IAP dependency of T-cell prolymphocytic leukemia identified by high-throughput drug screening Free

T-cell prolymphocytic leukemia (T-PLL) is an aggressive T-cell lymphoma characterized by the proliferation of mature, postthymic T-lymphocytes.¹ Its clinical course is usually rapid, with progressive lymphocytosis, splenomegaly, and lymphadenopathy.¹ T-PLL response to conventional chemotherapy is poor, and remissions induced with the monoclonal anti-CD52 antibody alemtuzumab are rarely maintained.¹ With a median overall survival of <20 months,^1,2 there is an urgent need to identify new treatment strategies. High-throughput ex vivo drug screening enables the systematic assessment of drug sensitivities in blood cancers. Previous studies identified T-PLL sensitivity to the B-cell lymphoma 2 (Bcl-2) homology 3 (BH3) mimetic venetoclax, p53 activators, as well as inhibition of histone deacetylase (HDAC) or the Janus kinase (JAK)–signal transducer and activator of transcription (STAT) pathway.^3-7 Early clinical data, however, suggest that these compounds are insufficient to provide long-term control of the disease, illustrating the need for additional treatment options.^6,8,9 We have previously reported on the ex vivo drug sensitivity of several blood cancers, including T-PLL.⁵ Here, we present a resource of 5 ex vivo drug screening data sets, along with a reanalysis of the previous data. In total we capture T-PLL drug response to >2800 unique compounds. We combine these data with combinatorial drug perturbations, bulk RNA sequencing (RNA-seq) and single-cell RNA-seq (scRNA-seq), as well as spectral flow cytometry, to study previously undescribed T-PLL pathway dependencies, including a particular sensitivity to inhibitor of apoptosis protein (IAP) inhibition.

Drug screens were conducted using viably frozen peripheral blood mononuclear cells (PBMCs) following previously described protocols.⁵ Details on the screens, the bulk RNA-seq, scRNA-seq, and spectral flow cytometry experiments can be found in the supplemental Methods, and supplemental Tables 2 and 4, available on the Blood website. Targeted sequencing of TP53 and western blots were performed using previously published approaches and are detailed in the supplemental Methods.^10,11 

The preprocessing and analysis of drug screen data, including quality filtering, normalization, incubation effect correction, drug-drug correlation, and the testing for synergistic or antagonistic drug combinations followed previously published approaches and are detailed in the supplemental Methods.^5,12 

The study was approved by the ethics committees in Heidelberg, Germany (University of Heidelberg; S-206/2011; S-356/2013) and Zurich, Switzerland (2019-01744).

To capture T-PLL drug sensitivities, we measured the drug response of 9 primary blood cancers in 5 drug screens ex vivo (referred to as screens A, B, C, D, and E, respectively). This included the previously published screen C, which we reanalyzed with a focus on T-PLL.⁵ 

For each screen, PBMCs were isolated from the blood and cultured with drugs for 48 hours, before cell viability was assessed by measuring adenosine triphosphate (Figure 1A). Screen A used a library of 2516 drugs, including chemotherapeutics, kinase inhibitors, and clinically used and investigational compounds, and was performed on 34 patient samples (n_T-PLL = 6; Figure 1B). Screen B contained primarily small-molecule inhibitors (n = 409) with multiple compounds per drug class at a wider concentration range in 141 patient samples (n_T-PLL = 8). Screens C and D were performed with a subselection of 63 compounds each, representing key drug classes, in a greater number of patient samples (n_{Patients Total} = 219 [screen C], 300 [screen D]; n_{PatientsT-PLL} = 25 [screen C], 10 [screen D]).

Finally, 50 drugs were chosen for the testing in a wider set of T-PLL (n = 12) and T-cell malignancies (n = 10), referred to as screen E (Figure 1B).

A full list of the compound libraries and the key screen characteristics is provided in Table 1, and supplemental Tables 1 and 2.

We began our analysis by ranking drugs based on their effect on viability. Of 2516 compounds in screen A, only a small fraction (n = 253; 10%) decreased the mean viability (v) of T-PLL patient samples by >10% and only 99 (3.9%) by >25% (Figure 1C). This limited response to drugs was observed across screens (supplemental Figure 1A-F).

