Optimal approach for classification of acute leukemia subtypes based on gene expression data

The classification of cancer subtypes, which is critical for successful treatment, has been studied extensively with the use of gene expression profiles from oligonucleotide chips or cDNA microarrays. Various pattern recognition methods have been successfully applied to gene expression data. However, these methods are not optimal, rather they are high-performance classifiers that emphasize only classification accuracy. In this paper, we propose an approach for the construction of the optimal linear classifier using gene expression data. Two linear classification methods, linear discriminant analysis (LDA) and discriminant partial least-squares (DPLS), are applied to distinguish acute leukemia subtypes. These methods are shown to give satisfactory accuracy. Moreover, we determined optimally the number of genes participating in the classification (a remarkably small number compared to previous results) on the basis of the statistical significance test. Thus, the proposed method constructs the optimal classifier that is composed of a small size predictor and provides high accuracy.     

Relationship between c-Kit expression and proliferation in acute myeloblastic leukemia cell lines

We have proposed that the transmembrane receptor encoded by the c-Kit protooncogene and its ligand play an important role in regulating the proliferation of blasts cells in acute myeloblastic leukemia (AML). To test this hypothesis, immunobeads were used to separate blasts from three Kit-expression positive cell lines into strongly Kit-protein positive and weakly Kit-protein positive fractions. The strongly positive fraction had greater proliferative potential than the weakly positive fraction as assessed both by colony-formation in methylcellulose and growth of clonogenic cells in suspension. The reproducibility of the percentage of each blast population found in the strongly and weakly positive fractions provided evidence that Kit-protein expression is regulated. Kinetic experiments provided evidence for reversible transitions between strong and weak Kit protein expression. Thus regulated expression of the Kit receptor may be a mechanism for controlling blast cell growth in culture. © 1993 Wiley-Liss, Inc.     

Synovial membrane protein expression differs between juvenile idiopathic arthritis subtypes in early disease

Introduction

Juvenile idiopathic arthritis (JIA) is the most common rheumatological disease of childhood with a prevalence of around 1 in 1,000. Without appropriate treatment it can have devastating consequences including permanent disability from joint destruction and growth deformities. Disease aetiology remains unknown. Investigation of disease pathology at the level of the synovial membrane is required if we want to begin to understand the disease at the molecular and biochemical level. The synovial membrane proteome from early disease-stage, treatment naive JIA patients was compared between polyarticular and oligoarticular subgroups.

Methods

Protein was extracted from 15 newly diagnosed, treatment naive JIA synovial membrane biopsies and separated by two dimensional fluorescent difference in-gel electrophoresis. Proteins displaying a two-fold or greater change in expression levels between the two subgroups were identified by matrix assisted laser desorption ionization-time of flight mass spectrometry with expression further verified by Western blotting and immunohistochemistry.

Results

Analysis of variance analysis (P ≤ 0.05) revealed 25 protein spots with a two-fold or greater difference in expression levels between polyarticular and oligoarticular patients. Hierarchical cluster analysis with Pearson ranked correlation revealed two distinctive clusters of proteins. Some of the proteins that were differentially expressed included: integrin alpha 2b (P = 0.04); fibrinogen D fragment (P = 0.005); collagen type VI (P = 0.03); fibrinogen gamma chain (P = 0.05) and peroxiredoxin 2 (P = 0.02). The identified proteins are involved in a number of different processes including platelet activation and the coagulation system.

Conclusions

The data indicate distinct synovial membrane proteome profiles between JIA subgroups at an early stage in the disease process. The identified proteins also provide insight into differentially perturbed pathways which could influence pathological events at the joint level.     

Comparative proteome analysis of acute myeloid leukemia with and without maturation

Acute myeloid leukemia (AML) is a severe, rapidly progressing disease triggered by blocking granulocyte or monocyte differentiation and maturation. Because of its heterogeneity, AML is divided into a number of subtypes. Unfortunately, so far very few correlations have been found between AML classification and its clinical course or patient response to treatment. In addition, as yet only a few subtype-specific AML biomarkers have been discovered. To solve these problems here, we focused on two AML subtypes M1 and M2 that are especially difficult to differentiate. Using 2D electrophoresis and mass spectrometry, we analyzed the protein profiles of peripheral blood (PB) and/or bone marrow (BM) samples collected from 38 AML-M1/M2 patients and 17 healthy volunteers. Comparative analysis of AML-M1/M2 and control PB/BM cells revealed 25 proteins that accumulated differentially. Hierarchical clustering of proteomic results clearly divided the AML samples into 2 groups (M1 and M2). Annexin III, L-plastin and 6-phosphogluconate dehydrogenase were found only in the M2 group. We also observed that the levels of annexin I and actin gamma 1 were correlated with resistance to treatment and the time of relapse. It appears that these five proteins can serve as potential AML biomarkers.     

