Proteogenomic profiling uncovers differential therapeutic vulnerabilities between TCF3::PBX1 and TCF3::HLF translocated B-cell acute lymphoblastic leukemia

Although therapy escalation has led to improved 5-year overall survival rates for patients with B-cell acute lymphoblastic leukemia (B-ALL), few effective treatment options are available for relapsed and treatment-resistant disease. This applies particularly to specific subtypes of B-ALL, such as patients harboring TCF3 (formerly E2A) fusions. TCF3, encoding members of the E protein (class I) family of helix-loop-helix transcription factors, is a master regulator of B-cell development and is involved in several chromosomal translocations associated with lymphoid malignancies, such as the translocation t(1;19)(q23;p13.3), resulting in the TCF3::PBX1 fusion (5% of pediatric B-ALL) or the translocation t(17;19)(q22;p13) generating the TCF3::HLF fusion (~0.5% of pediatric B-ALL).² Omics research for the discovery of novel treatment strategies in hematological cancer is still based largely on transcriptomics, although it is increasingly recognized that this does not translate well into the expression of proteins, which are the main targets of drugs and functional entities of biological processes. In this study, we comprehensively analyzed the proteomic landscapes of TCF3::HLF⁺ (N=6) and TCF3::PBX1⁺ (N=5) B-ALL employing primary patient-derived xenografts (PDX), liquid chromatography tandem mass spectrometry and data-dependent acquisition. Approval for the study reported here was granted by the Ethics Committee of the Medical Faculty of the Christian-Albrechts-University, Kiel, Germany (vote D508/13). We detected 6,863 proteins (6,123 without ≥2 missing values; Online Supplementary Table S1), which allowed a clear distinction between TCF3::HLF⁺ and TCF3::PBX1⁺ leukemia by unsupervised hierarchical clustering and principal component analysis (Figure 1A, B). Proteomic profiling proved a useful tool for prioritizing drug targets, as only 8.45% of the significantly differentially expressed genes (N=119/1,409; P<0.05 and minimal log₂ fold change of ±1) previously detected by RNA sequencing² showed differential expression on protein level confirmed by our proteomic analysis (Online Supplemenary Figure S1A). In contrast, 34.8% (N=119/342) of differentially regulated proteins detected by proteomics were also differentially expressed on RNA level. As a proof-of-concept, we examined overlap of differentially expressed genes (cutoffs: P<0.05 and minimal log₂ fold change of ±1) from RNA sequencing and proteomic analysis obtained from a previously published dataset of ETV6::RUNX1⁺ (N=9) and high hyperdiploid (N=18) primary ALL patient samples.³ While only 3.63% (N=82/2,262) of differentially expressed genes detected via RNA sequencing showed differential expression on protein level, 92.13% (N=82/89) of differentially regulated proteins were also differentially expressed on RNA level (Online Supplementary Figure S1B).