Most tyrosine kinase inhibitors demonstrated little activity, even when used at high concentrations (>10 μM). This included the JAK1/2 inhibitor ruxolitinib and the phosphoinositide 3-kinase inhibitor duvelisib, 2 drug candidates for the treatment of T-cell lymphoma (mean viability [v]_screen B = 0.95, 0.95; supplemental Figure 2A-B).^4,18,19 Chemotherapeutics were among the most toxic drugs. This included drugs used to treat T-PLL such as the antimetabolites cytarabine (v_screen B = 0.78) and fludarabine (v_screen C = 0.55; Figure 1D; supplemental Figure 1F).^1,20 Previous studies reported T-PLL sensitivity to p53 activation, BH3-mimetics, and HDAC inhibition.^3,5,6 In line with this, the p53 activator nutlin-3a was highly active across screens (v_screen C = 0.63); however, some patient samples appeared to be resistant (Figure 1E). Because TP53 mutation is a common mechanism of nutlin-3a resistance, we performed targeted sequencing of the TP53 gene, but did not detect any mutations (Table 1).^5,21 The Bcl-2 inhibitor venetoclax induced a more pronounced decrease in viability than other BH3 mimetics (navitoclax, obatoclax, sabutoclax, and TW-37), which possess a broader range of targets (Bcl-xl, Bcl-w), but lower binding affinity for Bcl-2 (screen B; adjusted P value <.001; Figure 1F-G).^13-17 This is in line with reports that describe T-PLL to primarily rely on the antiapoptotic signal of Bcl-2.^4,6 We confirmed the sensitivity of T-PLL to HDAC inhibitors (ie, panobinostat, v_screen B = 0.65), while also uncovering a susceptibility to other classes of epigenetic modifiers; specifically, histone demethylase (JIB04), histone methyltransferase (UNC0638), and BET inhibitors (OTX015, JQ1; v_screen B = 0.68, 0.8, 0.85, and 0.86, respectively; supplemental Figure 1E-F).³ Notably, we observed a high degree of correlation when comparing the drug response between screens for overlapping patient samples and compounds (R² = 0.59-0.81; supplemental Figure 2C).

We also uncovered previously undescribed pathway dependencies. T-PLL was especially vulnerable to nuclear export (selinexor and verdinexor), autophagy (thapsigargin and bafilomycin A1), and IAP (birinapant and GDC-0152) inhibition (v_screen B = 0.74, 0.7, 0.76, 0.75, 0.72, and 0.83, respectively; Figure 1H-J; supplemental Figures 1F and 2D).

We asked which of the drug effects were particularly pronounced in T-PLL by comparing the drug response of T-PLL with that of primary chronic lymphocytic leukemia and mantle cell lymphoma (MCL) samples (2-sided t test; false discovery rate [FDR] of <10%; and mean viability difference = Δ). We found T-PLL to be significantly more resistant to most drug classes, including the clinically used¹ chemotherapeutic fludarabine (Δ_screen D = 0.18; adjusted P value = .006), as well as the BH3 mimetics venetoclax (Δ_screen B = 0.18; adjusted P value = .03) and navitoclax (Δ_screen B = 0.17; adjusted P value = .014; Figure 2A-C; supplemental Figure 3A-D). Only a small set of compounds demonstrated a higher activity in T-PLL: the p53 activator nutlin-3a (Δ_screen A = −0.34; adjusted P value = 3 × 10⁻⁵), thalidomide (Δ_screen B = −0.052; adjusted P value = .078), the autophagy inhibitors thapsigargin (Δ_screen C = −0.38; adjusted P value = 3.41 × 10⁻⁷) and bafilomycin A1 (Δ_screen B = −0.18; adjusted P value = .02), the IAP inhibitor birinapant (Δ_screen B = −0.22; adjusted P value = .007), as well as the nuclear export inhibitor selinexor (Δ_screen D = −0.13; adjusted P value = .01; Figure 2A-C; supplemental Figure 3A-D). To explore whether these sensitivities are shared by other T-cell malignancies, we performed a validation screen (screen E) using a selection of compounds with T-PLL–specific activity, as well as a diverse range of 50 selective inhibitors and chemotherapeutics in 27 and 22 samples from patients with B-cell and T-cell lymphoma, respectively (Figure 1B). T-cell lymphoma samples (n = 10) demonstrated a response pattern similar to T-PLL (Figure 2D-E). This suggests that key sensitivities are shared across T-cell malignancies.

