High PDGFRA expression does not serve as an effective therapeutic target in ERG-deleted B-cell precursor acute lymphoblastic leukemia

Several oncogenic aberrations of receptor tyrosine kinases have been reported in B-cell precursor acute lymphoblastic leukemia (BCP-ALL). Herein we describe a subgroup of pediatric patients lacking common chromosomal aberrations (“B-other”) but with high expression of platelet-derived growth factor receptor alpha (PDGFRA). Oncogenic fusions involving the tyrosine kinase domain of PDGFRA are recurrent in eosinophilic leukemia and are responsive to kinase inhibitors.1 One case with a FIP1L1-PDGFRA fusion was reported in adult BCP-ALL.2 We, therefore, tested for genomic aberrations underlying the overexpression of PDGFRA and evaluated the inhibition of PDGFRα as a treatment option in pediatric BCP-ALL. We found that high PDGFRA expression was not caused by structural aberrations of PDGFRA, was associated with ERG deletions, and that the in vitro and ex vivo response to PDGFRα inhibitors was limited in Nalm6 and primary BCP-ALL cells, respectively.

Genetic aberrations that define distinct subtypes are essential for diagnosis and risk evaluation of ALL. Gene expression microarray analysis of 654 pediatric ALL cases (data deposited in GSE87070) identified high PDGFRA expression (2log expression probe-set 203131_at >5.9) in 26 of 574 BCP-ALL cases (4.5%). High PDGFRA expression was found exclusively in cases belonging to the “B-other” (n=22) and BCR-ABL1-like (n=4) subtypes, and was not present in ETV6-RUNX1, high hyperdiploid (51–65 chromosomes), TCF3-PBX1, BCR-ABL1 and MLL-rearranged BCP-ALL subtypes or T-ALL. (Figure 1A). PGFRA expression was confirmed by quantitative reverse transcriptase polymerase chain reaction analysis in 23 of 25 tested cases, was maintained in xenograft-derived cells of a PDGFRA-high case, and also found in the BCP-ALL cell line Nalm6 (Online Supplementary Figure S1).

Figure 1.Genomics of PDGFRA in pediatric B-cell precursor acute lymphoblastic leukemia. (A) Gene expression of PDGFRA in the genetic subtypes of pediatric BCP-ALL. Expression was measured on Affymetrix U133 plus 2.0 GeneChips using two different probe sets. (B) Schematic overview and representative results of interphase FISH probing the FIP1L1-PDGFRA locus in one case with high and one case with low PDGFRA expression. Red: probe spanning the CHIC2 gene; green: probes upstream of FIP1L1 and downstream of PDGFRA, respectively. Merge: overlay of red and green signals, blue: DNA stained by DAPI. (C) Targeted locus amplification performed from two viewpoints within the PDGFRA locus. Bars represent the mean coverage over a 10 kb sliding window plotted on the y-axis, while chromosomal position is indicated on the x-axis. Coverage around the viewpoint was capped for data visualization.⁷ Fourteen cases with high PDGFRA, one case with low PDGFRA, and two cell lines were tested. In case of translocations or interstitial deletions (e.g. causing the FIP1L1-PDGFRA fusion gene in EOL1), sequencing coverage drops steeply and rises again on a distant site. Samples marked by roman numerals (patients I to V) were used in ex vivo cytotoxicity assays in Figure 2. (D) Frequencies of secondary aberrations detected by means of multiplex ligation-dependent probe amplification and array comparative genomic hybridization in “B-other” patients with high or low PDGFRA expression. chr21 gain: partial or complete gain of chromosome 21, including germline trisomy. High CRLF2: high expression of CRLF2 by gene expression array as described above. Frequencies were compared using the Fisher exact test. *: P<0.05, **: P<0.01, ***: P<0.001. For details see also Online Supplementary Methods.

Since high PDGFRA expression was mainly found in the “B-other” subtype, we analyzed clinical characteristics within this group. Age, white blood cell count, gender, risk stratification arm, and the frequency of relapse, death, and non-response (together defining “events”) were not significantly different between PDGFRA-high and -low “B-other” cases (Table 1). However, cells from patients with high PDGFRA expression levels were more sensitive to prednisolone than were cells from patients with low PDGFRA expression levels, as measured with an ex vivo MTT assay previously described3 (P=0.03, see also Online Supplementary Figure S2). Ex vivo sensitivity towards L-asparaginase or vincristine was not different.

Table 1.Clinical characteristics of patients with “B-other“ B-cell precursor acute lymphoblastic leukemia with high PDGFRA expression.