In order to identify protein classes presenting specific therapeutic vulnerabilities, we performed gene set enrichment analysis (GSEA). We identified several gene sets enriched in either of the two subgroups (Figure 1C). RNA biology, mitochondrial translation and cellular respiration were the most prominent enriched gene sets for TCF3::HLF⁺ leukemia. In addition, strongly increased MYC expression and enrichment in MYC targets (Figure 1D, E) were detected, consistent with TCF3::HLF-driven activation of a MYC enhancer cluster previously shown using extensive functional genomics.⁴ For TCF3::PBX1⁺ leukemia, immune response/ cell cycle, actin cytoskeleton, cell morphogenesis and RTK signaling were among the most prominent enriched gene sets (Figure 1C). We validated therapeutic vulnerabilities indicated by GSEA using high-throughput drug screening. To this end, we tested the sensitivity of leukemic cell lines (TCF3::HLF⁺: HAL-01; TCF3::PBX1⁺: 697 and RCH-ACV) and mononuclear cells from peripheral blood of three healthy donors against a drug library of over 600 Food and Drug Administration-approved or clinical trial phase I-IV anti-cancer drugs. TCF3::HLF⁺ and TCF3::PBX1⁺ leukemic cells showed a differential response towards 109 drugs based on the area under the curve (AUC) as response parameter (Figure 2A; Online Supplementary Table S2; AUC<0.8 and >1.2 as a cutoff). Compared to our previous screening of bioactive compounds (N=98) employing the PDX samples,² the cell lines showed similarly increased sensitivity towards compounds, such as BCL2 and mTOR inhibitors for TCF3::HLF⁺, and aurora kinase and polo-like kinase inhibitors for TCF3::PBX1⁺ B-ALL (Figure 2A). In addition, we identified novel potential drug targets. These included MDM2 and DNA/RNA synthesis for TCF3::HLF⁺ and micro-tubule/tubulin and cyclin-dependent kinases (CDK) for TCF3::PBX1⁺ leukemic cells (Figure 2A). In order to confirm these findings, we chose drugs from those groups, which did not affect normal peripheral blood cells (Figure 2B-E), and treated TCF3::PBX1⁺ (RCH-ACV) and TCF3::HLF⁺ (HAL-01) cells with half half-maximal inhibitory concentration (IC₅₀), IC₅₀ and double IC₅₀ concentrations to investigate apoptosis induction. We demonstrated increased caspase 3/7 activity and apoptotic subG₁ cells in TCF3::PBX1⁺ B-ALL in response to the microtubule/tubulin inhibitor ixabepilone and the CDK inhibitor SNS-032. For TCF3::HLF⁺ leukemic cells, we verified increased apoptotic cell death upon idasanutlin (MDM2 inhibitor) and bleomycin sulfate (DNA/RNA synthesis inhibitor) treatment (Figure 2F-I).

Figure 1.Proteomic profiling distinguishes TCF3::HLF⁺ and TCF3::PBX1⁺ leukemia and uncovers therapeutic vulnerabilities for both subtypes. Unsupervised hierarchical clustering (A), principal component analysis (PCA) (B) and gene set enrichment analysis (GSEA) (C) was performed on the proteomic data of 6 TCF3::HLF⁺ (in red) and 5 TCF3::PBX1⁺ (in orange) B-cell acute lymphoblastic leukemia (B-ALL) patient-derived xenograft samples. For unsupervised hierarchical clustering, the 10% most variable proteins were used based on the standard deviation. PCA was performed on all proteins. Both subtypes clearly segregate into distinct clusters suggesting highly distinct proteomic landscapes. (C) GSEA is based on all proteins and identifies several gene sets enriched in either of the two subgroups. (D) Proteomic data shows enrichment of MYC protein expression in TCF3::HLF⁺ versus TCF3::PBX1⁺ leukemic samples in this study. In order to determine differential expression, non-parametric Mann-Whitney t test (two-tailed) was used. (E) Gene set enrichment plot of MYC targets showing a positive correlation with TCF3::HLF⁺ leukemia.