Classifying subtypes of acute lymphoblastic leukemia using silhouette statistics and genetic algorithms

Correct classification and prediction of tumor cells is essential for a successful diagnosis and reliable future treatment. In this study, we aimed at using genetic algorithms for feature selection and proposed silhouette statistics as a discriminant function to distinguish between six subtypes of pediatric acute lymphoblastic leukemia by using microarray with thousands of gene expressions. Our methods have shown a better classification accuracy than previously published methods and obtained a set of genes effective to discriminate subtypes of pediatric acute lymphoblastic leukemia. Furthermore, the use of silhouette statistics, offering the advantages of measuring the classification quality by a graphical display and by an average silhouette width, has also demonstrated feasibility and novelty for more difficult multiclass tumor prediction problems.     

Alpha-tocopherol content and alpha-tocopherol transfer protein expression in leukocytes of children with acute leukemia

-Tocopherol (a form of vitamin E) is a fat-soluble vitamin that can prevent lipid peroxidation of cell membranes. This antioxidant activity of -tocopherol can help to prevent cardiovascular disease, atherosclerosis and cancer. We investigated the -tocopherol level and the expression of -tocopherol transfer protein (-TTP) in the leukocytes of children with leukemia. The plasma and erythrocyte -tocopherol levels did not differ between children with leukemia and the control group. However, lymphocytes from children with leukemia had significantly lower -tocopherol levels than lymphocytes from the controls (58.4±39.0 ng/mg protein versus 188.9±133.6, respectively; p0.05), despite the higher plasma -tocopherol/cholesterol ratio in the leukemia group (5.83±1.64 μmol/mmol versus 4.34±0.96, respectively; p0.05). No significant differences in the plasma and leukocyte levels of isoprostanes (the oxidative metabolites of arachidonic acid) were seen between the leukemia patients and controls. The plasma level of acrolein, a marker of oxidative stress, was also similar in the two groups. Investigation of -TTP expression by leukocytes using real-time PCR showed no difference between the two groups. These findings suggest that there may be comparable levels of lipid peroxidation in children with untreated leukemia and controls, despite the reduced -tocopherol level in leukemic leukocytes.     

Simple rules underlying gene expression profiles of more than six subtypes of acute lymphoblastic leukemia (ALL) patients

MOTIVATIONS AND RESULTS: For classifying gene expression profiles or other types of medical data, simple rules are preferable to non-linear distance or kernel functions. This is because rules may help us understand more about the application in addition to performing an accurate classification. In this paper, we discover novel rules that describe the gene expression profiles of more than six subtypes of acute lymphoblastic leukemia (ALL) patients. We also introduce a new classifier, named PCL, to make effective use of the rules. PCL is accurate and can handle multiple parallel classifications. We evaluate this method by classifying 327 heterogeneous ALL samples. Our test error rate is competitive to that of support vector machines, and it is 71% better than C4.5, 50% better than Naive Bayes, and 43% better than k-nearest neighbour. Experimental results on another independent data sets are also presented to show the strength of our method. AVAILABILITY: Under http://sdmc.lit.org.sg/GEDatasets/, click on Supplementary Information.     

Transglutaminase 2 expression in acute myeloid leukemia: Association with adhesion molecule expression and leukemic blast motility

Acute myeloid leukemia (AML) is a heterogenous disease with differential oncogene association, outcome and treatment regimens. Treatment strategies for AML have improved outcome but despite increased molecular biological information AML is still associated with poor prognosis. Proteomic analysis on the effects of a range of leukemogenic oncogenes showed that the protein transglutaminase 2 (TG2) is expressed at greater levels as a consequence of oncogenic transformation. Further analysis of this observation was performed with 511 AML samples using reverse phase proteomic arrays, demonstrating that TG2 expression was higher at relapse than diagnosis in many cases. In addition elevated TG2 expression correlated with increased expression of numerous adhesion proteins and many apoptosis regulating proteins, two processes related to leukemogenesis. TG2 has previously been linked to drug resistance in cancer and given the negative correlation between TG2 levels and peripheral blasts observed increased TG2 levels may lead to the protection of the leukemic stem cell due to increased adhesion/reduced motility. TG2 may therefore form part of a network of proteins that define poor outcome in AML patients and potentially offer a target to sensitize AML stem cells to drug treatment.     

