Genome-wide expression analysis of paired diagnosis–relapse samples in ALL indicates involvement of pathways related to DNA replication, cell cycle and DNA repair, independent of immune phenotype


Almost a quarter of pediatric patients with acute lymphoblastic leukemia (ALL) suffer from relapses. The biological mechanisms underlying therapy response and development of relapses have remained unclear. In an attempt to better understand this phenomenon, we have analyzed 41 matched diagnosis–relapse pairs of ALL patients using genome-wide expression arrays (82 arrays) on purified leukemic cells. In roughly half of the patients, very few differences between diagnosis and relapse samples were found (‘stable group’), suggesting that mostly extra-leukemic factors (for example, drug distribution, drug metabolism, compliance) contributed to the relapse. Therefore, we focused our further analysis on 20 sample pairs with clear differences in gene expression (‘skewed group’), reasoning that these would allow us to better study the biological mechanisms underlying relapsed ALL. After finding the differences between diagnosis and relapse pairs in this group, we identified four major gene clusters corresponding to several pathways associated with changes in cell cycle, DNA replication, recombination and repair, as well as B-cell developmental genes. We also identified cancer genes commonly associated with colon carcinomas and ubiquitination to be upregulated in relapsed ALL. Thus, about half of the relapses are due to the selection or emergence of a clone with deregulated expression of genes involved in pathways that regulate B-cell signaling, development, cell cycle, cellular division and replication.


Approximately 25% of children with acute lymphoblastic leukemia (ALL) suffer from relapses.1 After relapse, the overall cure rate remains low, despite intensified chemotherapy and stem cell transplantation. Therefore, there is a strong need to develop novel prognostic factors at relapse and novel therapeutic strategies. To this end, insight into the molecular mechanisms underlying treatment outcome, therapy response and the biology of relapse are an absolute requirement. However, the biological mechanisms of relapse in ALL have remained largely elusive.2

Several mechanisms for leukemia relapse have been proposed. In a subset of patients, the original leukemic cell present at diagnosis is also present at relapse, likely because of insufficient therapy (for example, related to compliance, drug tissue distribution, drug metabolism and liver/kidney function), resistance of the ALL cells or other leukemic cell characteristics. In the remaining patients, a new (or modified) leukemic clone emerges at relapse, either because of the additional oncogenic hits during/after therapy or because of the outgrowth of minor subclones already present at diagnosis (referred to as clonal evolution). This may be caused by upregulation of multidrug resistance and chemotherapeutic detoxifying enzyme systems and finally the acquisition of additional genetic mutation(s) leading to deregulated cell growth.3, 4

Microarray technology for genome-wide expression profiling has been successfully applied to the classification and biology of both precursor-B-ALL and T-ALL and subtypes thereof.8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18 Microarray technology can also provide an unbiased means to identify genes involved in the emergence of resistant leukemic clones.5, 6, 7 For the identification of molecular mechanisms involved in relapse of ALL, pairwise comparison of matched diagnosis and relapses is, in principle, the most straightforward way to identify such genes. Several groups, including ours, have previously conducted such analyses.5, 6, 18 Surprisingly, the number of significantly differentially expressed genes identified was relatively small and changes in expression level between diagnosis and relapse were relatively small and only a handful of such genes were found, making it hard to interpret what biological mechanisms underlie the occurrence of relapse. Furthermore, only a minimal overlap was found between the gene sets identified in these studies. Given the solid efforts conducted before by other laboratories,5, 6, 7 it became apparent that the question of which changes in gene expression are related to relapse in ALL was not an easy one to tackle.

In this study, we therefore set out to perform a carefully controlled analysis on as large a set of samples as possible. We analyzed 41 matched diagnosis–relapse samples from both precursor-B-ALL and T-ALL patients, that is a total of 82 microarrays. We purified all leukemic cells, as in a previous study we discovered that, owing to the relatively low tumor load at relapse (typically 50% or less) compared with diagnosis (typically >80%), purification is required to obtain meaningful results.19 Finally, we applied state-of-the-art bioinformatics tools to process and interpret the data.

