Periodic Reporting for period 1 - PREDICT THE DRUG (Using ex vivo functional genomics as biomarker to predict the effectivity of Venetoclax in patients with acute leukemias)
Reporting period: 2024-01-01 to 2025-06-30
In 2018, 3.9 million new cancer cases arose, and 1.9 million cancer deaths occurred in Europe. Leukemia accounts for more than 500.000 cases worldwide per year. To provide these patients the most efficient therapy, it is crucial to identify biomarkers to select the best treatment option for each patient. In current clinical practice, cancer patients are treated according to their stratification in risk groups, with intensity increasing proportionally to the assigned risk. Leukemia cells are characterized based on different readout parameters, such as cell morphology, expression of certain surface molecules and chromosomal aberrations. Currently, correct stratification is very costly - from several hundred to few thousand Euros per patient - leading to a global hematology diagnostics market size of 6 billion USD per year.
Unfortunately, risk group stratification is often unspecific and has limited benefits for patients which is a major disadvantage of current clinical routine. As all patients within the same risk group are treated with the same drugs, numerous patients receive at least partially ineffective agents. Moreover, these drugs might be toxic resulting in avoidable adverse effects for the patients.
The identification of novel biomarkers is therefore essential to improve matching of the individual patient’s tumor with the most efficient drug. Two main strategies are followed: multi-omics sequencing enables drug selection based on descriptive genomic, epigenomic, transcriptomic and proteomic data, while ex vivo drug testing in cell or organoid cultures is used for drug selection on a functional level in preclinical settings. However, although being tested for several years or even decades, these techniques did not yet enter clinical routine. Despite major investments into ex vivo drug screening pipelines, their predictive value is limited, failing to correlate ex vivo drug data with their effectivity in vivo. Similarly, sequencing data are used by the Molecular Tumor Board to choose targeted therapies for individual patients. Regrettably, this approach proved little successful so far, suggesting that the genetic alteration alone is insufficient for response prediction. In contrast to the previous examples, measuring antigen expression levels proved to be predictive and clinically relevant for antibody-directed therapies, but is obviously limited to a subset of treatments. In general, currently used biomarkers are both highly resource-intensive and fail to effectively match a broad spectrum of antitumor therapies to individual patients who would most benefit from the treatment in terms of improved quality of life and extended survival.
As an example, the specific BCL-2 inhibitor Venetoclax is used for the treatment of several types of leukemia and various patient subgroups including elderly acute myeloid leukemia (AML) patients. Yearly costs for Venetoclax treatment sum up to 100.000€ per patient. Routinely, every patient within this group of elderly AML patients receives this drug. However, only a subset of patients benefits from this expensive treatment, while the rest of patients suffer from unnecessary side effects. To address the urgent need for a robust biomarker, several tests have been developed, such as BH3 profiling or the MACS-Score. Still, these tests are specific to this gene/drug pair and cannot be broadly applied to further genes or drugs.
Unfortunately, risk group stratification is often unspecific and has limited benefits for patients which is a major disadvantage of current clinical routine. As all patients within the same risk group are treated with the same drugs, numerous patients receive at least partially ineffective agents. Moreover, these drugs might be toxic resulting in avoidable adverse effects for the patients.
The identification of novel biomarkers is therefore essential to improve matching of the individual patient’s tumor with the most efficient drug. Two main strategies are followed: multi-omics sequencing enables drug selection based on descriptive genomic, epigenomic, transcriptomic and proteomic data, while ex vivo drug testing in cell or organoid cultures is used for drug selection on a functional level in preclinical settings. However, although being tested for several years or even decades, these techniques did not yet enter clinical routine. Despite major investments into ex vivo drug screening pipelines, their predictive value is limited, failing to correlate ex vivo drug data with their effectivity in vivo. Similarly, sequencing data are used by the Molecular Tumor Board to choose targeted therapies for individual patients. Regrettably, this approach proved little successful so far, suggesting that the genetic alteration alone is insufficient for response prediction. In contrast to the previous examples, measuring antigen expression levels proved to be predictive and clinically relevant for antibody-directed therapies, but is obviously limited to a subset of treatments. In general, currently used biomarkers are both highly resource-intensive and fail to effectively match a broad spectrum of antitumor therapies to individual patients who would most benefit from the treatment in terms of improved quality of life and extended survival.
As an example, the specific BCL-2 inhibitor Venetoclax is used for the treatment of several types of leukemia and various patient subgroups including elderly acute myeloid leukemia (AML) patients. Yearly costs for Venetoclax treatment sum up to 100.000€ per patient. Routinely, every patient within this group of elderly AML patients receives this drug. However, only a subset of patients benefits from this expensive treatment, while the rest of patients suffer from unnecessary side effects. To address the urgent need for a robust biomarker, several tests have been developed, such as BH3 profiling or the MACS-Score. Still, these tests are specific to this gene/drug pair and cannot be broadly applied to further genes or drugs.
