Clinical proteomics of myeloid leukemia

Myeloid leukemias are a heterogeneous group of diseases originating from bone marrow myeloid progenitor cells. Patients with myeloid leukemias can achieve long-term survival through targeted therapy, cure after intensive chemotherapy or short-term survival because of highly chemoresistant disease. Therefore, despite the development of advanced molecular diagnostics, there is an unmet need for efficient therapy that reflects the advanced diagnostics. Although the molecular design of therapeutic agents is aimed at interacting with specific proteins identified through molecular diagnostics, the majority of therapeutic agents act on multiple protein targets. Ongoing studies on the leukemic cell proteome will probably identify a large number of new biomarkers, and the prediction of response to therapy through these markers is an interesting avenue for future personalized medicine. Mass spectrometric protein detection is a fundamental technique in clinical proteomics, and selected tools are presented, including stable isotope labeling with amino acids in cell culture (SILAC), isobaric tags for relative and absolute quantification (iTRAQ) and multiple reaction monitoring (MRM), as well as single cell determination. We suggest that protein analysis will play not only a supplementary, but also a prominent role in future molecular diagnostics, and we outline how accurate knowledge of the molecular therapeutic targets can be used to monitor therapy response.

. Myeloid leukemia and normal hematopoiesis. Acute myeloid leukemia (AML; red box) is a hematological disease characterized by a block in differentiation and promotion of proliferation or survival. Recurrent genetic abnormalities include t(8;21)(q22;q22), inv(16)(p13.1q22), t(16;16)(p13.1q22), t(15;17)(q22;q21), and t(9;11)(p22;q23). Chronic myeloid leukemia (CML; green box) is characterized as a stem cell disease with hyperplastic myeloid cells, including both immature and mature myeloid cells. The disease is defined by occurrence of the BCR-ABL fusion gene in the Philadelphia chromosome. Juvenile myelomonocytic leukemia (JCMML) and chronic myelomonocytic leukemia (CMML) (blue box) are hematological diseases with features of a myeloproliferative neoplasm and a myelodysplastic syndrome. Characteristics are peripheral blood monocytosis >1 × 10 9 /l, no Philadelphia chromosome or BCR-ABL fusion gene, no rearrangement of platelet-derived growth factor receptor alpha polypeptide, or platelet-derived growth factor receptor beta polypeptide, and >20% blasts in the blood and bone marrow. The figure was kindly provided by Dr Line Wergeland, University of Bergen. of a cell, and in a specific type of cell, organ, tissue, or extra cellular fluid. Therefore, proteomics should be the ideal tool for the prediction of response to targeted therapy, as well as for monitoring targeted therapy. In this review we will focus on proteomic analysis of the leukemic cells in CML, juvenile chronic myelomono cytic leukemia (JCMML), adult chronic myelomonocytic leukemia (CMML), and AML (Figure 1), and illustrate how proteomics may create new diagnostics and mole cular classifications for the disease. Current proteomic studies are clearly technology driven and research is influenced by the recent technological advances of the field. We will discuss selected methodologies that are particularly powerful for biomarker screening or the analysis of limited clinical material. The methodological pipeline to identify promising biomarkers for disease and therapy response will be discussed, as well as how recognized and verified biomarkers may be transferred into an assay format that fits the routine laboratory ( Figure 2).