Despite these additional efforts, we still ended up with very few genes that were differentially expressed and change in expression was low. Further, analyzing each individual diagnosis–relapse pair, we found that in about half the samples, there was hardly any difference in gene expression between diagnosis and relapse (as if the same leukemic clone emerged). These samples were also most stable in the immunoglobulin (Ig) and T-cell receptor (TCR) gene rearrangement pattern with no or little clonal evolution. The resulting heterogeneity in relapse samples (containing both original and novel leukemic clones) severely hampers statistical tests used to find differential expression between diagnosis and relapse. Therefore, we decided to exclude samples without major changes in gene expression between diagnosis and relapse (hypothesized to be the original clones) as they would not give insights into the biology of the relapse mechanism. We focused our further analysis on 20 paired samples with clear differences in gene expression between matched diagnosis and relapse, reasoning that these would give a more homogeneous sample set on which to study the biological mechanisms underlying relapsed ALL.

Materials and methods


The study group consisted of 41 patients with a diagnosis of precursor-B-ALL or T-ALL treated with consecutive Dutch Childhood Oncology Group protocols. Patient characteristics are described briefly in Table 1 and more extensively in the Supplementary Table 1, which includes WBC, immunophenotype and cytogenetic data. The ALL cases were classified as T-ALL (14 patients) or precursor-B-ALL (27 patients). Precursor-B-ALL was further subclassified as pre-B-ALL (six patients), pro-B-ALL (three patients) or common ALL (18 patients). The only inclusion criterion for this study was availability at a minimum cell number of 107 to allow for sufficient material after purification for RNA extraction and microarray analysis. Matched diagnosis–relapse pair samples were processed and microarrays performed according to the consensus guidelines described for leukemia analyses by three European networks.20

Table 1 Patient characteristics


All precursor-B-ALL samples (both diagnosis and relapse) were purified using CD19-beads and T-ALL using CD7 beads on a Miltenyi autoMACS. Purity of samples always exceeded 90% as measured by flow cytometry, which we previously showed to be a good cutoff to minimize false-positive detection in microarray analyses.19

Microarray analyses

Purified leukemic cells were used to extract RNA, which was converted to cDNA and subsequently to biotinylated cRNA by standard methods as described.21 All cRNA samples were hybridized to Affymetrix HG-U133+2 microarrays (54 675 probe sets). Initially, more than 41 pairs were included, but seven patients needed to be excluded because of the low RNA yield in relapse samples or poor quality of the array (high 5′/3′GAPDH ratios20).

Data preprocessing

Robust multichip analysis22 was used to remove the background, quantile normalize the data across arrays23 and calculate expression values. These values were log2-transformed for further analysis, giving numbers between 0 and 16. Whenever subsets of the original data were used in subsequent analyses (for example, only T-ALLs), these subsets were independently normalized and expression values were recalculated to avoid influencing results by using data not present in the subset. The original and processed data have been deposited in the NCBI Gene Expression Omnibus (GEO; under GEO Series accession number GSE18497.


On the basis of the correlation between samples, the samples were first clustered using hierarchical clustering with complete linkage. To reduce the influence of noise (probe sets that give little signal), probe sets with strong signals were selected in an unbiased way by requiring at least a 4-fold change between minimal and maximal expression over all arrays. Correlation was then calculated only over the resulting 909 probe sets.

Differential expression

We applied significance analysis of microarrays (SAM24) to find differentially expressed probe sets in the robust multichip analysis-preprocessed data, using the paired-sample criterion, the null hypothesis being that the expression difference between diagnosis (D) and relapse (R) for each pair is equal to zero. SAM results in an empirically determined q-value, corresponding to the positive false discovery rate (pFDR). For each probe set, this value can be interpreted as the expected fraction of false positives in the set of probe sets called positive, when this probe set and all probe sets with q-values less or equal are included as a positive.

Patient subset selection

After the initial experiments, we proceeded to select a subset of patients that showed a large, concerted up- or downregulation of genes between D and R, thought to represent the emergence of a new leukemic clone. To do so, we first calculated the D/R expression ratio for each probe set. The number of probe sets that showed at least a 4-fold change between D and R, as well as the number of probe sets that showed the same difference between R and D, was then used to select patients: only 20 patients for whom these numbers differed by more than 100 probe sets in an absolute sense (see the ‘Skew’ column in Table 1) were selected. Note that the 4-fold change is not used here as a measure of significance, but to find probe sets carrying a strong signal, to be used to select a subset of patients informative for identifying genes involved in the biology of relapse. Afterward, the SAM procedure for assessing differential expression is applied on this subset as it was on the original full set.

Clustering of differentially expressed probe sets and patients

The log2(D/R) of probe sets found significantly differentially expressed between D and R samples of the 20 selected patients was clustered, again using hierarchical clustering with complete linkage. The distance measure used was Euclidean distance, which is more applicable for log2-ratios than for correlation. The optimal number of clusters k was then chosen as the number of clusters giving a local minimum in the Davies–Bouldin index,39 a measure of separation between clusters.

