhibitors may be particularly efficacious in this phase of the disease as opposed to during CML-CP, as experimentally proposed.23-25 Similarly, PARP inhibitors, which currently are used for the treatment of patients with breast cancer who carry breast cancer susceptibility gene 1 (BRCA1) mutations,26, 27 may be of benefit to patients with CML-BP given the high level of PARP expression in this phase of CML.
It is noteworthy that we identified several proteins that were overexpressed in the CD34-positive compartment, which also overlapped significantly with the set of proteins that was overexpressed in advanced-phase CML. Proteins that were overexpressed in both instances included AIF, PP2A, SMAD1, SMAD10, SSBP2α, PARP, and TRIM24, suggesting that similar signaling pathways may govern CML stem cell homeostasis and CML progression. This is critically important because it suggests that a common set of targets can be used to therapeutically tackle both processes. Notable among the latter are survivin and TCF4, which are important components of the Wnt/β-catenin pathway. Primary granulocyte-macrophage progenitor cells from patients with CML-BP and imatinib-resistant CML have elevated levels of nuclear β-catenin, which increases the self-renewal activity and leukemic potential of those cells. Inhibitors of this pathway are under development and are expected to be effective at eradicating minimal residual disease and for the treatment of CML-BP.
The current analysis is hindered by several obvious limitations. First, the number of patient samples analyzed is relatively small, particularly among those obtained from patients with CML-AP and CML-BP. This limitation prevents a clear differentiation regarding protein expression patterns between AP and BP. A transcriptomic analysis has suggested that CML progression is a 2-step process with little differences between the patterns of gene expression in AP and BP.10 In our proteomic analysis, AP and BP appear to share some commonalities but also some differences in protein expression. Given the unavoidable heterogeneity of biologic specimens, the lack of a larger collection of AP and BP samples prevents our ability to draw solid conclusions regarding differences between these 2 clinically (but perhaps not biologically) different phases of the disease. This limitation notwithstanding, several proteins very clearly are expressed differentially in CP compared with their expression during more advanced phases of the disease. It is noteworthy that we observed significant overlap between the levels of proteins that were expressed differentially in the RPPA and the levels of expression of their respective mRNA transcripts.10 The limited number of proteins tested is another shortcoming of this analysis, because it provides a limited snapshot of protein expression in CML. However, the majority of proteins included in these experiments are involved in key pathways in the pathogenesis of CML. We recently expanded the number of validated antibodies in our RPPA platforms to 200, which will provide more complete information regarding protein expression levels in CML in future experiments.
Overall, our findings indicate that protein signatures differ markedly between patients in CP and those with advanced-phase disease. Most important, such differences may serve as the basis for the development of targeted approaches for the management of AP and BP CML as well as for the eradication of minimal residual disease driven by CML stem cells based on differentially expressed proteins between CP and BP samples and between CD34-positive (putatively containing CML stem cells) and CD34 negative cells. RPPA analyses in a more extensive number of patient samples using a larger panel of antibodies may better define critical nodes in the signaling network governing the homeostasis of CML cells.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.