Leukemia derived from the myeloid lineage of hematopoietic stem cells
Myeloid leukemia includes some of the most well understood malignancies in humans, but also rare and less definable diseases. CML is characterized by a massive clonal expansion of mature granulocytes and precursors. The disease progresses through three distinct phases chronic phase, accelerated phase, and blast crisis during which the leukemic clone progressively loses its ability to differentiate. CML has been considered to be a unique model to understand the molecular mechanisms under lying the onset and progression of a leukemic process since it was the first recognized form of cancer to have a strong association with a recurrent chromosomal abnor mality. This abnormality is a reciprocal translocation between the long arms of chromosomes 9 and 22 (t(9;22)), which generates the socalled Philadelphia chromo some [7] (Table 1). This abnormality was later discovered to be a specific molecular defect, a hybrid BCR-ABL gene, coding for a tyrosinespecific protein kinase. The intro duc tion into clinical practice of the firstgeneration TKIs for example, imatinib represented a major breakthrough in the era of molecular targeted therapy [8,9].
In contrast to the relative homogeneity of CML, AML is a clinically and genetically heterogeneous disease that results from the transformation and proliferation of immature myeloid cells that suppress normal bone marrow function [1012]. Untreated patients have a median survival of only 2 to 4 months, and intensive chemo therapy, eventually in combination with allogeneic hematopoietic stem cell transplantation, is the only possibility for cure [10]. The first subclassification in AML was based on leukemic cell morphology and histo chemistry [10], and was named the FrenchAmerican British system [13], but this classification has proved to be of limited predictive value in determining prognosis and guidance for therapy. Acute promyelocyte leukemia (APL) is an important exception, since this morphological AML variant comprises a balanced reciprocal transloca tion between chromosomes 15 and 17 (t(15;17)) [14] ( Table 1). Patients with APL have a particularly good prog nosis after the introduction of vitamin A in the treatment [14] since the chimeric retinoid receptor in t(15;17) makes the leukemic cells sensitive for undergoing maturation and early cell death when treated with

Figure 2. Workflow for identification, verification and application of biomarkers in clinical diagnostics.
Stable isotope labeling with amino acids in cell culture (SILAC), difference gel electrophoresis (DIGE) and isobaric tags for relative and absolute quantification (iTRAQ) are all powerful tools for finding differences in protein production in separate samples. A marker is added to the samples during either the experiment or the preparation for analysis; the samples are then analyzed together and in the resulting data can be told apart on account of the different markers.
The aim is to find proteins that significantly differ in expression between the samples. For further validation, reverse-phase protein array (RPPA), and especially multiple reaction monitoring (MRM), are highly sensitive methods that can detect subtle differences in production of proteins identified as potential biomarkers. RPPA is an antibody-based assay that detects and quantifies protein production. MRM allows detection and absolute quantification of protein based on internal standard peptides. A suitable peptide, fulfilling the criteria to enable optimal analysis, is chosen from within the target protein and then produced with a heavy isotope amino acid incorporated. This synthetic peptide is added in known amounts to the sample. In the triple quadruple instrument (QQQ), the peptides of interest are selected (Q1), fragmented (Q2) and the resulting target peptide ions selected (Q3) for detection. As the amount of standard peptide added to the sample is known, peak comparison allows calculation of the amount of the target protein present in the sample. To apply identified and validated biomarkers in clinical diagnostics, the analytical method must be highly reproducible. Flow cytometry is a well-established method of analyzing hematological samples. With the application of mass spectrometry detection after flow cytometry selection (ICPTOF-MS), problems with multiplexing are overcome, and this method enables detection of up to 20 biomarker proteins. The nanofluidic proteomic assay (NIA) method allows quantitative detection of protein production in very limited material. The proteins are separated according to isoelectric point inside capillary glass tubes before immobilization and antibody detection.

Potential biomarkers
Validated biomarkers Nevertheless, approximately 50% of AML patients lack cytogenetic aberrations; because of this, gene expression profiles, DNA methylation patterns, micro RNA expression, protein production, chemokine production, and signal transduction responses have been used in subclassification [11,12,15,16]. Consequently, the treatment principle of 'one size fits all' has been challenged considerably over the past decades, even if mole cular classification has been translated into highly individualized chemotherapy in clinical trials [11].
An interesting third group of myeloid leukemia, which is difficult to fit into either the chronic or acute categories, is CMML and its pediatric variant JCMML [17]. By definition, CMML excludes the presence of the Philadelphia chromosome, but as many as 30% to 40% of cases show different cytogenetic abnormalities. In most cases of CMML, the critical genetic lesions remain un identi fied. However, the protooncogene family RAS seems to be of special interest. NRAS and KRAS are highly mutated in CMML, and recent research indicates that RAS is particularly important in the development and progression of the disease [18,19].