PCR analysis of Ig/TCR gene rearrangements

Diagnostic samples were screened for the presence of IGH, IGK-Kde, TCRG, TCRD, Vδ2-Jα and TCRB rearrangements using PCR heteroduplex analysis.25, 26, 27, 28, 29 Clonal rearrangements were subsequently analyzed in the paired relapse sample of the patient by mixed PCR heteroduplex analysis.30 In all patients who could be evaluated for Ig/TCR gene rearrangements, at least one rearrangement was preserved, indicating that the relapse did not result from de novo leukemia development. If single rearrangements differed between diagnosis and relapse, we assumed that some clonal evolution took place. Finally, when the changes concerned more than 50% of Ig/TCR gene rearrangements, this was considered unstable, as previously described.31


In an effort to unravel the biological mechanisms of relapsed ALL, we selected 41 patients from whom sufficient cell material was available at both D and R to purify tumor cells, extract RNA and be able to perform DNA microarray-based gene-expression profiling. The data of 82 microarrays (14 T-ALL and 27 precursor-B-ALL) were normalized, summarized into probe set expression values and clustered (Figure 1). As expected, precursor-B-ALL and T-ALL perfectly cluster in separate branches. For all but two patients, diagnosis and relapse samples cluster together, indicating high intra-patient similarity (likely because of the uniform purification step we use) and fairly large inter-patient variation.

Figure 1

Hierarchical clustering of all 41 diagnosis–relapse sample pairs along 909 probe sets showing high signal. Gene expression profiles of the full set of all patients were clustered and are depicted as a heat map.

Using SAM analysis, we identified probe sets significantly differentially expressed between diagnosis and relapse (Table 2). The results for this paired analysis (q<0.05) indicate that the precursor-B-ALL subsets are more homogeneous in the difference between diagnosis and relapse. The precursor-B-ALL subsets (pro-B-ALL, common ALL and pre-B-ALL) were too small to find statistically significant differences in expression within each subset. The T-ALL set is apparently more heterogeneous than the precursor-B-ALL set, which in turn influences the number of probe sets found to be differentially expressed in the overall ALL set. For the probe sets found to be differentially expressed, the absolute mean fold change between diagnosis and relapse was minimal (at most 2-fold, on average between 1.3- and 1.5-fold).

Table 2 SAM analysis

As the aim of this study was to gain better understanding of molecular mechanisms responsible for relapses, this small number of genes with minimal differential expression was not informative. We therefore looked for other ways to find gene signatures that would give insight into the mechanisms underlying relapse and therapy resistance. We hypothesized that our patient set actually consisted of (at least) two types of relapses: ‘true relapses’ of the identical leukemic clone and ‘modified relapses,’ in which a novel (sub)clone of the original ALL emerges. Analyzing all cases together seriously impedes finding differentially expressed genes, as we are looking for differences between diagnosis and relapse that are consistent over the entire set of samples. As we are mostly interested in the biology underlying the novel leukemic clones, we tried to discriminate between the two types on the basis of gene-expression data and Ig/TCR gene rearrangement patterns, that is between ‘true relapses’ and clonal evolution, thereby facilitating the discovery of genes that underlie the biology of relapse in ALL. In 21 cases, only minimal differences in gene expression were observed (Table 1), whereas in the remaining 20 patients clear differences between diagnosis and relapse were found. These 21 cases also showed the most stable Ig/TCR gene rearrangement pattern (stable group), with all unstable cases and many of the cases with clonal evolution residing in the novel clone group. Interestingly, a comparison of these two groups of patients at diagnosis indicated that there were two large groups of genes differentially expressed (SAM, q<0.05; Figure 2). One group of 2564 genes, more highly expressed in the novel clone patients, contained genes involved in PXR/RXR activation (such as ABC transporters and MDR genes), metabolism of xenobiotics by cytochrome P450, linoleic acid metabolism and FXR/RXR activation. Apparently, these leukemias constitutively expressed somewhat higher levels of drug metabolism genes than did the other group and may have ‘allowed’ selection of other genetic aberrations that develop during the development of relapse as well as outgrowth of subclones hidden in the original bulk tumor. The second large group, more highly expressed in the stable group, contained 3481 genes involved in cancer and cell growth.