Oncogenic fusions involving the tyrosine kinase domain of PDGFRA are recurrent in adult myeloproliferative neoplasms, but have not been reported in pediatric BCP-ALL. Using fusion gene-specific reverse transcriptase polymerase chain reactions, we did not detect any of the six PDGFRA fusions reported to date, which involve FIP1L1, STRN, BCR, CDK5RAP2, ETV6, and KIF5B (see Online Supplementary Table S1 and Online Supplementary Methods for the primers).741

Novel translocations or other structural aberrations affecting the PDGFRA locus may be responsible for high PDGFRA expression. We performed fluorescence in situ hybridization (FISH; Cytocell, Cambridge, UK) with probes located upstream of FIP1L1, between FIP1L1 and PDGFRA, and downstream of PDGFRA (Figure 1B). The cell line EOL1 served as a positive control, revealing one wildtype allele and two alleles with an interstitial deletion as reported (DSMZ website July 2017, Online Supplementary Figure S3). None of the 36 tested primary BCP-ALL samples and cell lines with high or low PDGFRA expression showed interstitial deletions between FIP1L1 and PDGFRA or split FISH signals (Figure 1B and Online Supplementary Figure S3).

We studied the PDGFRA locus at high resolution by targeted locus amplification in 17 samples (Cergentis BV, Utrecht, the Netherlands). This technique allows amplification and sequencing of genomic regions that are in close proximity to a region of interest and is described in more detail in the Online Supplementary Methods. Two viewpoints were selected in intron 2 and exon 15 of PDGFRA. In EOL1, we confirmed an interstitial deletion on chromosome 4 which rearranges FIP1L1 intron 12 to PDGFRA exon 12 (Figure 1C). The coverage pattern and sequence information obtained from Nalm6 and primary BCP-ALL samples did not indicate genomic aberrations in PDGFRA.

We compared recurrent copy number aberrations, detected by multiplex ligation-dependent probe amplification and array comparative genomic hybridization (as reported by Boer et al.8), between PDGFRA-high and -low “B-other” BCP-ALL cases (Figure 1D and Online Supplementary Table S2). The frequency of deletions in IKZF1, EBF1, RB1, BTG1, and TCF3 did not differ significantly between “B-other” patients with high or low PDGFRA expression. PAX5 and CDKN2A and/or B deletions were rare in patients with high PDGFRA (P=0.01 and P=0.03, respectively), and no loss of the entire chromosome-arm 9p was observed (P=0.02). No ETV6 deletion (P=0.003), high CRLF2 expression (P=0.04), or chromosome 21 gain (P=0.006) was found in PDGFRA-high patients. In contrast, 63% of patients with high PDGFRA expression carried ERG deletions (10/16), compared with 0% of PDGFRA-low “B-other” cases (0/71, P<0.001). Vice versa, all ERG-deleted cases expressed high levels of PDGFRA (Online Supplementary Figure S4). A distinct gene expression profile (referred to as R6 cluster) was described as being characteristic for, but not limited to, ERG-deleted cases.9 Recently, Lilljebjörn et al. reported strong overlap of DUX4 rearrangements with ERG deletions and provided evidence that the DUX4 rearrangement is causal for this exclusive gene expression profile.10 In this newly described BCP-ALL subtype, a truncated copy of the usually silenced germline transcription factor DUX4 is inserted into an actively transcribed region and aberrantly expressed.1110 Zhang et al. showed that aberrant DUX4 induces the expression of an alternative isoform of ERG, thereby directly linking the function of these two transcription factors.12 Importantly, not all DUX4-rearranged cases carried ERG deletions, and they can arise at various points during leukemogenesis of DUX4-rearranged leukemia.1312 The identification of DUX4 rearrangements at a genomic level is challenging because of varying integration sites, many repeats in the genome, and high GC content. We were unable to confirm DUX4 rearrangements in cases with high PDGFRA expression, but strong evidence supports our conclusion: Yasuda et al. have shown that Nalm6 carries a DUX4 rearrangement.11 In addition, in the extensive dataset on DUX4-rearranged leukemia published by Zhang et al.12, PDGFRA is highly expressed in DUX4-rearranged leukemia and its transcription start site was directly bound by DUX4 (Online Supplementary Tables S3a and S14 of reference 12). We therefore regard it as highly likely that the PDGFRA-high cases represent this novel DUX4-rearranged subtype.