Based on their pronounced effect on viability and their T-PLL– and T-cell lymphoma–specific activity, we selected birinapant, selinexor, and bafilomycin A1 for further exploration. We observed interpatient heterogeneity in the viability effects of these compounds (Figure 2E-F). To test whether this heterogeneity was simply indicative of overall drug sensitivity, we clustered drugs based on the similarity of their response in T-PLL and T-cell lymphoma samples (Figure 2G-H; supplemental Methods). We found distinct groups of compounds with similar pathway dependencies (Figure 2G). For instance, birinapant showed significant correlation with GDC-0152, a second, less potent IAP inhibitor (R² = 0.88; adjusted P value = 4.1 × 10⁻⁶; Figure 2G-H).²² No correlation was observed between birinapant and chemotherapeutics (fludarabine), p53 activators (nutlin-3a and serdemetan), HDAC inhibitors (dacinostat), and BH3 mimetics (venetoclax and navitoclax; Figure 2G-H). Selinexor demonstrated significant correlation with nutlin-3a, serdemetan, venetoclax, and navitoclax (R² = 0.85, 0.64, 0.56, and 0.69; adjusted P value = 2.6 × 10⁻⁵, .022, .061, and .009; respectively; Figure 2G; supplemental Figure 3E-F). Bafilomycin A1, meanwhile, showed high correlation with the HDAC inhibitor dacinostat (R² = 0.73; adjusted P value = .005; Figure 2G). Combined, these findings indicate similarity in the mode of action of selinexor and p53 activators and BH3 mimetics, as well as bafilomycin A1 and HDAC inhibitors. IAP inhibitors, however, appear to act in T-PLL and T-cell lymphoma via a unique mode of action.

To further characterize the drug response to birinapant, selinexor, and bafilomycin A1, we treated 9 samples from patients with T-PLL and 1 samples from a patient with T-large granular lymphocytic leukemia (T-LGL) with these compounds as well as 8 specific pathway inhibitors (nutlin-3a, pomalidomide, venetoclax, ruxolitinib, ibrutinib, everolimus, dacinostat, and motolimod, or dimethyl sulfoxide [DMSO]) for 24 hours and performed bulk RNA-seq (Figure 3A; supplemental Figure 3G). To uncover drug effects, we assessed differential gene expression by comparing each condition to the DMSO control (FDR < 10%, |log₂ fold change [FC]| > 0.25; Figure 3B-C; supplemental Figure 4A). The overall number of significantly differentially expressed genes (DEGs) varied greatly between treatments with dacinostat and pomalidomide displaying the highest number of DEGs (n = 7884, 6569), whereas other treatments such as nutlin-3a or birinapant showed more restricted expression changes (n = 682, 203) (Figure 3B-C; supplemental Figure 4A). This was also reflected in the number of significantly upregulated or downregulated pathways as identified by DEG network and gene set enrichment analysis (FDR < 10%; Figure 3D-E; supplemental Figure 4A-F). As expected, treatment with nutlin-3a resulted in a strong upregulation of p53 target genes (BAX, and MDM2, and CDKN1A; Figure 3E-F). Notably, selinexor showed an overlap with nutlin-3a in terms of DEGs as well as an upregulation of DNA-damage response gene sets (Figure 3G; supplemental Figure 4C-D). This included an upregulation of the TP53 gene itself (log₂FC = 0.99; adjusted P value = 6.89 × 10^–19), demonstrating that selinexor acts through the TP53 pathway in T-PLL (Figure 3G). Treatment with the autophagy inhibitor bafilomycin A1 had a strong transcriptional effect on T-PLL cells as judged by the number of DEGs (n = 4328). We observed a considerable overlap with the DEGs induced by the Toll-like receptor 8 (TLR8) agonist motolimod (Figure 3B; supplemental Figure 4A,G). Notably, this included an upregulation of several tumor necrosis factor α (TNF-α)/NF-κB signaling pathway members (TRAF2, BCL3, and NFKBIZ) as well as the proinflammatory cytokine interleukin 6 (IL6), which suggests that vacuolar-type adenosine triphosphatase inhibition not only prevents lysosomal acidification in T-PLL but also activates TLR signaling as has been demonstrated for human monocytes (Figure 3I).²⁸ 