Proteomic technologies for analysis of leukemia
The identification of proteins in cancer is predominantly performed by mass spectrometry (MS) analysis of fractionated proteins, or indirectly through probing with wellcharacterized antibodies. The strength of antibody based techniques lies in their signal amplification steps, which allow the detection of protein concentrations in the femtomolar range. The combination of antibodies and isoelectric protein focusing in a capillary tube format [20,21] allows the detection of specified proteins and their modifications with sensitivity down to 25 cells, depending on protein abundance. Certain MS methods, such as multiple reaction monitoring (MRM), are able to detect and quantify single proteins in the attomolar range [22,23], with a dynamic range extending over three orders of magnitude [24,25]. However, the fact that leukocytes include multiple cellular subsets requires a different technique able to distinguish between various cell types. Flow cytometric determination of blood cells using anti bodies against extra and intracellular targets provides the user with information about such cellular subsets, and represents a fundamental test in leukemia diag nostics. How can we extend the repertoire of antibodies recognizing relevant proteins in leukemia diagnostics? MS techniques such as stable isotope labeling with amino acids in cell culture (SILAC) can be used to identify new protein markers that can be developed for platforms such as flow cytometry.

Stable isotope labeling with amino acids in cell culture
SILAC is a technique based on MS that has been developed to detect quantitative differences in protein levels between two or more samples [26] (Figure 2). Different cell populations are grown in medium containing amino acids labeled with stable light or heavy isotopes. After a certain number of doublings, the isotopelabeled amino acids are metabolically incor porated into every protein of the cell. Equal amounts of protein from the two different cell populations are combined, followed by standard procedures for sample preparation for MS analysis. The metabolic incorporation of the isotopelabeled amino acids leads to a mass shift of corresponding peptides, with the ratio of the peak intensities reflecting the relative protein amount.
Until recently, the method was limited to the quanti fication of proteins from cultured cells. However, the development of a SILAC mouse has expanded its utiliza tion to include differential studies of tissues and biological fluids from animal models [27]. SILAC applica tions include studies of cellular signaling, posttrans lational modifications such as phosphorylation, and proteinprotein interactions, as well as protein expression profiling in normal versus diseased cells, and identifica tion of disease biomarkers and pharmacological targets. The method is frequently applied in mechanistic studies of druginduced alterations of cellular signaling in differ ent diseases, including myeloid leukemias; for example, Xiong and Wang [28] used SILAC to examine the mecha nisms underlying the cytotoxicity and therapeutic activity of arsenic trioxide, an ancient and effective secondline drug for therapy of acute promyelocytic leukemia. Others have applied a combination of SILAC and phospho proteomics: Pan et al. [29] studied the effects of different kinase inhibitors on entire phosphoprotein networks, while Liang et al. [30] quantified imatinibinduced changes in the phosphorylation of BCRABL kinase and its substrates in human CML cells. SILAC has also been used to study the molecular pathogenesis of myeloid leukemias, including studies of protein kinase regulation in myeloid leukemia cells compared with other cell types [31], and signaling induced by modulation of the FLT3 receptor tyrosine kinase in AML cells [32]. Some of the future applications of SILAC that hold great promise include the identification of biomarkers and pharmaco logical targets in myeloid/lymphoid or mixed lineage leukemia comprising t(4;11) translocations compared to other types of leukemia [33]. In summary, SILAC is a sensitive method for the study of novel signal trans duction pathways and pharmacological targets, and will probably contribute to the identification of biomarkers that can be brought into clinical use.

