Identification of a 24-gene prognostic signature that improves the European LeukemiaNet risk classification of acute myeloid leukemia: an international collaborative study

PURPOSE

To identify a robust prognostic gene expression signature as an independent predictor of survival of patients with acute myeloid leukemia (AML) and use it to improve established risk classification.

PATIENTS AND METHODS

Four independent sets totaling 499 patients with AML carrying various cytogenetic and molecular abnormalities were used as training sets. Two independent patient sets composed of 825 patients were used as validation sets. Notably, patients from different sets were treated with different protocols, and their gene expression profiles were derived using different microarray platforms. Cox regression and Kaplan-Meier methods were used for survival analyses.

RESULTS

A prognostic signature composed of 24 genes was derived from a meta-analysis of Cox regression values of each gene across the four training sets. In multivariable models, a higher sum value of the 24-gene signature was an independent predictor of shorter overall (OS) and event-free survival (EFS) in both training and validation sets (P < .01). Moreover, this signature could substantially improve the European LeukemiaNet (ELN) risk classification of AML, and patients in three new risk groups classified by the integrated risk classification showed significantly (P < .001) distinct OS and EFS.

CONCLUSION

Despite different treatment protocols applied to patients and use of different microarray platforms for expression profiling, a common prognostic gene signature was identified as an independent predictor of survival of patients with AML. The integrated risk classification incorporating this gene signature provides a better framework for risk stratification and outcome prediction than the ELN classification.

RUNX1 mutations are associated with poor outcome in younger and older patients with cytogenetically normal acute myeloid leukemia and with distinct gene and MicroRNA expression signatures

PURPOSE

To determine the association of RUNX1 mutations with therapeutic outcome in younger and older patients with primary cytogenetically normal acute myeloid leukemia (CN-AML) and with gene/microRNA expression signatures.

PATIENTS AND METHODS

Younger (< 60 years; n = 175) and older (≥ 60 years; n = 225) patients with CN-AML treated with intensive cytarabine/anthracycline-based first-line therapy on Cancer and Leukemia Group B protocols were centrally analyzed for RUNX1 mutations by polymerase chain reaction and direct sequencing and for established prognostic gene mutations. Gene/microRNA expression profiles were derived using microarrays.

RESULTS

RUNX1 mutations were found in 8% and 16% of younger and older patients, respectively (P = .02). They were associated with ASXL1 mutations (P < .001) and inversely associated with NPM1 (P < .001) and CEBPA (P = .06) mutations. RUNX1-mutated patients had lower complete remission rates (P = .005 in younger; P = .006 in older) and shorter disease-free survival (P = .058 in younger; P < .001 in older), overall survival (P = .003 in younger; P < .001 in older), and event-free survival (P < .001 for younger and older) than RUNX1 wild-type patients. Because RUNX1 mutations were more common in older patients and almost never coexisted with NPM1 mutations, RUNX1 mutation-associated expression signatures were derived in older, NPM1 wild-type patients and featured upregulation of genes normally expressed in primitive hematopoietic cells and B-cell progenitors, including DNTT, BAALC, BLNK, CD109, RBPMS, and FLT3, and downregulation of promoters of myelopoiesis, including CEBPA and miR-223.

CONCLUSION

RUNX1 mutations are twice as common in older than younger patients with CN-AML and negatively impact outcome in both age groups. RUNX1-mutated blasts have molecular features of primitive hematopoietic and lymphoid progenitors, potentially leading to novel therapeutic approaches.

Abstract 1307: Heritable polymorphism predisposes to high expression of BAALC in cytogenetically normal acute myeloid leukemia (CN-AML)

Introduction: Risk-adapted therapy of AML patients (pts) is presently based on cytogenetic and molecular findings. However, the outcome remains poor for the majority of pts. Overexpression of BAALC is implicated in leukemogenesis and poor survival of CN-AML pts, but the underlying mechanisms are unknown. We hypothesized that the elevated expression of BAALC might result from a genetic variant in or close to the gene (in cis). Methods and Results: BAALC and RUNX1 expression were measured in pre-therapy samples from a test set of 253 CN-AML pts (Cancer And Leukemia Group B) using Affymetrix U133 plus 2.0 arrays. Pts were grouped as high or low BAALC and high or low RUNX1 expressers using the median expression value as the cut-off and screened for single nucleotide polymorphisms (SNPs) by direct sequencing of the BAALC gene. Nine informative SNPs were identified. By genotyping, marker rs62527607, a non-coding SNP in the promoter region correlated with high BAALC expression (rs62527607G&gt;T: genotypes TT/GT vs. GG: P=2.01E-04). Luciferase reporter constructs (pGL4.11) spanning the respective alleles demonstrated that the T allele (“risk allele”) of rs62527607 increased luciferase activity compared to the G allele (P=1.0E-02). The T allele of rs62527607 is predicted to create a binding site for the transcription factor RUNX1. Indeed, RUNX1-cotransfection showed an increase in luciferase activity for the T allele of rs62527607 (P=2.0E-02) but not for the G allele (P=2.7E-01). An Electrophoretic Mobility Shift Assay comparing the binding activity of the two alleles showed highly increased binding activity of RUNX1 to the T allele vs. the G allele. Testing of the pt samples for a potential correlation of RUNX1 and BAALC expression levels revealed a positive correlation in pts with genotypes TT/GT but not GG when restricting the analysis to the high RUNX1 expressing group (P=1.18E-04). For validation of our findings, a set of 105 CN-AML pts comprising 52 high and 53 low BAALC expressers (AML Study Group [AMLSG], quantitative Real-Time PCR [qPCR]) was genotyped for rs62527607, determined for RUNX1 expresser status (qPCR) and analyzed alone and in combination with the test set. A non-significant association of TT/GT vs. GG with high BAALC expression (31% vs. 19%; P=1.6E-01) was seen. Analyzing both sets combined resulted in a P-value of 9.3E-05 (Odds Ratio [OR]=2.59 [1.62, 4.22; 95%CI]). Combining the results of the CALGB & AMLSG series showed a highly significant association between RUNX1 and BAALC expression in pts with genotypes TT/GT belonging to the high RUNX1 expressing group (P=7.97E-05, OR=4.02 [2.05, 8.21; 95%CI]). Conclusion: We show that the risk allele of rs62527607 creates a RUNX1 binding site, thereby leading to increased transcription of BAALC. The observation that a heritable trait is involved in the leukemogenesis process might initiate a new era of personalized medicine in AML.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 1307. doi:1538-7445.AM2012-1307