Figure 2

Comparison of patient samples thought to originate from the original leukemic clone (stable, removed) and from a novel (sub)clone, (skewed, retained for further analysis), based on differential gene expression at diagnosis between the two groups: 20 retained (top) vs 21 removed (bottom). The retained samples show higher initial expression (at diagnosis) of genes involved in xenobiotic metabolism.

For the 20 novel clone patients, we again used SAM to find differentially expressed probe sets between diagnosis and relapse. At q<0.05, 2663 significant probe sets were found in all ALL, clearly more than when all pairs were left in. Next, we clustered the log2(D/R) values for these 2663 probe sets, that is the log-fold changes per patient, and clustered the probe sets. The optimal number of probe set clusters assessed by the Davies–Bouldin index is 11; although a clustering into three clusters gives a lower index, the resulting clusters are highly uninformative (Supplementary Figure 1). Clustering results are shown in Figure 3. Seven of the 11 clusters are very small (less than 10 probe sets) and likely contain outliers. The larger clusters contain 1492, 692, 492 and 21 probe sets, respectively (Table 3). Interestingly, these four clusters do not cluster along patients according to immunophenotype: both precursor-B-ALL and T-ALL can be found in all four clusters, indicating that the differences between D and R are independent of immunophenotype.

Figure 3

Clustering of 2663 differentially expressed probe sets between diagnosis and relapse in the skewed group.

Table 3 Clusters identified by hierarchical clustering of the 2663 probe sets differentially expressed between D and R in the subset of 20 patients (skewed group)

The four larger clusters identified were further examined by Ingenuity pathway analysis. All the genes contained in these clusters are provided in Supplementary Table 2 (Supplementary Tables). Each of these represents overlapping but different functionally related genes and pathways:

  • Cluster 2, the largest cluster, contains many genes involved in cell cycle regulation, DNA replication and repair, including BRCA2, CDC kinases and RAD51, as well as genes involved in cell cycle checkpoints and cancer genes.

  • Cluster 3 involves genes involved in molecular transport, cellular assembly and protein trafficking, as well as genes in ubiquitination and genes involved in B-lymphocyte signaling and pre-BCR, involved in normal B-cell development (Figure 4 and Supplementary Figure 2 containing the same data including log-fold changes), as well as several other B-cell signaling genes, such as the small GTPase adapter protein RAC1.

    Figure 4

    Ingenuity analysis of differentially expressed genes related to the B-cell signaling pathway.

  • Cluster 4 involves cancer and cell cycle genes, as well as very prominent metabolic genes involved in amino acid and purine metabolism, hallmarks of highly proliferating cells. Interestingly, in both clusters 2 and 4, many genes were found that are typically associated with colon carcinoma, but which are expressed in hematopoietic tissues as well. Examples include CCNB2, LIMS1, TEGT, CACYBP, GNB1, WWOX and PPIH. For colon carcinoma, aberrations in Wnt signaling are causative factors that have also been proposed and are increasingly seen in hematological malignancies (reviewed in Refs. 32, 33).

  • Cluster 5 contains only 21 genes mainly involved in cell cycle regulation (14 of 21 genes), including TOP2A (3-fold upregulated in relapsed samples), encoding a DNA topoisomerase, that controls and alters the topologic states of DNA during transcription and is often upregulated in a wide variety of cancers; and KIF 11 (2.1-fold up), a kinesin family member 11 motor protein that controls spindle dynamics. The function of this gene product includes chromosome positioning, centrosome separation and establishing a bipolar spindle during cell mitosis (Figure 5 and Supplementary Figure 3 containing the same data including log-fold changes).

    Figure 5

    Ingenuity analysis of differentially expressed genes involved in the cell cycle checkpoint.

Finally, we investigated whether timing of relapse is associated with different gene expression profiles. To this end we performed an analysis similar to the one by Kirschner-Schwabe et al.34 and found a few probe sets significantly differentially expressed between relapse samples of late and early relapsed patients. We used the following two subdivisions: early (30 months after diagnosis, 15 patients) vs late (>30 months, five patients) relapse and very early (12 months after diagnosis, eight patients) vs late (>30 months, five patients) relapse. The first analysis resulted in only three significantly differentially expressed probe sets, namely the Zn finger DNA binding factor ZDHHC17, a ribosomal protein and an unknown EST (Supplementary Table 3). The second comparison resulted in 10 probe sets, including two Ring finger-containing proteins and genes associated with non-hematological cancers (colon carcinoma, meningioma; Supplementary Table 3).