Two-dimensional difference gel electrophoresis
Twodimensional electrophoresis is an effective method ology to separate proteins based on charge and size, and this protein separation technique combined with MS protein identification has already led to the successful identification of experimental therapy targets in AML [34,35]. However, limitations in reproducibility and quanti fication have led to the development of the powerful twodimensional differential gel electrophoresis (2DDIGE) [36,37]. Largely correcting for technical variations, 2DDIGE multiplexes two fluorescentlabeled protein samples with an internal standard [36]. The method improves the accuracy of protein quantification and provides highly reproducible results. It has been widely used in the discovery of disease biomarkers, treat ment response and verification of disease classifica tion. In myeloid leukemia, the accumulation of ubiquitinated proteins in the cytosol identified in the APL cell line HL60 treated with methotrexate was found to be caused by the apoptosisassociated downregulation of structural and regulatory proteasome subunits [38]. Shen and colleagues [39] used 2DDIGE to discover the involve ment of CRKlike protein (CRKL) in multidrug resistance of the CML cell line K562/ADM, thereby supporting previous findings that suggested that the protein is impli cated in the pathogenesis of CML. Even though 2DDIGE has clear limitations with respect to the representation of proteins analyzed, it allows the researcher to narrow down the number of potential protein candidates for identification and mapping of protein modification, thereby resulting in a less complex sample set to be further analyzed by MS to identify proteins with func tional impact on disease.

Isotope-coded affinity tags
Introduced by Gygi and coworkers in 1999 [40], isotope coded affinity tags (ICAT) was one of the first develop ments in quantitative proteomics. The reagents used are biotinylated derivates of iodoacetamide that react with the cysteine side chains of reduced and denatured proteins. Heavy and light versions of the reagents are used to label the two different samples before they are combined and digested. Streptavidin affinity chromato graphy excludes peptides that do not contain cysteine residues with biotin, and reduces the complexity of the sample. Before the development of cleavable ICAT [41,42], the presence of biotin complicated the analysis due to the additional weight of the biomarker. However, cleavable ICAT allows the cleavage of biotin from the peptide before the samples are analyzed by MS. The ratio between the heavy and light peptides is used to calculate the relative abundance of the peptides before they are identified by tandem MS (MS/MS). This method has high specificity and sensitivity, and can be applied to many different kinds of samples. The obvious disadvantage with ICAT is its dependence on cysteine residues, as this excludes proteins without this amino acid. Thus far, ICAT has been used to map basic protein interactions in myeloid cell lines [43] and mechanisms of differentiation [44].

Isobaric tags for relative and absolute quantification
Isobaric tags for relative and absolute quantification (iTRAQ) allow the analysis of up to eight separate samples simultaneously [45]; this is a higher level of multiplexing than that of any of the other labeling methods to date. After digestion, the peptides in the samples are labeled with an isobaric tag on the primary amines ( Figure 2). The samples are then combined before MS/MS analysis. The isobaric tags consist of a charged reporter group and a neutral balance group, which are combined so the different tags have identical molecular mass. In MS analysis, the combined samples appear as a single peak, with the different reporters becoming visible after MS/MS. During fragmentation, the two reporter groups split, releasing the neutral reporter. This makes the differently charged tags visible and the intensity of these ion peaks are a measurement of the relative abun dance of the peptide in the different samples. Griffiths and colleagues [46] used iTRAQ to explore imatinib induced effects on the proteome in CML CD34+ cells from patients presenting with chronic phase CML. They found AspGluAlaAsp (DEAD)box protein 3, heat shock protein 105 kDa, and peroxiredoxin3 to be poten tial markers for response to imatinib treatment [46]. Furthermore, AML cells that produce wildtype FLT3, internal tandem duplication FLT3, or D835Y point mutated FLT3 have been examined by iTRAQ to identify and quantify phosphotyrosines. The analysis showed differences in protein phosphorylation of JAK2, signal trans ducer and activator of transcription protein (STAT)5a, and SH2containing protein phosphate (SHP1), and it was concluded that the mutations FLTITD and FLT3 D835Y cause divergent signal responses in AML [47].