Birinapant inhibits cIAP-1 and cIAP-2, 2 IAP family members that promote apoptosis resistance and cell survival by blocking prodeath signals through the TNF receptor 1, Fas receptor (CD95), and TLR signaling pathways.^29-32 Birinapant treatment led to a marked upregulation of the TNF-α/NF-κB signaling pathway (NFKBIA, NFKB1, NFKB2, TRAF1, TRAF2, and BIRC3; Figure 3H-I). Notably, the effect was distinct from that of the TLR8 agonist motolimod, because birinapant treatment only had a limited effect on the expression of TLR-associated proinflammatory cytokines (IL1A, IL1B, IL6, CXCL1, CXCL2, and CXCL3; Figure 3I; supplemental Figure 4H). This is in line with reports, which observed improved tolerability of birinapant over other IAP inhibitors because of birinapant’s limited effect on XIAP, a key regulator of inflammasome activation.^33,34 

Combined, these findings suggest distinct pathway dependencies of birinapant and selinexor on the TNF-α/NF-κB and p53 pathways, whereas a broader transcriptional effect was observed for bafilomycin A1.

We were intrigued by the potent effects of birinapant on T-PLL and T-cell lymphoma viability as well as the highly specific transcriptional effect. Given birinapant’s manageable safety profile in a phase 2 clinical trial,³⁵ it represented a promising treatment candidate, which led us to investigate it further. To capture the effect of birinapant on healthy and malignant cell subsets we treated patient-derived samples (3 T-PLL, 3 MCL, 2 hairy cell leukemia [HCL], and 1 T-LGL) with birinapant and 4 specific pathway inhibitors (nutlin-3a, ibrutinib, everolimus, and selumetinib, or DMSO) for 48 hours, followed by scRNA-seq (Figure 4A). After quality control, we captured 49 664 cells. For 32 301 (65.0%) cells we additionally retrieved the B- or T-cell receptor sequences. Visualization by uniform manifold approximation and projection embedding revealed a strong separation of T-PLL, HCL, and MCL cells (monoclonal B- and T-cell receptor) from healthy B- and T-cells (polyclonal B- and T-cell receptor; Figure 4B-E; supplemental Figure 5A-F). To assess the transcriptional effect of drug perturbations on nonmalignant T-cells, we randomly sampled an equal number of cells per cell type and treatment, comparing each condition with the DMSO control (FDR < 10%, |log₂FC| > 0.25). Comparing the DEGs between cell types (HCL, MCL, T-PLL, and healthy T-cells) demonstrated that the effects of birinapant, ibrutinib, and everolimus were mainly restricted to the malignant cells, whereas nutlin-3a affected both malignant and healthy cell subsets (Figure 4F-J). As a control, we included the BRAF inhibitor selumetinib, which has a strong effect especially on HCL, given the dependency of these cells on a BRAF V600E mutation (Figure 4F).³⁶ We found that all lymphoma types and patient samples responded similarly to birinapant, with an upregulation of the TNF-α/NF-κB pathway, suggesting a conserved transcriptional response to IAP inhibition across lymphoma types (Figure 4K). We observed that this transcriptional response was homogenous across different clusters of malignant cells and did not identify a primary unresponsive subset of T-PLL cells (Figure 4L; supplemental Figure 5G-I). This homogenous pattern of TNF-α/NF-κB pathway activation contrasts the heterogeneous effects we previously observed for birinapant on cell viability, prompting us to further investigate the mechanisms of cell death in birinapant-treated T-PLL cells.