Another related question is whether at diagnosis a late vs an early relapse can be predicted by gene expression profiling. By setting the same threshold of the survival time at 30 months as above, a binary problem is created, comparing the five late relapse cases with the 15 early ones. SAM analysis (q<0.05) results in 417 probe sets that are predictive of a late relapse (Supplementary Table 4). Interestingly, a relatively large proportion of these genes is involved in mitochondrial function (20%) and oxidative phosphorylation (16%) or ubiquitination (12%) as found by the Ingenuity pathway analysis.


A large number of investigators have performed microarray studies on various types of ALL to define novel subgroups, come up with novel risk classification or predict therapy response.2, 5, 6, 7, 13, 35, 36, 37 Relatively few studies have focused on relapse samples and we are aware of only three previous studies using matched diagnosis–relapse pairs to better understand the process of relapsed ALL.5, 6, 18 One of these studies was a small pilot study we performed a few years ago using seven patient pairs;18 the other two are larger studies, one Australian (Beesley et al.5) and the other American (Bhojwani et al.6), who studied 11 and 35 matched diagnosis–relapse samples, respectively, and additional diagnosis samples. Although both seminal studies were performed carefully, they found only small numbers of genes to be differentially expressed between diagnosis and relapse samples, in general with changes that were relatively small (1.5- to 2-fold per gene), which are not easily interpreted in terms of mechanisms underlying the relapse process. Beesley et al. found a maximum change of 2.3-fold in expression for GRP58 (glucose-regulated protein of 58 kD), whereas Bjowani et al. found only changes smaller than 2-fold. Both studies performed extensive additional analyses on the pathways involved and studied whether signatures could be defined predicting early vs late relapse (an analysis we also performed), but with poor results: a small number of genes whose biological function makes them unlikely to have a major impact on these genes in survival. Thus, despite these efforts, the understanding of the relapse process remained unsatisfactory.

In our current study, we set out to improve on these studies by increasing the sample size, purifying all leukemic samples (>90%) and using tried and tested bioinformatic tools to allow us to identify genes relevant for understanding the mechanism leading to the emergence of relapses. In all, we started with 41 pairs of matched D–R samples. Our initial analyses again showed only genes with small differential expression between diagnosis and relapse, hardly explaining the biological mechanism underlying ALL relapse.

We then realized that in some cases relapse may result from therapy-induced selection of minor clones present at diagnosis, rather than from genetic adaptation of the original tumor cells. Gene expression profiling cannot clearly distinguish between these cases, and even an analysis of Ig/TCR gene rearrangements can only give partial insights into the extent of clonal evolution. Closely examining each D/R pair, we discovered that about half of the pairs showed very little change in gene expression, whereas in the other half more changes occurred. We therefore devised a ‘skew’ measure for concerted up- or downregulation of gene expression between D and R (Table 1). D–R pairs for which this measure is low may represent the very same leukemic clone that escaped therapy because of limited compliance, too low a dosage of the drugs or rapid drug catabolism in the individual patients (for example, caused by differential liver and kidney function between patients). These patients are not very informative in understanding the leukemic cell-related mechanisms underlying ALL relapse, and in fact hamper the interpretation of the overall set. Patients with a large skew measure, however, may correspond to novel leukemic clones, and represent a possible source to understand what is going on at relapse. We were left with 20 patients (40 arrays), which clearly limited our sample size and hence the possibilities of finding differences, in spite of our initial efforts to study a large group. This is further exacerbated by the fact that between-patient variation is larger than variation between diagnosis and relapse, as almost all D–R samples cluster together as patient pairs.

Nevertheless, we identified four different major clusters of pathways with linked functional annotations, which we could not have identified without focusing solely on patients with novel clones (‘skewed’ patients). Interestingly, all four clusters were found in both T-ALL and precursor-B-ALL and all subtypes of precursor-B-ALL were represented in the different clusters. This indicates that the mechanism governing relapses is independent of the lineage and immunophenotype of the leukemia involved. The clusters found not only involved cycle regulation, DNA replication and repair, but also genes involved in ubiquitination and in B-lymphocyte signaling, and the pre-BCR complex, involved in normal B-cell development.

We investigated to what extent our gene list overlapped with those in previously published studies. With Beesley et al.,5 probe sets are shared, corresponding to four distinct genes (MPO, SRRT, CRKL and WHSC1). With Bhojwani et al., no probe sets overlapped.