Age-related prognostic impact of different types of DNMT3A mutations in adults with primary cytogenetically normal acute myeloid leukemia

PURPOSE

To determine the frequency of DNMT3A mutations, their associations with clinical and molecular characteristics and outcome, and the associated gene- and microRNA-expression signatures in primary cytogenetically normal acute myeloid leukemia (CN-AML).

PATIENTS AND METHODS

Four hundred fifteen previously untreated adults were analyzed for DNMT3A mutations and established prognostic gene mutations and expression markers. Gene- and microRNA-expression profiles were derived using microarrays.

RESULTS

Younger (< 60 years; n = 181) and older (≥ 60 years; n = 234) patients had similar frequencies of DNMT3A mutations (35.3% v 33.3%). Missense mutations affecting arginine codon 882 (R882-DNMT3A) were more common (n = 92; 62%) than those affecting other codons (non-R882-DNMT3A). DNMT3A-mutated patients did not differ regarding complete remission rate, but had shorter disease-free survival (DFS; P = .03) and, by trend, overall survival (OS; P = .07) than DNMT3A-wild-type patients. In multivariable analyses, DNMT3A mutations remained associated with shorter DFS (P = .01), but not with shorter OS. When analyzed separately, the two DNMT3A mutation types had different significance by age group. Younger patients with non-R882-DNMT3A mutations had shorter DFS (P = .002) and OS (P = .02), whereas older patients with R882-DNMT3A mutations had shorter DFS (P = .005) and OS (P = .002) after adjustment for other clinical and molecular prognosticators. Gene- and microRNA-expression signatures did not accurately predict DNMT3A mutational status.

CONCLUSION

DNMT3A mutations are frequent in CN-AML, and their clinical significance seems to be age dependent. DNMT3A-R882 mutations are associated with adverse prognosis in older patients, and non-R882-DNMT3A mutations are associated with adverse prognosis in younger patients. Low accuracy of gene- and microRNA-expression signatures in predicting DNMT3A mutation status suggested that the role of these mutations in AML remains to be elucidated.

Clinical outcome and gene- and microRNA-expression profiling according to the Wilms tumor 1 (WT1) single nucleotide polymorphism rs16754 in adult de novo cytogenetically normal acute myeloid leukemia: a Cancer and Leukemia Group B study

BACKGROUND

The alleles of the Wilms tumor 1 (WT1) polymorphism rs16754 harbor adenine (A) or guanine (G). Recently, rs16754 has been reported to affect the outcome of patients with cytogenetically normal acute myeloid leukemia. To validate this finding, we investigated pretreatment features and outcome associated with rs16754 in a large cohort of patients with cytogenetically normal acute myeloid leukemia.

DESIGN AND METHODS

Four-hundred and thirty-three intensively treated and molecularly characterized cytogenetically normal patients with de novo acute myeloid leukemia (18-83 years old) were analyzed for rs16754. To gain biological insights, we studied the gene- and microRNA-expression profiles for associations with rs16754.

RESULTS

Three-hundred and nine (71%) patients were homozygous for A (WT1(AA)), 112 (26%) were heterozygous (WT1(AG)) and 12 (3%) were homozygous for G (WT1(GG)). For comparison with previous studies, we grouped WT1(AG) and WT1(GG) patients and compared them with WT1(AA) patients divided into younger (<60 years) and older (≥60 years) adults. We found no independent prognostic impact of WT1(AA). However, WT1(GG) patients, who were less often Caucasian than WT1(AG) (P=0.001) or WT1(AA) (P=0.008) patients, and had TET2 mutations more often than WT1(AG) (P=0.02) patients, had, among patients with FLT3-internal tandem duplication and/or NPM1 wild-type, better disease-free (P=0.02) and overall survival (P=0.04) than WT1(AA) and WT1(AG) patients combined. Unsupervised and supervised analyses of the gene- and microRNA-expression profiles suggested that there were no distinct expression patterns associated with any rs16754 genotype.

CONCLUSIONS

We did not observe the previously reported adverse impact of WT1(AA) but found favorable outcomes associated with the homozygous WT1(GG). Considering its low frequency, confirmatory studies are necessary. The biological significance of rs16754 remains questionable as no distinct expression profiles were associated with the genotypes.

TET2 mutations improve the new European LeukemiaNet risk classification of acute myeloid leukemia: a Cancer and Leukemia Group B study

PURPOSE

To determine the frequency of TET2 mutations, their associations with clinical and molecular characteristics and outcome, and the associated gene- and microRNA-expression signatures in patients with primary cytogenetically normal acute myeloid leukemia (CN-AML).

PATIENTS AND METHODS

Four-hundred twenty-seven patients with CN-AML were analyzed for TET2 mutations by polymerase chain reaction and direct sequencing and for established prognostic gene mutations. Gene- and microRNA-expression profiles were derived using microarrays.