The pharmacological or genetic depletion of IAPs has been shown to induce cell death in cancer cells through apoptosis or a form of organized necrosis, termed necroptosis.^31,37 This process has been linked to the formation of distinct multiprotein signaling complexes, consisting amongst others of FADD, FLIP_S, FLIP_L, caspase-8, RIPK1, RIPK3, and MLKL (Figure 5A).^31,38,39 To test whether the interpatient heterogeneity in birinapant response may be explained by differential abundances in these proteins, we measured the expression of key apoptotic and necroptotic proteins in samples from patients with T-PLL (n = 8; Figure 5B; supplemental Figure 6A). We observed variable expression of cIAP1, cIAP2, cFLIP_L, RIPK1, RIPK3, and FADD across T-PLL samples (Figure 5B; supplemental Figure 6A). Notably, we found a trend for association between birinapant response and cIAP2 (BIRC3) and cFLIP_L (CFLAR) expression, which was reflected both on the protein and gene expression levels (protein: R² = −0.41, −0.54; RNA: R² = −0.65, −0.62; Figure 5C; supplemental Figure 6B-E). To explore the mode of birinapant-induced cell death in T-PLL, we performed a combinatorial drug screen with specific inhibitors of the apoptosis and necroptosis cascades. For this, we measured the viability of patient-derived samples (11 T-PLL and 1 T-LGL) in response to birinapant at 5 concentrations (0.016-10 μM) as single compound and in combination with the caspase inhibitor, quinoline-val-asp-difluorophenoxymethylketone (Q-VD-OPh); the RIPK1 inhibitor, necrostatin-1; the RIPK3 inhibitor, GSK-872; and the MLKL oligomerization inhibitor necrosulfonamide (NSA; Figure 5A). Drug combination synergy and antagonism (cell death blockage) were analyzed using a Bliss independence model (supplemental Figure 6F; supplemental Methods). Necrostatin-1, GSK-872, and NSA but not Q-VD-OPh prevented birinapant-induced cell death in T-PLL (mean combination index [CI] = 1.7, 1.18, 1.37; adjusted P value = .001, .021, .006; Figure 5D-H; supplemental Figure 7A-C). In contrast, Q-VD-OPh addition seemed to further sensitize individual T-PLL samples to IAP inhibitor–induced cell death (CI = 0.94; adjusted P value = .116; Figure 5G-H; supplemental Figure 7D). These findings indicate that IAP inhibitors lead to RIPK1-dependent and necroptotic cell death in T-PLL.^31,37 Because of the similar transcriptional response of the different lymphoma entities to birinapant treatment in our scRNA-seq data, we asked whether there were differences in the mode of cell death between lymphoma types. For this, we repeated the experiment in a larger set of 49 lymphoma samples (n_T-PLL = 12), testing birinapant at 10 concentrations (0.00076-15 μM) alone and in combination with Q-VD-OPh and necrostatin-1. In both T-PLL and T-cell lymphoma, birinapant-induced cell death was decreased upon the addition of necrostatin-1 (CI = 1.35, 1.18; adjusted P value = 3 × 10⁻⁴, 0.058) but not Q-VD-OPh (CI = 0.89, 1.06; adjusted P value = .08, .694; Figure 5I). In contrast, caspase inhibition prevented birinapant-induced cell death in MCL and chronic lymphocytic leukemia (CI = 1.21, 1.13; adjusted P value = .041, .016), whereas necrostatin-1 had little effect (CI = 0.99, 0.98; adjusted P value = .288, .076; Figure 5I; supplemental Figure 8A). Similar results were obtained when using GDC-0152 instead of birinapant and NSA instead of necrostatin-1 (Figure 5J; supplemental Figures 7E and 8B-C). Combined, these findings suggest that IAP inhibitors lead to RIPK1-dependent cell death by necroptosis in T-PLL and other T-cell lymphomas but induce RIPK1-independent apoptosis in B-cell entities.

The antiapoptotic function of IAP proteins involves blocking of prodeath signals through TNF receptor 1, Fas receptor (CD95), and TLRs.^29-32 Furthermore, autocrine production of TNF-α is required for cell death upon IAP inhibition in many in vitro models.^40,41 To test the relevance of these pathways for IAP inhibitor–induced cell death in T-cell lymphoma, we cultured samples from patients with T-PLL and T-LGL with birinapant at 5 concentrations (0.016-10 μM) as a single compound as well as in combination with the anti–TNF-α blocking antibody adalimumab (n = 10); recombinant CD95L (n = 10); the TLR8 agonist, motolimod (n = 12); and the Bruton tyrosine kinase/IL-2–inducible T-cell kinase inhibitor ibrutinib (n = 12). The addition of adalimumab and ibrutinib reduced birinapant-induced cell death in T-PLL and T-LGL, whereas TLR8 stimulation promoted T-PLL and T-LGL cell killing (CI = 1.75, 1.46, 0.9; adjusted P value = .006, .011, .085; Figure 5K-M; supplemental Figure 8D-F). Furthermore, CD95L binding sensitized individual patient samples to birinapant killing (CI = 0.93; adjusted P value = .106; supplemental Figure 8G). This was correlated with the CD95 cell surface expression levels of the malignant cells, as assessed by spectral flow cytometry (R² = −0.84; adjusted P value = .009; Figures 5N and 6A). Combined, these findings suggest that birinapant-induced cell death in T-PLL and T-LGL requires the binding of TNF-α and the activity of IL-2–inducible T-cell kinase, whereas TLR, and possibly CD95 pathway signaling, sensitize malignant T-cells.