Figure 2.High-throughput drug screening and functional analysis confirm therapeutic vulnerabilities for TCF3::HLF⁺ and TCF3::PBX1⁺ leukemic cells. (A) HAL-01 (TCF3::HLF⁺), 697 and RCH-ACV (both TCF3::PBX1⁺) were exposed to a drug library, consisting of over 600 compounds. The supervised heatmap is based on the area under the curve (AUC, minimal fold change of <0.8 or >1.2 as a cutoff) as response parameter. Presented are groups of drugs with a differential response between TCF3::PBX1⁺ and TCF3::HLF⁺ leukemic cells. High sensitivity is shown in blue and low sensitivity is shown in white. (B-E) Presented are the dose-response curves for ixabepilone (B), SNS-032 (C), bleomycin sulfate (D) and idasanutlin (E), showing a differential response of TCF3::HLF⁺ versus TCF3::PBX1⁺ leukemic cells. Human peripheral blood cells from 3 healthy donors (in black) served as control cells (B-E). (F-I) RCH-ACV and HAL-01 were treated with ixabepilone (10 nM, 20 nM, 40 nM), SNS-032 (133 nM, 266 nM, 532 nM), idasanutlin (131 nM, 262 nM, 524 nM) or bleomycin sulfate (235 nM, 470 nM, 940 nM) dissolved in dimethyl sulfoxide (DMSO) or DMSO as negative control. (F, G) Bar graphs representing the fold change in caspase 3/7 activity 24 hours after treatment. Caspase 3/7 activity was determined using the Caspase-Glo 3/7 Assay (Promega). (H, I) Bar graphs represent the proportion of apoptotic cells (subG₁) 48 hours after treatment. Cell cycle profiles were generated by flow cytometric measurement of propidium iodide intercalation into DNA after partial cell lysis in hypotonic buffer (0.1% sodium citrate, 0.1% Triton X-100, 0.5 mg/mL RNase A containing 40 µg/mL propidium iodide). Values shown in (F-I) represent mean ± standard error of the mean of 3 biologically independent replicates. NS: not significant; *P<0.05; **P<0.01; ***P<0.001 (t test).

Besides the detection of vulnerabilities to specific drug classes, we aimed to identify novel targets for drug development. In our proteomic analyses, the B-lymphoid tyrosine kinase (BLK) was the most significantly upregulated protein for the TCF3::PBX1⁺ subtype (Figure 3A; Online Supplementary Figure S1C; Online Supplementary Table S1; minimal log₂ fold change of ±1 and significance level of P<0.05 as cutoffs). BLK encodes a non-receptor tyrosine kinase of the src family of proto-oncogenes that plays an important role in precursor (pre) B-cell receptor (BCR) signaling and early B-cell development.⁵ RNA-sequencing and expression microarray data by us and others supported this finding (Figure 3B-D). We examined human gene expression data derived from four independent data sets of >3,000 leukemia cases^6-9 available at the R2: genomics analysis and visualization platform (http://r2.amc.nl). These data indicated a subpopulation of leukemia samples that highly co-expresses BLK and PBX1 (Figure 3B; Online Supplementary Figure S1D-F). In two of the data sets, information on chromosomal translocations was available. There, the BLK and PBX1 co-expressing subpopulation was specifically associated with the TCF3::PBX1 fusion (Figure 3B; Online Supplementary Figure S1D).^8,9 In the Microarray Innovations in LEukemia (MILE) study⁸ all 36 cytogenetically identified TCF3::PBX1⁺ cases were BLK^high expressing (N=1,897 other leukemia or myelodysplastic syndrome [MDS], N=71 normal controls). Similarly, in another study⁹ all six TCF3::PBX1⁺ cases and none of the other samples (N=185 other B-ALL, N=3 normal controls) were both PBX1 and BLK high expressing. RNA-sequencing data of our cohort showed high RNA expression of BLK in all TCF3::PBX1⁺ leukemia cases (N=5 at diagnosis, N=8 after transplantation into NSG mice)² compared to TCF3::HLF⁺ cases (N=5 at diagnosis, N=22 after transplantation) (Figure 3C, D).

Thus, we hypothesized that interference with BLK signaling might present a potential treatment strategy for TCF3::PBX1-rearranged B-ALL in particular. In order to test this, we treated TCF3::PBX1⁺ BLK^high (RCH-ACV) and TCF3::HLF⁺ BLK^low (HAL-01) cells with a first selective irreversible BLK inhibitor BLK-IN-2.¹⁰ BLK^high cells responded in a dose-dependent manner starting at nanomolar concentrations (IC₅₀=0.2169 μM), whereas BLK^low cells showed little or no response (≥167-fold less, IC₅₀=36.20 μM) (Figure 3E; Online Supplementary Figure S1G-I). We further tested the impact of BLK-IN-2 on other genetic subtypes of B-ALL and noticed preferential sensitivity of only the TCF3::PBX1⁺ subtype to BLK-IN-2 (Online Supplementary Figure S1I). In order to test if BLK inhibition synergizes with the specific vulnerabilities identified in our proteomic screen, we performed combined treatment with ixabepilone (microtubule/tubulin inhibitor). Indeed, both drugs synergized in TCF3::PBX1⁺ cell lines, but not in the TCF3::HLF⁺ cell line HAL-01 (Figure 3F-H). We further tested, if interference with pre-BCR signaling including BTK inhibitors would have the same impact. To this end, we tested the response of TCF3::PBX1⁺ B-ALL cell lines to ibrutinib and other BTK-targeting drugs (acalabrutinib, spebrutinib, LFM-A13). The response, however, was low and did not differ from cells lacking pre-BCR expression.