RESULTS

TET2 mutations, found in 23% of patients, were associated with older age (P < .001) and higher pretreatment WBC (P = .04) compared with wild-type TET2 (TET2-wt). In the European LeukemiaNet (ELN) favorable-risk group (patients with CN-AML who have mutated CEBPA and/or mutated NPM1 without FLT3 internal tandem duplication [FLT3-ITD]), TET2-mutated patients had shorter event-free survival (EFS; P < .001) because of a lower complete remission (CR) rate (P = .007), and shorter disease-free survival (DFS; P = .003), and also had shorter overall survival (P = .001) compared with TET2-wt patients. TET2 mutations were not associated with outcomes in the ELN intermediate-I-risk group (CN-AML with wild-type CEBPA and wild-type NPM1 and/or FLT3-ITD). In multivariable models, TET2 mutations were associated with shorter EFS (P = .004), lower CR rate (P = .03), and shorter DFS (P = .05) only among favorable-risk CN-AML patients. We identified a TET2 mutation-associated gene-expression signature in favorable-risk but not in intermediate-I-risk patients and found distinct mutation-associated microRNA signatures in both ELN groups.

CONCLUSION

TET2 mutations improve the ELN molecular-risk classification in primary CN-AML because of their adverse prognostic impact in an otherwise favorable-risk patient subset. Our data suggest that these patients may be candidates for alternative therapies.

Prognostic significance of expression of a single microRNA, miR-181a, in cytogenetically normal acute myeloid leukemia: a Cancer and Leukemia Group B study

PURPOSE

To evaluate the prognostic significance of expression levels of a single microRNA, miR-181a, in the context of established molecular markers in cytogenetically normal acute myeloid leukemia (CN-AML), and to gain insight into the leukemogenic role of miR-181a.

PATIENTS AND METHODS

miR-181a expression was measured in pretreatment marrow using Ohio State University Comprehensive Cancer Center version 3.0 arrays in 187 younger (<60 years) adults with CN-AML. Presence of other molecular prognosticators was assessed centrally. A gene-expression profile associated with miR-181a expression was derived using microarrays and evaluated by Gene-Ontology analysis.

RESULTS

Higher miR-181a expression associated with a higher complete remission (CR) rate (P=.04), longer overall survival (OS; P=.01) and a trend for longer disease-free survival (DFS; P=.09). The impact of miR-181a was most striking in poor molecular risk patients with FLT3-internal tandem duplication (FLT3-ITD) and/or NPM1 wild-type, where higher miR-181a expression associated with a higher CR rate (P=.009), and longer DFS (P<.001) and OS (P<.001). In multivariable analyses, higher miR-181a expression was significantly associated with better outcome, both in the whole patient cohort and in patients with FLT3-ITD and/or NPM1 wild-type. These results were also validated in an independent set of older (≥60 years) patients with CN-AML. A miR-181a-associated gene-expression profile was characterized by enrichment of genes usually involved in innate immunity.

CONCLUSION

To our knowledge, we provide the first evidence that the expression of a single microRNA, miR-181a, is associated with clinical outcome of patients with CN-AML and may refine their molecular risk classification. Targeted treatments that increase endogenous levels of miR-181a might represent novel therapeutic strategies.

MicroRNA expression in acute myeloid leukemia

Acute myeloid leukemia (AML) is a group of diseases that are very heterogeneous with regard to cytogenetic aberrations, gene mutations, and changes in expression of numerous genes. A new class of genes known as microRNAs recently was found to be involved in myeloid leukemogenesis. These genes are transcribed into regulatory, noncoding RNAs that control mRNA and protein expression of target genes. Genome-wide analyses of microRNA expression have revealed signatures associated with selected cytogenetic and molecular subsets of AML and have led to the recognition of previously unreported molecular pathways involved in myeloid leukemogenesis. In cytogenetically normal AML, microRNA-expression profiling has also provided prognostic information in addition to that obtained from cytogenetics and analyses of gene mutations and aberrant gene expression. This article reviews recent studies that were focused on the alterations of microRNA expression in AML and their diagnostic and prognostic significance.

IDH1 and IDH2 gene mutations identify novel molecular subsets within de novo cytogenetically normal acute myeloid leukemia: a Cancer and Leukemia Group B study

PURPOSE To analyze the frequency and associations with prognostic markers and outcome of mutations in IDH genes encoding isocitrate dehydrogenases in adult de novo cytogenetically normal acute myeloid leukemia (CN-AML). PATIENTS AND METHODS Diagnostic bone marrow or blood samples from 358 patients were analyzed for IDH1 and IDH2 mutations by DNA polymerase chain reaction amplification/sequencing. FLT3, NPM1, CEBPA, WT1, and MLL mutational analyses and gene- and microRNA-expression profiling were performed centrally. Results IDH mutations were found in 33% of the patients. IDH1 mutations were detected in 49 patients (14%; 47 with R132). IDH2 mutations, previously unreported in AML, were detected in 69 patients (19%; 13 with R172 and 56 with R140). R172 IDH2 mutations were mutually exclusive with all other prognostic mutations analyzed. Younger age (< 60 years), molecular low-risk (NPM1-mutated/FLT3-internal tandem duplication-negative) IDH1-mutated patients had shorter disease-free survival than molecular low-risk IDH1/IDH2-wild-type (wt) patients (P = .046). R172 IDH2-mutated patients had lower complete remission rates than IDH1/IDH2wt patients (P = .007). Distinctive microarray gene- and microRNA-expression profiles accurately predicted R172 IDH2 mutations. The highest expressed gene and microRNAs in R172 IDH2-mutated patients compared with the IDH1/IDH2wt patients were APP (previously associated with complex karyotype AML) and miR-1 and miR-133 (involved in embryonal stem-cell differentiation), respectively. CONCLUSION IDH1 and IDH2 mutations are recurrent in CN-AML and have an unfavorable impact on outcome. The R172 IDH2 mutations, previously unreported in AML, characterize a novel subset of CN-AML patients lacking other prognostic mutations and associate with unique gene- and microRNA-expression profiles that may lead to the discovery of novel, therapeutically targetable leukemogenic mechanisms.