High expression of heat shock protein 90 alpha and its significance in human acute leukemia cells

This study investigated the expression of heat shock protein 90 alpha (Hsp90α) in acute leukemia cells. The expression of Hsp90α was investigated in leukemia cell lines and human bone marrow mononuclear cells derived from acute leukemia patients and from healthy individuals using polymerase chain reaction, Western blot, and enzyme-linked immunosorbent assay. Compared with cells from healthy individuals, the expression of Hsp90α in the untreated patients was higher. Similarly high levels were observed in remission patients. Significantly higher expression levels were observed in all the tested cell lines, and in cells from refractory and relapsed patients. No obvious relationship was observed between the occurrence of graft versus host disease and the expression of Hsp90α. The untreated patients showing higher expression levels of Hsp90α had lower complete remission rates. During remission of untreated patients, the expression of Hsp90α decreased and reached the lowest level after transplantation, but the expression increased again before relapse. Hsp90α was highly expressed in leukemia cells. The expression level of Hsp90α was associated with leukemia prognosis. However, no obvious relationship was observed between the occurrence of graft versus host disease and the expression of Hsp90α.     

Human but not murine toll-like receptor 2 discriminates between tri-palmitoylated and tri-lauroylated peptides

Toll-like receptors (TLRs) mediate activation of the immune system upon challenge with microbial agonists, components of disintegrating cells of the body, or metabolic intermediates of lipidic nature. Comparison of murine (m) and human (h) TLR2 primary sequences revealed 65% of identical residues within the extracellular domains in contrast to 84% in the intracellular domains. Comparative analysis of TLR2-driven cell activation by various TLR2 agonists showed that the tri-lauroylated lipopeptide analog (Lau(3)CSK(4)) is recognized efficiently through mTLR2 but not hTLR2. Genetically complemented human embryonic kidney 293 cells and murine TLR2(-/-) embryonic fibroblasts, as well as human and murine macrophage cells, were used for this analysis. In contrast to cellular activation, which depended on blockable access of the TLR2-ligand to TLR2, cellular uptake of Lau(3)CSK(4) and tri-palmitoylated peptide (P(3)CSK(4)) was independent of TLR2. A low-conserved region spanning from leucine-rich repeat (LRR) motif 7 to 10 was found to control TLR2 species-specific cell activation. Exchange of mLRR8 for hLRR8 in mTLR2 abrogated mTLR2-typical cell activation upon cellular challenge with Lau(3)CSK(4) but not P(3)CSK(4), implicating mLRR8 as a central element of Lau(3)CSK(4) recognition. The point mutation L112P within LRR3 abrogated hTLR2-dependent recognition of lipopeptides but merely attenuated mTLR2 function, whereas deletion of the N-terminal third of each LRR-rich domain (LRRs 1 to 7) had the opposite effect on P(3)CSK(4) recognition. Despite similar domain structure of both TLR2 molecules, species-specific properties thus exist. Our results imply distinct susceptibilities of humans and mice to challenge with specific TLR2 ligands.     

AF-DX 116 discriminates between muscarinic M2 receptors of the heart and vasculature

The newly developed muscarinic antagonist, AF-DX 116, has been reported to have a higher affinity in vivo for muscarinic receptors in the heart than in the vasculature. Therefore two in vitro preparations, the rabbit ear artery and spontaneously beating right atrium, were compared. AF-DX 116 had a 29 times greater affinity for muscarinic receptors in the cardiac preparation than in the ear artery, with a pA2 in the heart of 7.42 compared to a value of 5.95 in the ear artery. Thus AF-DX 116 shows promise as an approach to differentiating sub-classes of muscarinic M2 receptors.     

Calnexin discriminates between protein conformational states and functions as a molecular chaperone in vitro.