These patients have a favorable outcome, but the risks of chemotherapy still demand new therapeutic approaches.13 Because transcription factors are difficult to target with small molecule inhibitors, PDGFRα inhibition could represent a treatment option. First, we confirmed protein expression and activation of PDGFRα by western blotting using Nalm6 cells. These cells were first starved and afterwards briefly stimulated with 10 ng/mL recombinant PDGFR-ligand PDGF-BB. Western blot analysis revealed that the ligand activates the phosphorylation of PDGFRα at position Y762 and reduces total PDGFRα levels (Figure 2A). Exposure to the PDGFR inhibitors imatinib and CP673451 abrogated the ligand-induced activation of PDGFRα, and prevented the phosphorylation of Y762.

Figure 2.PDGFRα signaling and inhibition in pediatric B-cell precursor acute lymphoblastic leukemia cells. (A) Western blot for phosphorylated (Y762) and total PDGFRα protein in the cell line Nalm-6. Cells were treated or not with 10 mM imatinib or 1 mM CP673451 and 10 ng/mL recombinant human PDGF-BB for 10 min. Arrow heads indicate the expected band. Total protein load was measured by staining for β-actin. It should also be noted that the closely related PDGFRB is not expressed in cases with high PDGFRA expression. (B) and (C) Ex vivo sensitivity of cell lines and primary leukemic cells to PDGFRα inhibitors in an ex vivo co-culture model. Cell lines (4 to 6 replicates) (B) and primary leukemic cells (C) were cultured with or without mesenchymal stromal cells (MSC) and exposed to a range of imatinib conv (0.2 μM – 1.6 μM – 12.5 μM) or CP673451 (0.02 μM – 0.16 μM – 1.25 μM) for 4 days. Cell survival was analyzed by flow cytometry and is depicted relative to the untreated control. Vehicle represents the respective vehicle control for imatinib (water) or CP673451 (DMSO) treated conditions. For details see the Online Supplementary Methods. The results of individual samples are plotted and connected by lines.

The sensitivity of BCP-ALL cells to PDGFRα inhibition was evaluated in an ex vivo co-culture model including bone marrow mesenchymal stromal cells (MSC). This model has been previously shown to improve the survival of primary BCP-ALL cells and also serves as a model of microenvironment-mediated drug resistance.14 The cell line EOL1 was highly sensitive to PDGFR inhibitors (imatinib and CP673451) in culture with and without MSC, while Nalm6 showed reduced viability only at the highest inhibitor concentrations (Figure 2B). Primary BCP-ALL samples with high PDGFRA expression were marginally sensitive to imatinib, but not to CP673451, in mono-culture (Figure 2C, left side,) and in co-culture with MSC (Figure 2C, right side). No cytotoxic effect of PDGFR inhibitors was observed in cases with low PDGFRA expression (Figure 2C). For Nalm6 and three BCP-ALL samples with sufficient material, the assay was repeated including recombinant human PDGF-BB, but ligand exposure did not sensitize to PDGFR inhibition (Online Supplementary Figure S5). We, therefore, conclude that inhibition of PDGFRα signaling with imatinib or CP673451 does not strongly reduce viability in PDGFRA-expressing primary cells, nor in the DUX4-rearranged PDGFRA-high expressing Nalm6 cell line. This is in contrast to a strong reduction in survival seen for the positive control, the FIP1L1-PDGFRA-rearranged cell line EOL1.

In summary, we have identified a group of BCP-ALL cases with high expression of the receptor tyrosine kinase PDGFRα. No genomic aberrations affecting the PDGFRA locus were detected: fusion-specific polymerase chain reactions to all known translocations were negative, FISH to the PDGFRA locus showed no interstitial deletion or translocation, and targeted locus amplification revealed no structural aberrations. However, two-thirds of the cases carried ERG deletions and, vice versa, all ERG-deleted cases showed high PDGFRA expression. Therefore, high PDGFRA expression likely characterizes the newly discovered DUX4-rearranged subtype with frequent ERG deletions.

Western blot experiments in the Nalm6 cell line demonstrated ligand-dependent activation of PDGFRα which was inhibited by imatinib and CP673451. Ex vivo exposure to these inhibitors revealed that activation of PDGFRα is not predictive for cytotoxicity in PDGFRA-high BCP-ALL cells. The baseline clinical characteristics of PDGFRA-high cases were not significantly different from those of “B-other” cases with low PDGFRA expression, although the small group size may have limited the analysis. Interestingly, we found a marked ex vivo sensitivity towards prednisolone in PDGFRA-high cases, in line with the good prognosis reported for ERG-deleted/DUX4-rearranged leukemia.131110

In conclusion, despite high PDGFRA expression, PDGFRα signaling was not essential to BCP-ALL cell survival ex vivo. Although in vivo data are warranted to corroborate these results, we conclude that targets other than PDGFRα should be explored for the cure of PDGFRA-high BCP-ALL.