Sp1/NFkappaB/HDAC/miR-29b regulatory network in KIT-driven myeloid leukemia

The biologic and clinical significance of KIT overexpression that associates with KIT gain-of-function mutations occurring in subsets of acute myeloid leukemia (AML) (i.e., core binding factor AML) is unknown. Here, we show that KIT mutations lead to MYC-dependent miR-29b repression and increased levels of the miR-29b target Sp1 in KIT-driven leukemia. Sp1 enhances its own expression by participating in a NFkappaB/HDAC complex that further represses miR-29b transcription. Upregulated Sp1 then binds NFkappaB and transactivates KIT. Therefore, activated KIT ultimately induces its own transcription. Our results provide evidence that the mechanisms of Sp1/NFkappaB/HDAC/miR-29b-dependent KIT overexpression contribute to leukemia growth and can be successfully targeted by pharmacological disruption of the Sp1/NFkappaB/HDAC complex or synthetic miR-29b treatment in KIT-driven AML.

Favorable prognostic impact of NPM1 mutations in older patients with cytogenetically normal de novo acute myeloid leukemia and associated gene- and microRNA-expression signatures: a Cancer and Leukemia Group B study

PURPOSE

To analyze the prognostic significance of NPM1 mutations, and the associated gene- and microRNA-expression signatures in older patients with de novo, cytogenetically normal acute myeloid leukemia (CN-AML) treated with intensive chemotherapy.

PATIENTS AND METHODS

One hundred forty-eight adults age >or= 60 years with de novo CN-AML, enrolled onto Cancer and Leukemia Group B protocols 9720 and 10201, were studied at diagnosis for NPM1, FLT3, CEBPA, and WT1 mutations, and gene- and microRNA-expression profiles.

RESULTS

Patients with NPM1 mutations (56%) had higher complete remission (CR) rates (84% v 48%; P < .001) and longer disease-free survival (DFS; P = .047; 3-year rates, 23% v 10%) and overall survival (OS; P < .001; 3-year rates, 35% v 8%) than NPM1 wild-type patients. In multivariable analyses, NPM1 mutations remained independent predictors for higher CR rates (P < .001) and longer DFS (P = .004) and OS (P < .001), after adjustment for other prognostic clinical and molecular variables. Unexpectedly, the prognostic impact of NPM1 mutations was mainly observed in patients >or= 70 years. Gene- and microRNA-expression profiles associated with NPM1 mutations were similar across older patient age groups and similar to those in younger (< 60 years) patients with CN-AML. These profiles were characterized by upregulation of HOX genes and their embedded microRNAs and downregulation of the prognostically adverse MN1, BAALC, and ERG genes.

CONCLUSION

NPM1 mutations have favorable prognostic impact in older patients with CN-AML, especially those age >or= 70 years. The gene- and microRNA-expression profiles suggest that NPM1 mutations constitute a marker defining a biologically homogeneous entity in CN-AML that might be treated with specific and/or targeted therapies across age groups.

Prognostic importance of MN1 transcript levels, and biologic insights from MN1-associated gene and microRNA expression signatures in cytogenetically normal acute myeloid leukemia: a cancer and leukemia group B study

PURPOSE To determine the prognostic importance of the meningioma 1 (MN1) gene expression levels in the context of other predictive molecular markers, and to derive MN1 associated gene- and microRNA-expression profiles in cytogenetically normal acute myeloid leukemia (CN-AML). PATIENTS AND METHODS MN1 expression was measured in 119 untreated primary CN-AML adults younger than 60 years by real-time reverse-transcriptase polymerase chain reaction. Patients were also tested for FLT3, NPM1, CEBPA, and WT1 mutations, MLL partial tandem duplications, and BAALC and ERG expression. Gene- and microRNA-expression profiles were attained by performing genome-wide microarray assays. Patients were intensively treated on two first-line Cancer and Leukemia Group B clinical trials. Results Higher MN1 expression associated with NPM1 wild-type (P < .001), increased BAALC expression (P = .004), and less extramedullary involvement (P = .01). In multivariable analyses, higher MN1 expression associated with a lower complete remission rate (P = .005) after adjustment for WBC; shorter disease-free survival (P = .01) after adjustment for WT1 mutations, FLT3 internal tandem duplications (FLT3-ITD), and high ERG expression; and shorter survival (P = .04) after adjustment for WT1 and NPM1 mutations, FLT3-ITD, and WBC. Gene- and microRNA-expression profiles suggested that high MN1 expressers share features with high BAALC expressers and patients with wild-type NPM1. Higher MN1 expression also appears to be associated with genes and microRNAs that are active in aberrant macrophage/monocytoid function and differentiation. CONCLUSION MN1 expression independently predicts outcome in CN-AML patients. The MN1 gene- and microRNA-expression signatures suggest biologic features that could be exploited as therapeutic targets.

Prognostic significance of, and gene and microRNA expression signatures associated with, CEBPA mutations in cytogenetically normal acute myeloid leukemia with high-risk molecular features: a Cancer and Leukemia Group B Study

PURPOSE

To evaluate the prognostic significance of CEBPA mutations in the context of established molecular markers in cytogenetically normal (CN) acute myeloid leukemia (AML) and gain biologic insights into leukemogenesis of the CN-AML molecular high-risk subset (FLT3 internal tandem duplication [ITD] positive and/or NPM1 wild type) that has a significantly higher incidence of CEBPA mutations than the molecular low-risk subset (FLT3-ITD negative and NPM1 mutated).

PATIENTS AND METHODS

One hundred seventy-five adults age less than 60 years with untreated primary CN-AML were screened before treatment for CEBPA, FLT3, MLL, WT1, and NPM1 mutations and BAALC and ERG expression levels. Gene and microRNA (miRNA) expression profiles were obtained for the CN-AML molecular high-risk patients.