Next, we sought to study the effect of IAP inhibitors on proliferation and cytokine expression in different cellular subsets including nonmalignant cells. To this end, we treated 10 T-PLL, 1 T-LGL and 4 healthy age-matched PBMC samples with either a sublethal concentration of birinapant (0.2 μM) or DMSO for 40 hours, followed by 4 hours stimulation ex vivo. We then performed spectral flow cytometry using a 30-marker panel (Figure 6A-E; supplemental Figure 9A). In addition to resolution of the T-cell compartment, and the major B-cell and natural killer (NK) cell types (supplemental Figure 9B-C), this panel included 3 distinct cell death markers, Apotracker Green (early and late cell death), LIVE/DEAD staining (late cell death), and cleaved caspase-3 (apoptotic) to explore the mode of birinapant-induced cell death. Birinapant treatment significantly reduced the frequency of proliferating lymphoma cells but not that of nonmalignant B, T, and NK cells (adjusted P value = .076, .918, .12, .417; Figure 6F; supplemental Figure 9D). The effect of birinapant on cytokine expression was cell type dependent (linear mixed model, FDR < 10%). Birinapant treatment significantly increased the expression of TNF-α and granulocyte-macrophage colony-stimulating factor in NK cells and showed a trend for increased IL-2 expression in T cells (adjusted P value = .092, .073, .227; Figure 6G; supplemental Figures 9E-J and 10A-B). Next, we sought to explore the mode of birinapant-induced cell death. To this end, we cultured 10 T-PLL, 1 T-LGL, and 4 healthy PBMC samples with either selinexor, birinapant (1 μM) with and without Q-VD-OPh, or DMSO, followed by spectral flow cytometry using the same 30-marker panel (Figure 6H-K; supplemental Figure 10C). Both selinexor and birinapant significantly reduced the fraction of live T-PLL and T-LGL cells compared with DMSO (2-sided paired t test, FDR < 10%; Δ = −0.096, −0.234; adjusted P value = .009, .023; supplemental Figure 10D). Although selinexor led to a significant enrichment of apoptotic cells, birinapant-induced cell death was primarily nonapoptotic (apoptotic: adjusted P value = .04, .801; nonapoptotic: adjusted P value = .023, .003; Figure 6L-M). Notably, the addition of Q-VD-OPh greatly increased birinapant killing of lymphoma cells but not of nonmalignant B, T, and NK cells, a finding consistent with necroptosis through caspase-8 inhibition (CI = 0.39, 1.38, 0.88, 0.95; adjusted P value = .065, .834, .898, .898; supplemental Figure 10D).^31,42,43 Among the nonmalignant cells, selinexor treatment significantly decreased the viability of B, T, and NK cells (Δ = −0.35, −0.22, −0.14; adjusted P value = .006, .006, .013; supplemental Figure 10D). This effect was pronounced for B and NK cells and resulted in a relative depletion of B-cell and NK cell subtypes among the living cells of the healthy PBMC donors (Figure 6N). Birinapant also decreased the fraction of live B and NK cells but did not affect the overall fraction of live nonmalignant T cells (Δ = −0.14, −0.08, −0.08; adjusted P value = .023, .002, .124; supplemental Figure 10D). We did, however, observe a trend toward the enrichment of naive and central memory CD4 and CD8 T cells among the living cells of healthy donor PBMCs, suggesting that birinapant treatment might differentially affect T-cell subsets (Figure 6O). In summary, spectral flow cytometry demonstrated that birinapant reduces T-cell lymphoma proliferation at sublethal concentrations and induces nonapoptotic death, whereas showing only a limited effect on nonmalignant immune cells.