While we were concluding this study and analyzing the data, an elegant study by the group of Downing et al. was published.38 These investigators did not use genome-wide gene expression profiling, as we did, but whole genome copy number analysis using SNP microarrays on paired D–R samples that were purified as well. Despite different technologies used, similar conclusions were reached, namely, that in about half of their cohort clear clonal evolution occurred, which could not be attributed to a few genetic alterations but involved multiple pathways. Similar to our results, the authors found genes involved in cell cycle regulation, proliferation and normal development and B-cell signaling and activation.

Thus, on the basis of the study reported here and the results of Mullighan et al.,38 we conclude that the processes underlying treatment outcome, therapy response and the biology of relapse are complex, heterogeneous between patients and involve multiple pathways. Clearly, microarray-based analysis is only one of the methods to improve our insight into these mechanisms. SNP analyses, genome-wide identification of miRNAs and analyses of methylation status at the genome level are all techniques that will contribute to a better understanding of the complex mechanism that operates in relapsed ALL. Nevertheless, the pathways and genes identified may help unravel the mechanisms of disease progression or even provide prognostic markers, and a number of them may represent novel drug targets to improve therapy outcome in those ALL patients refractory to current treatment protocols. It is intriguing to note that our attempts to identify genes involved in a late or an early relapse identified clearly distinct pathways (mitochondrial function, ubiquitination). Although the number of such cases in our current cohort is relatively small (but the pathways identified are statistically significant), such an analysis on a larger set of patients may confirm our identification of pathways amenable to drug target screenings, thereby opening up possibilities for novel treatment regimens in refractory patients.

Conflict of interest

The authors declare no conflict of interest.


  1. 1

    Pieters R, Carroll WL . Biology and treatment of acute lymphoblastic leukemia. Pediatr Clin North Am 2008; 55: 1–20, ix.

    Article  Google Scholar 

  2. 2

    Carroll WL, Bhojwani D, Min DJ, Moskowitz N, Raetz EA . Childhood acute lymphoblastic leukemia in the age of genomics. Pediatr Blood Cancer 2006; 46: 570–578.

    Article  Google Scholar 

  3. 3

    Gaynon PS . Childhood acute lymphoblastic leukaemia and relapse. Br J Haematol 2005; 131: 579–587.

    Article  Google Scholar 

  4. 4

    Henderson MJ, Choi S, Beesley AH, Sutton R, Venn NC, Marshall GM et al. Mechanism of relapse in pediatric acute lymphoblastic leukemia. Cell Cycle 2008; 7: 1315–1320.

    CAS  Article  Google Scholar 

  5. 5

    Beesley AH, Cummings AJ, Freitas JR, Hoffmann K, Firth MJ, Ford J et al. The gene expression signature of relapse in paediatric acute lymphoblastic leukaemia: implications for mechanisms of therapy failure. Br J Haematol 2005; 131: 447–456.

    CAS  Article  Google Scholar 

  6. 6

    Bhojwani D, Kang H, Moskowitz NP, Min DJ, Lee H, Potter JW et al. Biologic pathways associated with relapse in childhood acute lymphoblastic leukemia: a Children's Oncology Group study. Blood 2006; 108: 711–717.

    CAS  Article  Google Scholar 

  7. 7

    Choi S, Henderson MJ, Kwan E, Beesley AH, Sutton R, Bahar AY et al. Relapse in children with acute lymphoblastic leukemia involving selection of a preexisting drug-resistant subclone. Blood 2007; 110: 632–639.

    CAS  Article  Google Scholar 

  8. 8

    Ferrando AA, Look AT . Gene expression profiling in T-cell acute lymphoblastic leukemia. Semin Hematol 2003; 40: 274–280.

    CAS  Article  Google Scholar 

  9. 9

    Ferrando AA, Look AT . DNA microarrays in the diagnosis and management of acute lymphoblastic leukemia. Int J Hematol 2004; 80: 395–400.

    CAS  Article  Google Scholar 

  10. 10

    Hoffmann K, Firth MJ, Beesley AH, de Klerk NH, Kees UR . Translating microarray data for diagnostic testing in childhood leukaemia. BMC Cancer 2006; 6: 229.

    Article  Google Scholar 

  11. 11

    Holleman A, Cheok MH, den Boer ML, Yang W, Veerman AJ, Kazemier KM et al. Gene-expression patterns in drug-resistant acute lymphoblastic leukemia cells and response to treatment. N Engl J Med 2004; 351: 533–542.

    CAS  Article  Google Scholar 

  12. 12

    Palomero T, McKenna K, J ON, Galinsky I, Stone R, Suzukawa K et al. Activating mutations in NOTCH1 in acute myeloid leukemia and lineage switch leukemias. Leukemia 2006; 20: 1963–1966.