As previously reported by Geng et al., BLK is a signature gene of adult TCF3::PBX1⁺ B-ALL.⁹ Combining chromatin immoprecipitation sequencing, DNA methylation and expression profiling, the authors identified hypomethylation and overexpression of BLK in adult TCF3::PBX1⁺ B-ALL. In this study, upregulated genes targeted by TCF3::PBX1 included pre-BCR components and pre-BCR downstream signaling molecules.¹¹ Ligand-independent autonomous tonic preBCR activation via self-aggregation is a main mechanism for pre-BCR activation and leads to constitutive activation of BLK¹¹ (indicated by phosphorylation of the activating tyrosine Y388) observed in several TCF3::PBX1⁺ cell lines and primary B-ALL.¹¹ Pre-BCR function induces activation of the transcription factor BCL6, which further increases pre-BCR signaling in a self-enforcing positive feedback loop and directly activates BLK transcription. More than 90% of TCF3::PBX1⁺ leukemia cases are pre-BCR⁺ and critically rely on pre-BCR-dependent signaling for proliferation.⁵ Thus, targeting BLK to abrogate pre-BCR downstream signaling presents an attractive approach for therapeutic intervention in TCF3::PBX1⁺ B-ALL.

Figure 3.Proteogenomic profiling detects BLK as a marker of TCF3::PBX1⁺ B-cell acute lymphoblastic leukemia targetable by the selective BLK inhibitor BLK-IN-2. (A) The volcano plot presents significantly (red) or non-significantly (grey) dysregulated proteins in TCF3::PBX1⁺ (left) or TCF3::HLF⁺ (right) leukemias. BLK is the most significantly dysregulated protein for the TCF3::PBX1⁺ subtype (minimal log₂ fold change of ±1 and significance level of P<0.05 as cutoffs). (B) Gene expression data of human patient samples is derived from data sets available at the R2: genomics analysis and visualization platform (http://r2.amc.nl). Dot plot presents PBX1 and BLK expression in healthy controls (black), TCF3::PBX1⁺ (orange) and other unstratified leukemic samples (grey). Data of the Mixed Leukemia - MILE - 2004 - MAS5.0 - u133p2 study⁸ is shown. Three further studies are presented in the Online Supplementary Figure S1D-F. (C,D) BLK RNA expression in TCF3::HLF⁺ (N=5) and TCF3::PBX1⁺ (N=5) B-ALL at the time of diagnosis (C) and after transplantation into NSG mice (TCF3::HLF⁺: N=22; TCF3::PBX1⁺: N=8). (D). Data is derived from our previous study.² In order to determine differential expression, non-parametric Mann-Whitney t test (two-tailed) was used. (E) Dose-response curves for BLK-IN-2 show a differential response of TCF3::HLF⁺ (red) versus TCF3::PBX1⁺ (orange) leukemic cells. (F-H) Synergy drug screening of BLK-IN-2 and Ixabepilone reveals a strong synergistic effect in TCF3::PBX1⁺ cell lines, while no such effect is detected in the TCF3::HLF⁺ cell line HAL-01. Representative synergy plots of 3 independent experiments are shown. Drug concentration ranges were chosen according to the predetermined half-maximal inhibitory concentration (IC₅₀) values of each cell line (RCH-ACV. and 697: 5-500 nM BLK-IN-2, 0.25-25 nM ixabepilone; HAL-01: 0.25-25 μM BLK-IN-2, 5-500 nM ixabepilone). Dimethyl sulfoxide (DMSO) was used as a negative control. ZIP synergy score analysis was conducted using the synergyfinder package version 3.0.14 with additional baseline correction.