During our ERC-CoG 68152 work, we established many patient-derived xenograft (PDX) models of AML and acute lymphoblastic leukemia (ALL). Via lentiviral transduction, we genetically engineered these models to express certain transgenes or for knockdown (KD) or knockout (KO) experiments. Recently, we established a CRISPR/Cas9 KO dropout screening pipeline, allowing for the detection of several AML and ALL dependencies which might serve as potential new therapeutic targets. A strong correlation between molecular targeting of these dependency genes and the activity of a corresponding targeted drug could be shown on a patient-specific level. These promising results led us to develop a functional genomics approach which offers a new working principle for the prediction of the activity of targeted drugs. Specifically, we designed a test based on examining the ex vivo viability of tumor cells upon CRISPR/Cas9-mediated KO of specific genes for drug response prediction.
Our aim for this proposal’s period was to establish the test pipeline and thus set the ground for a prospective clinical trial. Further steps, such as standardization, commercialization and implementation of the test were beyond the scope of the current proposal.
To optimize the set-up of the test, we used the gene/drug pair BCL-2/Venetoclax. We delivered ribonucleoprotein (RNP) complexes targeting BCL-2 via electroporation into AML PDX cells as well as primary patients’ material ex vivo and established a highly efficient protocol which enabled us to consistently obtain KO efficiencies above 90-95% without affecting the cell’s viability. Successful KO was confirmed via intracellular BCL-2 staining and flow cytometry measurement. Of note, we used a competitive assay design by mixing BCL-2 KO cells and control cells electroporated with RNPs targeting the non-essential gene PRB1 (Ctrl) in a 1:1 ratio. This approach increased the test’s sensitivity and minimized inter-well variability across replicates. The relative proportions of KO and control populations were monitored over a 14-day period to assess differential survival.
Based on the level of reduction of the BCL-2 KO population, we classified the samples as sensitive, partially sensitive or resistant. We could show that the test system could be robustly used for AML and ALL PDX models as well as with primary patient material, which is essential for further translation into the clinical setting. Additionally, we showed that both fresh and frozen material is amenable to the test system, leading to highly comparable results if the same sample is tested in both settings. Furthermore, we correlated our test’s results to previously generated in vivo therapy trials with Venetoclax using the same PDX models. Albeit the small sample size, by generating functional genomics data on the sensitivity upon BCL-2 KO, our test system was able to predict the in vivo response of the corresponding samples towards Venetoclax treatment with a Spearman correlation value of ~0.5. These data indicate that our test system harbors a huge potential to predict individual patients’ response to Venetoclax treatment. Importantly, unlike other BCL-2/Venetoclax test systems, our test can be easily adapted to further gene/drug pairs while using the same working principle. In addition, the test system can potentially be further adapted to solid cancers again broadening the test’s importance and putative impact for anti-cancer treatment.
Our aim for this proposal’s period was to establish the test pipeline and thus set the ground for a prospective clinical trial. Further steps, such as standardization, commercialization and implementation of the test were beyond the scope of the current proposal.
To optimize the set-up of the test, we used the gene/drug pair BCL-2/Venetoclax. We delivered ribonucleoprotein (RNP) complexes targeting BCL-2 via electroporation into AML PDX cells as well as primary patients’ material ex vivo and established a highly efficient protocol which enabled us to consistently obtain KO efficiencies above 90-95% without affecting the cell’s viability. Successful KO was confirmed via intracellular BCL-2 staining and flow cytometry measurement. Of note, we used a competitive assay design by mixing BCL-2 KO cells and control cells electroporated with RNPs targeting the non-essential gene PRB1 (Ctrl) in a 1:1 ratio. This approach increased the test’s sensitivity and minimized inter-well variability across replicates. The relative proportions of KO and control populations were monitored over a 14-day period to assess differential survival.
Based on the level of reduction of the BCL-2 KO population, we classified the samples as sensitive, partially sensitive or resistant. We could show that the test system could be robustly used for AML and ALL PDX models as well as with primary patient material, which is essential for further translation into the clinical setting. Additionally, we showed that both fresh and frozen material is amenable to the test system, leading to highly comparable results if the same sample is tested in both settings. Furthermore, we correlated our test’s results to previously generated in vivo therapy trials with Venetoclax using the same PDX models. Albeit the small sample size, by generating functional genomics data on the sensitivity upon BCL-2 KO, our test system was able to predict the in vivo response of the corresponding samples towards Venetoclax treatment with a Spearman correlation value of ~0.5. These data indicate that our test system harbors a huge potential to predict individual patients’ response to Venetoclax treatment. Importantly, unlike other BCL-2/Venetoclax test systems, our test can be easily adapted to further gene/drug pairs while using the same working principle. In addition, the test system can potentially be further adapted to solid cancers again broadening the test’s importance and putative impact for anti-cancer treatment.
We established a highly efficient electroporation protocol which enabled us to omit any enrichment strategy. Using a competitive approach, we could increase the test sensitivity and avoid any inter-well differences within the cell culture dishes. Additionally, we showed that both fresh and frozen material is amenable to the test system. Furthermore, our test system was able to predict the in vivo response of the corresponding samples towards Venetoclax treatment, outperforming the predictive value of other test systems currently used to predict Venetoclax response in patients.