Although calnexin is thought to function as a molecular chaperone for glycoproteins, a prevalent view is that it cannot distinguish between protein conformational states, binding solely through its lectin site to monoglucosylated oligosaccharides. Using purified components in vitro, calnexin effectively prevented the aggregation not only of glycoproteins bearing monoglucosylated oligosaccharides but also proteins lacking N-glycans, an effect enhanced by ATP. It also suppressed the thermal denaturation of nonglycosylated proteins and enhanced their refolding in conjunction with other cellular components. Calnexin formed stable complexes with unfolded conformers of these proteins but not with the native molecules. Therefore, in addition to being a lectin, calnexin functions as a bona fide molecular chaperone capable of interacting with polypeptide segments of folding glycoproteins.     

Encoding of Ca2+ signals by differential expression of IP3 receptor subtypes

Inositol 1,4,5-trisphosphate (IP3) plays a key role in Ca2+ signalling, which exhibits a variety of spatio-temporal patterns that control important cell functions. Multiple subtypes of IP3 receptors (IP3R-1, -2 and -3) are expressed in a tissue- and development-specific manner and form heterotetrameric channels through which stored Ca2+ is released, but the physiological significance of the differential expression of IP3R subtypes is not known. We have studied the Ca2+-signalling mechanism in genetically engineered B cells that express either a single or a combination of IP3R subtypes, and show that Ca2+-signalling patterns depend on the IP3R subtypes, which differ significantly in their response to agonists, i.e. IP3, Ca2+ and ATP. IP3R-2 is the most sensitive to IP3 and is required for the long lasting, regular Ca2+ oscillations that occur upon activation of B-cell receptors. IP3R-1 is highly sensitive to ATP and mediates less regular Ca2+ oscillations. IP3R-3 is the least sensitive to IP3 and Ca2+, and tends to generate monophasic Ca2+ transients. Furthermore, we show for the first time functional interactions between coexpressed subtypes. Our results demonstrate that differential expression of IP3R subtypes helps to encode IP3-mediated Ca2+ signalling.     

Drug concentration-dependent expression of multidrug resistance-associated protein and P-glycoprotein in the doxorubicin-resistant acute myelogenous leukemia sublines.

The multidrug resistance of cancer cells can be mediated by an overexpression of the human MDR1 and MRP genes, which encode the transmembrane efflux pumps, the 170 kDa P-glycoprotein (Pgp) and the 190 kDa multidrug resistance-associated protein (MRP), respectively. In this study, we investigate which protein is preferentially overexpressed in the function of doxorubicin concentrations in the acute myelogenous leukemia cell line (OCI/AML-2). Multidrug-resistant AML-2 sublines were isolated in doxorubicin concentrations of 20, 100, 250, and 500 ng/ml. MRP was at first expressed at low concentrations of less than 5 x IC50 (100 ng/ml) of doxorubicin followed by the overexpression of Pgp with concentrations of more than 12.5 x IC50 (250 ng/ml) of doxorubicin. In addition, it appeared that increased amounts of MRP and its mRNA in AML-2/DX20 and /DX100 decreased gradually in both AML-2/DX250 and /DX500 overexpressing Pgp. In conclusion, it is thought that the overexpression of MRP or Pgp is dependent upon drug concentrations. It could be implicated that the overexpression of MRP might be negatively related to that of Pgp.     

Image morphometry of acute leukemias. Comparison between lymphoid and myeloid subtypes

OBJECTIVE: To determine if image morphometry has any role in distinguishing blasts of acute lymphoblastic leukemia (ALL-L2) from those of acute myeloid leukemia (AML-M1) and (AML-M2). STUDY DESIGN: Ten cases each of ALL-L2, AML-M1 and AML-M2 diagnosed according to the French-American-British criteria were studied. In all cases May-Grünwald-Giemsa-stained bone marrow aspiration smears were obtained. At least 100 blast cells from each case were subjected to analysis randomly with an image cytometer using Leica Quantimet 600 software (Cambridge, U.K.). The area, convex area, length, width, perimeter, convex perimeter, roundness, total optical density, average optical density and pixel grey value variance of the nuclei were measured by random selection of cells using a 40:1 objective (1 pixel = 0.446 micron). RESULTS: Mann Whitney's nonparametric test showed that there was considerable overlap of morphometric variables between the 3 subtypes. Though statistical significance was found in "roundness" between blasts of AML-M1 and ALL-L2, power analyses (sample size of 100 blasts of each subtype) did not show sufficient power for this variable. However, between blasts of ALL-L2 and AML-M2, "average optical density" and "pixel grey value variance" were statistically significant with full power using power analyses. CONCLUSION: Image morphometry may be helpful in differentiating blasts from lymphoid and myeloid leukemic subtypes.