Reverse-phase protein arrays
Reversephase protein array (RPPA) is a rapid, high throughput technology for the analysis of patient samples. Originally applied to the investigation of micro dissected tissue from prostate, colon, and head and neck cancer, it is also highly applicable to hematopoietic cells [4850]. In short, cells are lysed and protein samples are spotted onto nitrocellulosecoated glass slides with an automated printer. Slides are blocked, blotted, and incu bated with antibodies before measuring signal intensity by scanning and quantification of results ( Figure 2). Results from RPPA have been shown to be correlated with western blotting [51]. Tibes and colleagues [51] validated the method for hematological cells, proving it to be highly reproducible, even in samples as scarce as the proteomic equivalent to three cells. Kornblau and co workers [52] used RPPA to propose that high levels of phosphorylation of the tumor suppressor forkhead transcription factor 3A is an unfavorable, though thera peutically targetable, prognostic factor in AML. Pigazzi and colleagues [53] observed that the microRNA miR34b mediated downregulation of several target genes (for example, those encoding cyclins A1, B and D1) of cAMP response elementbinding (CREB) in myeloid cell lines and bone marrow samples from pediatric AML patients, and this could possibly explain some of the cell cycle abnormalities found in myeloid cell lines. RPPA has also largely confirmed the prognostic value of Bax/Bcl2 protein production ratio and Mcl1 production [54], as previously reported by flow cytometric analysis [55,56]. These studies show the potential for RPPA in high through put analysis screening for variation in protein production in large patient cohorts.

Multiple reaction monitoring
An emerging strategy for the analysis of potential protein markers is the absolute quantification (AQUA) strategy used in selective reaction monitoring; this is also known as MRM [25,57]. The high sensitivity of this method enables quantitative detection of lowabundance proteins (concentrations down to attomolar levels) in complex mixtures, with dynamic range extending over three orders of magnitude [2225]. The AQUA strategy is based on specific detection of target proteins enabled by internal standard peptides, chosen within the target protein, and produced with a heavy isotope amino acid incorporated [58] (Figure 2). This selective detection of the peptides makes this method the most sensitive of all the MS methods. MRM has, to date, not been applied in studies of myeloid leukemia, but as it becomes more wide spread it is natural to assume that it will take on a more prominent role as a tool for validation of bio markers and clinical diagnostics.

Multiparameter flow cytometry, mass cytometry, and single-cell proteomics
Flow cytometry has been used for the analysis of myeloid malignancies for several decades and its applications include classification, diagnosis and prognosis, therapy response prediction, and monitoring of therapy response. Methodological advances in the form of multiparameter flow cytometry and analysis of entire signaling networks have brought this methodology into the group of techniques available for largescale protein studies. In 2004, the Nolan group demonstrated a novel approach for the analysis of signaling networks of cancer cells [59], in which they used multiparameter flow cytometry for detection of induced phosphoprotein responses at the singlecell level in AML cells. They were thereby able to correlate specific signaling profiles with genetic features and clinical outcome [59]. The same team employed the technique to identify a specific STAT5 signaling signature in myeloid malignancies, and this correlated with specific clinical and biologic correlates [60]. The analysis of signaling profiles at the singlecell level makes it possible to identify pathways that are activated in therapy resistant cells, as well as potential biomarkers for patient diagnosis and prognosis. A major advantage of multi parameter flow cytometry is the small amount of material needed, making it ideal for analysis of patient samples. However, there are several limitations to the method ology, including few detection channels, spectral overlap between signals from fluorescent labels of the antibodies used, and, not least, antibody specificity and epitope blocking in a cluttered cell. Recently, limitations in multiplexing have been overcome by the utilization of element tagging of antibodies, followed by analysis using a mass cytometer, combining the principles of flow cyto metry and MS [61]. Many available stable isotopes (up to 100) can be used in the tags, enabling simultaneous detection of proteins and gene transcripts in individual cells in a quantitative manner using inductively coupled plasma timeofflight mass spectrometry with high resolu tion, sensitivity, and speed of analysis. Simul taneous detection of 20 surface antigens in singlecell analysis of human leukemia cell lines and samples from patients with AML [62] has demonstrated the advantages of the approach, avoiding problems with detection channels, sample matrix, and stability of the experiment setup, in addition to the modest requirements of patient material. All in all, this is a promising approach for the development of a rapid, sensitive, automated method for concomitant detection of many biomarkers in individual cells. Further development of data collection and pro cess ing, including multiparameter clustering algorithms, may provide a novel methodology for investigation of signaling pathways and disease detection.