RESULTS

CEBPA mutations predicted better event-free (P = .007), disease-free (P = .014), and overall survival (P < .001) independently of other molecular and clinical prognosticators. Among patients with CEBPA mutations, 91% were in the CN-AML molecular high-risk group. Within this group, CEBPA mutations predicted better event-free (P < .001), disease-free (P = .004), and overall survival (P = .009) independently of other molecular and clinical characteristics and were associated with unique gene and miRNA expression profiles. The major features of these profiles were upregulation of genes (eg, GATA1, ZFPM1, EPOR, and GFI1B) and miRNAs (ie, the miR-181 family) involved in erythroid differentiation and downregulation of homeobox genes.

CONCLUSION

Pretreatment testing for CEBPA mutations identifies CN-AML patients with different outcomes, particularly in the molecular high-risk group, thus improving molecular risk-based classification of this large cytogenetic subset of AML. The gene and miRNA expression profiling provided insights into leukemogenesis of the CN-AML molecular high-risk group, indicating that CEBPA mutations are associated with partial erythroid differentiation.

Gene expression profiling reveals similarities between the in vitro and in vivo responses of immune effector cells to IFN-alpha

PURPOSE

The precise molecular targets of IFN-alpha therapy in the context of malignant melanoma are unknown but seem to involve signal transducers and activators of transcription 1 signal transduction within host immune effector cells. We hypothesized that the in vitro transcriptional response of patient peripheral blood mononuclear cells (PBMC) to IFN-alpha would be similar to the in vivo response to treatment with high-dose IFN-alpha.

EXPERIMENTAL DESIGN

The gene expression profiles of PBMCs and immune cell subsets treated in vitro with IFN-alpha were evaluated, as were PBMCs obtained from melanoma patients receiving adjuvant IFN-alpha.

RESULTS

Twenty-seven genes were up-regulated in PBMCs from normal donors after treatment with IFN-alpha in vitro for 18 hours (>2-fold, P < 0.001). A subset of these genes (in addition to others) was significantly expressed in IFN-alpha-treated T cells, natural killer cells, and monocytes. Analysis of gene expression within PBMCs from melanoma patients (n = 13) receiving high-dose IFN-alpha-2b (20 MU/m(2) i.v.) revealed significant up-regulation (>2-fold) of 21 genes (P < 0.001). Also, the gene expression profile of in vitro IFN-alpha-stimulated patient PBMCs was similar to that of PBMCs obtained from the same patient after IFN-alpha therapy.

CONCLUSIONS

This report is the first to describe the transcriptional response of T cells, natural killer cells, and monocytes to IFN-alpha and characterize the transcriptional profiles of PBMCs from melanoma patients undergoing IFN-alpha immunotherapy. In addition, it was determined that microarray analysis of patient PBMCs after in vitro stimulation with IFN-alpha may be a useful predictor of the in vivo response of immune cells to IFN-alpha immunotherapy.

MicroRNA expression in cytogenetically normal acute myeloid leukemia

BACKGROUND

A role of microRNAs in cancer has recently been recognized. However, little is known about the role of microRNAs in acute myeloid leukemia (AML).

METHODS

Using microRNA expression profiling, we studied samples of leukemia cells from adults under the age of 60 years who had cytogenetically normal AML and high-risk molecular features--that is, an internal tandem duplication in the fms-related tyrosine kinase 3 gene (FLT3-ITD), a wild-type nucleophosmin (NPM1), or both. A microRNA signature that was associated with event-free survival was derived from a training group of 64 patients and tested in a validation group of 55 patients. For the latter, a microRNA compound covariate predictor (called a microRNA summary value) was computed on the basis of weighted levels of the microRNAs forming the outcome signature.

RESULTS

Of 305 microRNA probes, 12 (including 5 representing microRNA-181 family members) were associated with event-free survival in the training group (P<0.005). In the validation group, the microRNA summary value was inversely associated with event-free survival (P=0.03). In multivariable analysis, the microRNA summary value remained associated with event-free survival (P=0.04) after adjustment for the allelic ratio of FLT3-ITD to wild-type FLT3 and for the white-cell count. Using results of gene-expression microarray analysis, we found that expression levels of the microRNA-181 family were inversely correlated with expression levels of predicted target genes encoding proteins involved in pathways of innate immunity mediated by toll-like receptors and interleukin-1beta.

CONCLUSIONS

A microRNA signature in molecularly defined, high-risk, cytogenetically normal AML is associated with the clinical outcome and with target genes encoding proteins involved in specific innate-immunity pathways.

STAT1-dependent and STAT1-independent gene expression in murine immune cells following stimulation with interferon-alpha

PURPOSE

The precise molecular targets of interferon-alpha (IFN-alpha) therapy of melanoma are unknown but likely involve signal transducer and activator of transcription 1 (STAT1) signal transduction within host immune effector cells. We hypothesized that microarray analysis could be utilized to identify candidate molecular targets important for mediating the anti-tumor effect of exogenously administered IFN-alpha.

EXPERIMENTAL METHODS

To identify the STAT1-dependent genes regulated by IFN-alpha, the gene expression profile of splenocytes from wild type (WT) and STAT1(-/-) mice was characterized.

RESULTS

This analysis identified 30 genes that required STAT1 signal transduction for optimal expression in response to IFN-alpha (p < 0.001). These genes include granzyme b (Gzmb), interferon regulatory factor 7 (Irf7), Fas death domain-associated protein (Daxx), and lymphocyte antigen 6 complex, locus C (Ly6c). The expression of 20 genes was found to be suppressed in the presence of STAT1 including chemokine ligand 2 (Ccl2), Ccl5, and Ccl7. Nineteen genes were significantly upregulated in murine splenocytes following treatment with IFN-alpha regardless of the presence of STAT1 including CD86, lymphocyte antigen 6 complex, locus A (Ly6a), and Tap binding protein (Tapbp). The expression of representative IFN-responsive genes was confirmed at the transcriptional level by Real Time PCR.