    CAS  Article  Google Scholar 

  13. 13

    Raetz EA, Perkins SL, Bhojwani D, Smock K, Philip M, Carroll WL et al. Gene expression profiling reveals intrinsic differences between T-cell acute lymphoblastic leukemia and T-cell lymphoblastic lymphoma. Pediatr Blood Cancer 2006; 47: 130–140.

    Article  Google Scholar 

  14. 14

    Yeoh EJ, Ross ME, Shurtleff SA, Williams WK, Patel D, Mahfouz R et al. Classification, subtype discovery, and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling. Cancer Cell 2002; 1: 133–143.

    CAS  Article  Google Scholar 

  15. 15

    Armstrong SA, Staunton JE, Silverman LB, Pieters R, den Boer ML, Minden MD et al. MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia. Nat Genet 2002; 30: 41–47.

    CAS  Article  Google Scholar 

  16. 16

    Ross ME, Mahfouz R, Onciu M, Liu HC, Zhou X, Song G et al. Gene expression profiling of pediatric acute myelogenous leukemia. Blood 2004; 104: 3679–3687.

    CAS  Article  Google Scholar 

  17. 17

    Ross ME, Zhou X, Song G, Shurtleff SA, Girtman K, Williams WK et al. Classification of pediatric acute lymphoblastic leukemia by gene expression profiling. Blood 2003; 102: 2951–2959.

    CAS  Article  Google Scholar 

  18. 18

    Staal FJ, van der Burg M, Wessels LF, Barendregt BH, Baert MR, van den Burg CM et al. DNA microarrays for comparison of gene expression profiles between diagnosis and relapse in precursor-B acute lymphoblastic leukemia: choice of technique and purification influence the identification of potential diagnostic markers. Leukemia 2003; 17: 1324–1332.

    CAS  Article  Google Scholar 

  19. 19

    de Ridder D, van der Linden CE, Schonewille T, Dik WA, Reinders MJ, van Dongen JJ et al. Purity for clarity: the need for purification of tumor cells in DNA microarray studies. Leukemia 2005; 19: 618–627.

    CAS  Article  Google Scholar 

  20. 20

    Staal FJ, Cario G, Cazzaniga G, Haferlach T, Heuser M, Hofmann WK et al. Consensus guidelines for microarray gene expression analyses in leukemia from three European leukemia networks. Leukemia 2006; 20: 1385–1392.

    CAS  Article  Google Scholar 

  21. 21

    Staal FJ, Weerkamp F, Baert MR, van den Burg CM, van Noort M, de Haas EF et al. Wnt target genes identified by DNA microarrays in immature CD34+ thymocytes regulate proliferation and cell adhesion. J Immunol 2004; 172: 1099–1108.

    CAS  Article  Google Scholar 

  22. 22

    Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U et al. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 2003; 4: 249–264.

    Article  Google Scholar 

  23. 23

    Bolstad BM, Irizarry RA, Astrand M, Speed TP . A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 2003; 19: 185–193.

    CAS  Article  Google Scholar 

  24. 24

    Tusher VG, Tibshirani R, Chu G . Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci USA 2001; 98: 5116–5121.

    CAS  Article  Google Scholar 

  25. 25

    Verhagen OJ, Willemse MJ, Breunis WB, Wijkhuijs AJ, Jacobs DC, Joosten SA et al. Application of germline IGH probes in real-time quantitative PCR for the detection of minimal residual disease in acute lymphoblastic leukemia. Leukemia 2000; 14: 1426–1435.

    CAS  Article  Google Scholar 

  26. 26

    Pongers-Willemse MJ, Seriu T, Stolz F, d’Aniello E, Gameiro P, Pisa P et al. Primers and protocols for standardized detection of minimal residual disease in acute lymphoblastic leukemia using immunoglobulin and T cell receptor gene rearrangements and TAL1 deletions as PCR targets: report of the BIOMED-1 CONCERTED ACTION: investigation of minimal residual disease in acute leukemia. Leukemia 1999; 13: 110–118.

    CAS  Article  Google Scholar 

  27. 27

    Szczepanski T, van der Velden VHJ, Hoogeveen PG, De Bie M, Jacobs DCH, Van Wering ER et al. V{delta}2-J{alpha} gene rearrangements are frequent in precursor-B-acute lymphoblastic leukemia but rare in normal lymphoid cells. Blood 2004; 103: 3798–3804.