Nanofluidic proteomic assay
The nanofluidic proteomic assay aims to overcome the limitations in the detection of potential biomarkers in limited material [20,21]. This method, based on iso electric focusing of cell lysate in capillary glass tubes and detection using antibodies and chemiluminescence, can identify and quantify protein production in as little as 25 cells (Figure 2). The isoelectric separation can separate proteins with different phosphorylation states, which can provide insights into the regulation of the protein. Fan and coworkers [20] used this method to detect as little as 2 pg protein in a 4 nl sample, with a dynamic range of three orders of magnitude. In the K562 CML cell line, they detected changes in phosphorylation of STAT3 and STAT5, and changes in activation of caspase 3 and extra cellular signalrelated kinase (ERK)2 after treatment with imatinib [20]. The obvious limitation of the method is the dependency on antibodies and the need for combination with flowbased cell sorting if cell type specificity is to be obtained; however, if these obstacles are overcome, this method can be a powerful tool in diagnostics and the monitoring of disease.

Prospects for protein-based diagnostics and monitoring of response to therapy
Lessons from recent achievements in CML research may illustrate the future of proteinbased diagnostics. CML is the single malady with apparently perfect correlation between the pathognomonic gene aberration BCR-ABL and a therapeutic kinase inhibitor that effectively inhibits ABL tyrosine kinase activity. A proteinbased immuno assay for detection of the gene product BCRABL is in development, and will allow diagnostics of ABLBCR positive leukemia using cellular protein extracts or single cell flow cytometry [63]. Future diagnosis of CML could thereby be based on protein diagnostics, using PCR based DNA diagnostics only as confirmatory analysis.
Proteomic analysis of CML cell lines treated with imatinib has revealed the modulation of several phospho protein targets [30]. SILAC analysis of imatinibtreated CML cells has demonstrated a 90% reduction in phos phorylation of BCRABL kinase, SHIP2 and Dok2, and other modulated proteins, including SHIP1, SH2 contain ing protein and Casitas Blineage lymphoma protooncogene. Imatinibtreated CML cells show an attenuated activation of ERK1 and ERK2, mitogen activated kinase1, STAT3 and STAT5, and cJun N terminal kinase [20]. PhosphoCRKL has been proposed as a marker for monitoring patients with CML treated with imatinib and nilotinib [64], and may be a pseudo marker for detecting resistance against TKI treatment. Direct and indirect detection of resistance in BCRABL positive CML is underscored by reports of significantly lower levels of BCRABL, CRKL (Tyr207) and AKT (Ser473) in resistant patients. CML may be the first disease where we achieve fully proteinbased diagnostics and therapy response monitoring.

Conclusions
Future use of proteinbased diagnostics in cancer may reflect clearer the epigenetics and genome alterations in leukemia, and represent a test more directly related to the cellular protein targets of therapeutics. It is still unclear which technology platform will be the dominat ing workhorse in clinical proteomics, but a technology that is 'open' in terms of target selection seems to hold the strongest appeal. Maybe the lessons from modern vitamin and hormone analysis indicate a promising path [65,66], as a high number of molecules can be identified and quantified in minute clinical samples. In diagnostics of leukemia, validated MS analysis of peptides, with high flexibility and high sensitivity, may allow determination of classification and therapeutic targets [67]. Celltype specific analysis is necessary for diagnostic information in leukemia and other hematological malignancies; this analytic quality is currently provided by flow cytometry only. Detection of modified proteins in intracellular signal transduction pathways may represent a novel diag nostic tool, as exemplified in JCMML, where phospho specific flow cytometry may replace cumbersome and timeconsuming growth assays [60]. Furthermore, protein analyses may play a role in future monitoring of therapy response evaluation, but carefully designed clinical trials will be needed to determine the role of proteomics in therapy guidance.