CONCLUSION

This report is the first to demonstrate that STAT1-mediated signal transduction plays a major role in the transcriptional response of murine immune cells to IFNalpha.

Appropriateness of some resampling-based inference procedures for assessing performance of prognostic classifiers derived from microarray data

The goal of many gene-expression microarray profiling clinical studies is to develop a multivariate classifier to predict patient disease outcome from a gene-expression profile measured on some biological specimen from the patient. Often some preliminary validation of the predictive power of a profile-based classifier is carried out using the same data set that was used to derive the classifier. Techniques such as cross-validation or bootstrapping can be used in this setting to assess predictive power, and if applied correctly, can result in a less biased estimate of predictive accuracy of a classifier. However, some investigators have attempted to apply standard statistical inference procedures to assess the statistical significance of associations between true and cross-validated predicted outcomes. We demonstrate in this paper that naïve application of standard statistical inference procedures to these measures of association under null situations can result in greatly inflated testing type I error rates. Under alternatives of small to moderate associations, confidence interval coverage probabilities may be too low, although for very large associations coverage probabilities approach their intended values. Our results suggest that caution should be exercised in interpreting some of the claims of exceptional prognostic classifier performance that have been reported in prominent biomedical journals in the past few years.

Oxygen limitation modulates pH regulation of catabolism and hydrogenases, multidrug transporters, and envelope composition in Escherichia coli K-12

BACKGROUND

In Escherichia coli, pH regulates genes for amino-acid and sugar catabolism, electron transport, oxidative stress, periplasmic and envelope proteins. Many pH-dependent genes are co-regulated by anaerobiosis, but the overall intersection of pH stress and oxygen limitation has not been investigated.

RESULTS

The pH dependence of gene expression was analyzed in oxygen-limited cultures of E. coli K-12 strain W3110. E. coli K-12 strain W3110 was cultured in closed tubes containing LBK broth buffered at pH 5.7, pH 7.0, and pH 8.5. Affymetrix array hybridization revealed pH-dependent expression of 1,384 genes and 610 intergenic regions. A core group of 251 genes showed pH responses similar to those in a previous study of cultures grown with aeration. The highly acid-induced gene yagU was shown to be required for extreme-acid resistance (survival at pH 2). Acid also up-regulated fimbriae (fimAC), periplasmic chaperones (hdeAB), cyclopropane fatty acid synthase (cfa), and the "constitutive" Na+/H+ antiporter (nhaB). Base up-regulated core genes for maltodextrin transport (lamB, mal), ATP synthase (atp), and DNA repair (recA, mutL). Other genes showed opposite pH responses with or without aeration, for example ETS components (cyo,nuo, sdh) and hydrogenases (hya, hyb, hyc, hyf, hyp). A hypF strain lacking all hydrogenase activity showed loss of extreme-acid resistance. Under oxygen limitation only, acid down-regulated ribosome synthesis (rpl,rpm, rps). Acid up-regulated the catabolism of sugar derivatives whose fermentation minimized acid production (gnd, gnt, srl), and also a cluster of 13 genes in the gadA region. Acid up-regulated drug transporters (mdtEF, mdtL), but down-regulated penicillin-binding proteins (dacACD, mreBC). Intergenic regions containing regulatory sRNAs were up-regulated by acid (ryeA, csrB, gadY, rybC).

CONCLUSION

pH regulates a core set of genes independently of oxygen, including yagU, fimbriae, periplasmic chaperones, and nhaB. Under oxygen limitation, however, pH regulation is reversed for genes encoding electron transport components and hydrogenases. Extreme-acid resistance requires yagU and hydrogenase production. Ribosome synthesis is down-regulated at low pH under oxygen limitation, possibly due to the restricted energy yield of catabolism. Under oxygen limitation, pH regulates metabolism and transport so as to maximize alternative catabolic options while minimizing acidification or alkalinization of the cytoplasm.

Independent confirmation of a prognostic gene-expression signature in adult acute myeloid leukemia with a normal karyotype: a Cancer and Leukemia Group B study

Patients with acute myeloid leukemia (AML) and normal karyotype are classified in an intermediate-risk group, albeit this subset is heterogeneous for clinical outcome. A recent complementary DNA microarray study identified a gene-expression signature that--when used to cluster normal karyotype patients--separated them into 2 prognostically relevant subgroups. We sought the first independent validation of the prognostic value of this signature. Using oligonucleotide microarrays to measure gene expression in samples from uniformly treated adults with karyotypically normal AML, we performed cluster analysis based on the previously identified signature. We also developed a well-defined classification rule using the signature to predict outcome for individual patients. Cluster analysis confirmed the prognostic utility of the signature: patient clusters differed in overall (P = .001) and disease-free (P = .001) survival. The signature-based classifier identified groups with differences in overall (P = .02) and disease-free (P = .05) survival. A strong association of the outcome classifier with the prognostically adverse FLT3 internal tandem duplication (FLT3 ITD) potentially explained the prognostic significance of the signature. However, in the subgroup of patients without FLT3 ITD there was a moderate difference in survival for the classifier-derived groups. Our analysis confirms the applicability of the gene-expression profiling strategy for outcome prediction in cytogenetically normal AML.