    CAS  Article  Google Scholar 

  28. 28

    Szczepanski T, Willemse MJ, van Wering ER, van Weerden JF, Kamps WA, van Dongen JJM . Precursor-B-ALL with D(H)-J(H) gene rearrangements have an immature immunogenotype with a high frequency of oligoclonality and hyperdiploidy of chromosome 14. Leukemia 2001; 15: 1415–1423.

    CAS  Article  Google Scholar 

  29. 29

    van Dongen JJM, Langerak AW, Bruggemann M, Evans PAS, Hummel M, Lavender FL et al. Design and standardization of PCR primers and protocols for detection of clonal immunogloulin and T-cell receptor gene recombinations in suspect lymphoproliferations. Leukemia 2003; 17: 2257–2317.

    CAS  Article  Google Scholar 

  30. 30

    Szczepanski T, van der Velden VH, Raff T, Jacobs DC, van Wering ER, Bruggemann M et al. Comparative analysis of T-cell receptor gene rearrangements at diagnosis and relapse of T-cell acute lymphoblastic leukemia (T-ALL) shows high stability of clonal markers for monitoring of minimal residual disease and reveals the occurrence of second T-ALL. Leukemia 2003; 17: 2149–2156.

    CAS  Article  Google Scholar 

  31. 31

    Szczepanski T, Willemse MJ, Brinkhof B, van Wering ER, van der Burg M, van Dongen JJ . Comparative analysis of Ig and TCR gene rearrangements at diagnosis and at relapse of childhood precursor-B-ALL provides improved strategies for selection of stable PCR targets for monitoring of minimal residual disease. Blood 2002; 99: 2315–2323.

    CAS  Article  Google Scholar 

  32. 32

    Staal FJ, Luis TC, Tiemessen MM . WNT signalling in the immune system: WNT is spreading its wings. Nat Rev 2008; 8: 581–593.

    CAS  Google Scholar 

  33. 33

    Weerkamp F, van Dongen JJ, Staal FJ . Notch and Wnt signaling in T-lymphocyte development and acute lymphoblastic leukemia. Leukemia 2006; 20: 1197–1205.

    CAS  Article  Google Scholar 

  34. 34

    Kirschner-Schwabe R, Lottaz C, Todling J, Rhein P, Karawajew L, Eckert C et al. Expression of late cell cycle genes and an increased proliferative capacity characterize very early relapse of childhood acute lymphoblastic leukemia. Clin Cancer Res 2006; 12: 4553–4561.

    CAS  Article  Google Scholar 

  35. 35

    Beesley AH, Weller RE, Kees UR . The role of BSG (CD147) in acute lymphoblastic leukaemia and relapse. Br J Haematol 2008; 142: 1000–1002.

    CAS  Article  Google Scholar 

  36. 36

    Boag JM, Beesley AH, Firth MJ, Freitas JR, Ford J, Hoffmann K et al. Altered glucose metabolism in childhood pre-B acute lymphoblastic leukaemia. Leukemia 2006; 20: 1731–1737.

    CAS  Article  Google Scholar 

  37. 37

    Hoffmann K, Firth MJ, Beesley AH, Freitas JR, Ford J, Senanayake S et al. Prediction of relapse in paediatric pre-B acute lymphoblastic leukaemia using a three-gene risk index. Br J Haematol 2008; 140: 656–664.

    CAS  Article  Google Scholar 

  38. 38

    Mullighan CG, Phillips LA, Su X, Ma J, Miller CB, Shurtleff SA et al. Genomic analysis of the clonal origins of relapsed acute lymphoblastic leukemia. Science (New York, NY) 2008; 322: 1377–1380.

    CAS  Article  Google Scholar 

  39. 39

    Davies DL, Bouldin D . A cluster separation measure. IEEE Trans Pattern Anal Mach Intell 1979; 1: 224–227.

    CAS  Article  Google Scholar 

Download references


We thank Patricia Hoogeveen for technical assistance and Dr Ton Langerak for stimulating discussions. We also thank the Dutch Childhood Oncology Group (head: Dr Valerie de Haas) for help in providing samples and patient data.

Author information



Corresponding author

Correspondence to F J T Staal.

Additional information

Supplementary Information accompanies the paper on the Leukemia website (

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Staal, F., de Ridder, D., Szczepanski, T. et al. Genome-wide expression analysis of paired diagnosis–relapse samples in ALL indicates involvement of pathways related to DNA replication, cell cycle and DNA repair, independent of immune phenotype. Leukemia 24, 491–499 (2010).

Download citation


  • ALL
  • relapse
  • microarray

Further reading