Previously, interference with pre-BCR signaling has been suggested as a therapeutic option for TCF3-rearranged ALL¹² and the inhibitors ibrutinib, dasatinib and idelalisib to be effective against TCF3::PBX1⁺ B-ALL.^5,12,13 Inhibition of BTK, a downstream signaling kinase of the BCR, by ibrutinib is clinically beneficial in BCR⁺ B-cell malignancies such as non-Hodgkin lymphomas and multiple myeloma. However, in our analyses, the response of TCF3::PBX1⁺ B-ALL cell lines to BTK-targeting drugs was low and not corresponding to pre-BCR expression. This is in line with the observation that ibrutinib exerts a cytostatic rather than a cytotoxic effect on pre-BCR⁺ B-ALL cells⁵ and is further supported by the lack of in vivo effectivity against TCF3::PBX1⁺ primogafts.¹⁴ However, TCF3::PBX1⁺ PDX samples responded well to the tyrosine kinase inhibitor dasatinib.² Still, high doses might be required for targets other than BCR::ABL1 and might be limited in the relapse setting due to toxicity.¹³

Taken together, proteomic-based profiling is a powerful tool to discover highly specific and sensitive cancer biomarkers and oncogenic pathway activation.¹⁵ Here, we uncovered proteomic alterations associated with TCF3::HLF⁺ or TCF3::PBX1⁺ B-ALL and revealed potential therapeutic options for these subtypes. These include previously known sensitivities for TCF3::HLF⁺ (e.g., BCL2 and mTOR) and TCF3::PBX1⁺ B-ALL (e.g., aurora kinase and polo-like kinase), as well as potential novel drug targets, such as MDM2 and DNA/RNA synthesis for TCF3::HLF⁺ or microtubule/tubulin and CDK for TCF3::P-BX1⁺ leukemic cells. Our data suggest that TCF3::PBX1⁺ B-ALL might be sensitive to treatment with selective BLK inhibitors, especially in combination with microtubule/tubulin targeting drugs, such as ixabepilone. Due to high BLK expression in TCF3::PBX1⁺ B-ALL cells, such inhibitors could selectively eradicate leukemic cells at doses eliciting less side-effects on normal tissue. A limitation of our study is that this was not tested in mouse models. In future studies, it would be interesting to apply BLK inhibition to suitable mouse models of TCF3::PBX1⁺ leukemia and to test synergism with other drugs. Larger numbers of patient samples need to be tested to show the applicability for TCF3::PBX1⁺ leukemia.

Footnotes

• Received July 17, 2023
• Accepted February 15, 2024

Correspondence

^#MS, J-PB, AB, MR, OA and UF contributed equally as senior authors

U. FISCHER - ute.fischer@med.uni-duesseldorf.de

Disclosures

No conflicts of interest to disclose.

Contributions

LB, JB, MS, J-PB, AB, MR, OA and UF planned and directed the study. Patient-derived xenograft models were provided by BB, BM and J-PB. FB, JTD, VM, FD, DL and OA conducted proteomic profiling and analyzed proteomic data. DP provided bioinformatic analyses. LB designed and performed the in vitro experiments, supported by VJ, JS-D and RH. NQ generated dose-response curves. SB provided intellectual contributions to the project and to the interpretation of the results. LB, VJ and UF wrote the manuscript. Figures were designed and drafted by LB, DP, NQ, VJ, JS-D and UF. All authors edited and contributed to the final manuscript.

Funding

Acknowledgments

Frauke-Dorothee Meyer, Sarah Göbbels and Bianca Killing are acknowledged for excellent technical assistance.