Albumin turnover: FcRn-mediated recycling saves as much albumin from degradation as the liver produces

It is now understood that the nonclassical major histocompatibility complex-I molecule FcRn binds albumin and retrieves it from an intracellular degradative fate. Whether FcRn in the liver modulates albumin turnover through effects on biosynthesis and production is not known. Thus we quantified the appearance of biosynthetically labeled albumin in plasma after an intravenous bolus injection of [(3)H]leucine in FcRn-deficient mice. The production rates for both albumin (FcRn substrate) and transferrin (nonsubstrate) are increased by approximately 20% in FcRn-deficient mice compared with normal mice, likely compensating for the lowered plasma oncotic pressure caused by hypoalbuminemia in FcRn-deficient mice. Determining the magnitude of FcRn-mediated effects on albumin turnover, we then measured the steady-state plasma concentrations of biosynthetically labeled albumin and transferrin during [(3)H]leucine infusion. The concentration of albumin was approximately 40% lower in FcRn-deficient mice compared with normal mice. Furthermore, the approximately 40% lower plasma albumin concentration in FcRn-deficient mice along with the approximately 20% increase in albumin production indicate, by the mass-balance equation, that albumin degradation in FcRn-deficient mice is twice that of normal mice. These studies of biosynthetically labeled, and thus native, albumin support our previous finding that FcRn protects albumin from degradation. Permitting quantification of the magnitude of FcRn-mediated recycling, they further indicate that FcRn has extraordinary capacity: the amount of albumin saved from degradation by FcRn-mediated recycling is the same as that produced by the liver.

Overexpression of the ETS-related gene, ERG, predicts a worse outcome in acute myeloid leukemia with normal karyotype: a Cancer and Leukemia Group B study

PURPOSE

To test the prognostic significance of ETS-related gene (ERG) expression in cytogenetically normal primary acute myeloid leukemia (AML).

PATIENTS AND METHODS

Pretreatment blood samples from 84 cytogenetically normal AML patients aged less than 60 years, who were characterized for BAALC expression, FLT3 internal tandem duplication (ITD), and MLL partial tandem duplication (PTD) and uniformly treated on Cancer and Leukemia Group B 9621 protocol, were analyzed for ERG expression by real-time reverse transcriptase polymerase chain reaction. Patients were divided into quartiles according to ERG levels and were compared for clinical outcome. High-density oligonucleotide arrays were used to identify genes differentially expressed between high and low ERG expressers.

RESULTS

With a median follow-up of 5.7 years, patients with the upper 25% of ERG expression values had a worse cumulative incidence of relapse (CIR; P < .001) and overall survival (OS; P = .011) than the remaining patients. In a multivariable analysis, high ERG expression (P < .001) and the presence of MLL PTD (P = .027) predicted worse CIR. With regard to OS, an interaction was observed between expression of ERG and BAALC (P = .013), with ERG overexpression predicting shorter survival only in low BAALC expressers (P = .002). ERG overexpression was an independent prognostic factor even when the unfavorable group of FLT3 ITD patients lacking an FLT3 wild-type allele was included. High ERG expression was associated with upregulation of 112 expressed-sequenced tags and named genes, many of which are involved in cell proliferation, differentiation, and apoptosis.

CONCLUSION

ERG overexpression in AML patients with normal cytogenetics predicts an adverse clinical outcome and seems to be associated with a specific molecular signature.

A paradigm for class prediction using gene expression profiles

We propose a general framework for prediction of predefined tumor classes using gene expression profiles from microarray experiments. The framework consists of 1) evaluating the appropriateness of class prediction for the given data set, 2) selecting the prediction method, 3) performing cross-validated class prediction, and 4) assessing the significance of prediction results by permutation testing. We describe an application of the prediction paradigm to gene expression profiles from human breast cancers, with specimens classified as positive or negative for BRCA1 mutations and also for BRCA2 mutations. In both cases, the accuracy of class prediction was statistically significant when compared to the accuracy of prediction expected by chance. The framework proposed here for the application of class prediction is designed to reduce the occurrence of spurious findings, a legitimate concern for high-dimensional microarray data. The prediction paradigm will serve as a good framework for comparing different prediction methods and may accelerate the development of molecular classifiers that are clinically useful.

Methods for assessing reproducibility of clustering patterns observed in analyses of microarray data

MOTIVATION

Recent technological advances such as cDNA microarray technology have made it possible to simultaneously interrogate thousands of genes in a biological specimen. A cDNA microarray experiment produces a gene expression 'profile'. Often interest lies in discovering novel subgroupings, or 'clusters', of specimens based on their profiles, for example identification of new tumor taxonomies. Cluster analysis techniques such as hierarchical clustering and self-organizing maps have frequently been used for investigating structure in microarray data. However, clustering algorithms always detect clusters, even on random data, and it is easy to misinterpret the results without some objective measure of the reproducibility of the clusters.

RESULTS

We present statistical methods for testing for overall clustering of gene expression profiles, and we define easily interpretable measures of cluster-specific reproducibility that facilitate understanding of the clustering structure. We apply these methods to elucidate structure in cDNA microarray gene expression profiles obtained on melanoma tumors and on prostate specimens.

Design of studies using DNA microarrays

DNA microarrays are assays that simultaneously provide information about expression levels of thousands of genes and are consequently finding wide use in biomedical research. In order to control the many sources of variation and the many opportunities for misanalysis, DNA microarray studies require careful planning. Different studies have different objectives, and important aspects of design and analysis strategy differ for different types of studies. We review several types of objectives of studies using DNA microarrays and address issues such as selection of samples, levels of replication needed, allocation of samples to dyes and arrays, sample size considerations, and analysis strategies.

Graph models of oncogenesis with an application to melanoma

We describe several analytical techniques for use in developing genetic models of oncogenesis including: methods for the selection of important genetic events, construction of graph models (including distance-based trees, branching trees, contingency trees and directed acyclic graph models) from these events and methods for interpretation of the resulting models. The models can be used to make predictions about: which genetic events tend to occur early, which events tend to occur together and the likely order of events. Unlike simple path models of oncogenesis, our models allow dependencies to exist between specific genetic changes and allow for multiple, divergent paths in tumor progression. A variety of genetic events can be used with the graph models including chromosome breaks, losses or gains of large DNA regions, small mutations and changes in methylation. As an application of the techniques, we use a recently published cytogenetic analysis of 206 melanoma cases [Nelson et al. (2000), Cancer Genet. Cytogenet.122, 101-109] to derive graph models for chromosome breaks in melanoma. Among our predictions are: (1) breaks in 6q1 and 1q1 are early events, with 6q1 preferentially occurring first and increasing the probability of a break in 1q1 and (2) breaks in the two sets [1p1, 1p2, 9q1] and [1q1, 7p2, 9p2] tend to occur together. This study illustrates that the application of graph models to genetic data from tumor sets provide new information on the interrelationships among genetic changes during tumor progression.

Waiting times to appearance and dominance of advantageous mutants: estimation based on the likelihood

The germinal center reaction (GCR) of vertebrate immunity provides a remarkable example of evolutionary succession, in which an advantageous phenotype arises as a spontaneous mutation from the parental type and eventually displaces the parental type altogether. In the case of the immune response to the hapten (4-hydroxy-3-nitrophenyl)acetyl (NP), as with several other designed immunogens, the process is dominated by a single key mutation, which greatly simplifies the modeling of and analysis of data. We developed a two-stage model of this process in which the primary stage represents the appearance and establishment of the mutant population as a stochastic process while the second stage represents the growth and dominance of the clone as a deterministic process, conditional on its time of establishment from stage one. We applied this model to the analysis of population samples from several germinal center (GC) reactions and used maximum-likelihood methods to estimate the waiting times to arrival and to dominance of the mutant clone. We determined the sampling properties of the maximum-likelihood estimates using Monte Carlo methods and compared them to their asymptotic distributions. The methods we present here are well-suited for use in the analysis of other systems, such as tumor growth and the experimental evolution of bacteria.

Chromosome abnormalities in malignant melanoma: clinical significance of nonrandom chromosome abnormalities in 206 cases

We report the cytogenetic abnormalities from a series of 206 primary malignant melanoma specimens referred to a single institution. A total of 169 out of 206 unique cases had chromosome breakpoints. A previously described statistical method was used to detect nonrandom distribution of chromosome breakpoints at the level of chromosome regions. Nonrandom occurrence of chromosome breakpoints (indicating that the observed number of breaks significantly exceeded the expected number of breaks) was detected in 28 regions, suggesting a hierarchy of genetic abnormalities in melanoma. Clinical variables and tumor characteristics were analyzed for associations with the presence of any nonrandom chromosome breakpoints; with individual, nonrandomly involved chromosome regions; and with paired, nonrandomly involved chromosome regions. No nonrandomly involved chromosome regions or pairs of regions appeared to significantly affect survival. These results identify recurring, nonrandom chromosome abnormalities in malignant melanoma. These results suggest that recurring, nonrandom chromosome alterations play a key role in the etiology and/or progression of malignant melanoma and identify targets within the genome for molecular genetic studies.

Estimation of tamoxifen's efficacy for preventing the formation and growth of breast tumors

BACKGROUND

Several randomized clinical trials have tested the hypothesis that tamoxifen is effective in preventing breast cancer. The largest such trial, the National Surgical Adjuvant Breast and Bowel Project's Breast Cancer Prevention Trial (BCPT), reported a 49% reduction in risk of invasive breast cancer for the tamoxifen group. However, it is unclear whether the effect of tamoxifen in this trial was mainly due to prevention of newly forming tumors or to treatment of occult disease.

METHODS

We used various tumor growth models (i.e., exponential and Gompertzian [growth limited by tumor size]) and a computer simulation to approximate the percentage of detected tumors that were initiated after study entry. Maximum likelihood techniques were then used to estimate separately the efficacy of tamoxifen in treating occult disease and in preventing the formation and growth of new tumors.

RESULTS

Under the assumptions of most of the growth models, the trial was sufficiently long for substantial numbers of new tumors to form, grow, and be detected during the trial. With the Gompertzian model and all available incidence data from the BCPT, it was estimated that 60% (95% confidence interval [CI] = 40%-80%) fewer new tumors were detected in the tamoxifen group than in the placebo group. Likewise, 35% (95% CI = 6%-63%) fewer occult tumors were detected in the tamoxifen group. With this model, the estimated incidence rate of invasive breast cancer among women in the placebo group of the BCPT was 7.7 (95% CI = 6.6-8.9) per 1000 women per year. Similar results were obtained with three exponential tumor growth models.

CONCLUSIONS

These results support the concept that tamoxifen reduced cancer incidence in the BCPT through both treatment of occult disease and prevention of new tumor formation and growth. However, data from prevention trials may never be sufficient to completely distinguish prevention of new tumor formation from treatment of occult disease.

Predicted and inferred waiting times for key mutations in the germinal centre reaction: evidence for stochasticity in selection

The germinal centre reaction (GCR) is a fundamental component of the immune response to T-dependent antigens, during which the immunoglobulin (Ig) genes of B cells experience somatic hypermutation and selection. A maximum-likelihood method on DNA sequence data from 16 individual germinal centres was used to infer that the waiting time for position 33 key (high-affinity) mutations in the anti-(4-hydroxy-3-nitrophenyl) acetyl (NP) response is 8.3 days. This is in marked contrast to the prediction of a key mutant each generation (waiting time about 1/3 day) obtained from a simple model and parameters available in the literature. This disagreement is resolved in part by the finding that the targeted base occurs in a cold spot for hypermutation, raising the predicted waiting time to 2.3 days, although this value remains significantly lower than that inferred from the sequence data. It is proposed that the remaining disparity is attributable to some further stochastic process in the GCR: many early key mutations arise but fail to 'take root' within the GC, either due to emigration or failure of cognate T cell/B cell interaction. Furthermore, it is argued that the frequency with which position 33 mutations are found in secondary responses to NP indicates the presence of selection after the GCR.