Here, we report on drug vulnerabilities of T-PLL to exportin 1 (selinexor), autophagy (bafilomycin A1), and IAP (birinapant) inhibition, that were shared by other T-cell malignancies. To characterize the biological processes behind these vulnerabilities, we used spectral flow cytometry and bulk and scRNA-seq to study T-PLL drug response. We found the response of selinexor to be significantly correlated with that of nutlin-3a in our drug screening data and could subsequently demonstrate the upregulation of TP53 and its target genes after selinexor treatment. Bafilomycin A1, a strong inhibitor of autophagy, demonstrated considerable transcriptional overlap with the TLR8 agonist motolimod. This suggests that in addition to its effects on cell metabolism, bafilomycin A1 promotes TLR pathway activation in T-PLL, as has been proposed for human monocytes.²⁸ We explored the sensitivity of T-PLL to the IAP inhibitor birinapant, which is known to induce cell death through either apoptosis or necroptosis.^31,37 The dependence on RIPK1 and RIPK3 kinase activity, as well as MLKL activation in our drug combination experiments, combined with the lack of caspase-3 cleavage in our spectral flow cytometry data indicate that birinapant-induced cell death in T-PLL is primarily necroptotic. Indeed, the addition of the pan-caspase inhibitor Q-VD-OPh further sensitized lymphoma but not nonmalignant cell types to birinapant killing, likely by promoting necroptosis through caspase-8 inhibition.^31,42,43 The opposite effect was observed in B-cell lymphomas, in which birinapant-induced cell death was primarily RIPK1-independent and apoptotic. This was surprising given that T-PLL, T-LGL, MCL, and HCL cells showed a similar transcriptional activation of the TNF-α/NF-κB pathway after birinapant treatment in our scRNA-seq data and highlights the intricacy of the multiprotein signaling complexes that form after IAP depletion. We observed a limited effect of birinapant on nonmalignant T-cell transcriptomes and viability. These findings are in line with reports that show a limited impact of birinapant on healthy T-cell viability and even enhanced tumor cell killing of chimeric antigen receptor T cells after birinapant treatment.^44,45 Combined, these findings suggest that IAP inhibitors are not generally toxic for healthy T cells.

Previous studies have described T-PLL sensitivity to BH3 mimetics, p53 activators, as well as HDAC and JAK-STAT inhibitors.^3-7 Early clinical data, however, suggest that these compounds are insufficient to provide long-term control of the disease.^6,8 Recently, Jan et al have reported synergistic drug combinations for the treatment of T-PLL.⁹ By screening 8 drug classes with previously reported single-agent activity, the authors demonstrated encouraging results for combinations involving the p53 activator idasanutlin, the HDAC inhibitor romidepsin, and the ribonucleotide reductase inhibitor cladribine, based on the superior ability of these combinations to activate the p53 pathway.⁹ We now report on T-PLL sensitivity to >2800 compounds and identify hitherto unexplored drug vulnerabilities of T-PLL. Notably, each of the compounds (birinapant, selinexor, and bafilomycin A1), demonstrated distinct pathway dependencies, making them optimal candidates to explore drug combinations to circumvent intrinsic and acquired resistances (ie, TP53 escape mutations). Furthermore, birinapant and selinexor have demonstrated manageable toxicity in phase 2 clinical trials, which could facilitate their further clinical development in T-PLL and other T-cell malignancies.^35,46 

The authors thank Jan Dürig and Marco Herling for the contribution of samples; Berend Snijder, Julien Mena, Steffen Böttcher, and Roman Schimmer for input on the design and the printing of combinatorial drug plates; and André Samson for his insights on the necroptotic signaling cascade.

This work was supported by the KFSP Precision Hematology and Oncology, the Helmut Horten Foundation, and The LOOP Zurich (INTeRCePT).

M.F.P. is an MD thesis candidate at the University of Zurich, and this work is submitted in partial fulfillment of the requirement for the MD thesis.

Contribution: J. Lewis, W.H., and T.Z. devised the research strategy; A.M. cosupervised the research; M.F.P., K.P., S.S., L.B.T., T. Walther, S. Kummer, T. Wertheimer, M.L., T.H.L.D., K.H., J.M., S. Kisele, J. Kim, and J. Kivioja performed the experiments; M.F.P. and J. Lu performed data analyses; J. Lu, M.F.P., and K.B. developed the tools; K.B., X.F., W.H., W.W.-L.W., and T.Z. helped to interpret results; T.Z., J. Lewis, W.H., S.D., A.M., B. Becher, and B. Bornhauser provided resources; M.F.P., J. Lu, X.F., and T.Z. wrote the manuscript; and all authors reviewed the manuscript before submission.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Thorsten Zenz, Department of Medical Oncology and Hematology, University Hospital Zurich, Rämistr 100, CH-8091 Zürich, Switzerland; email: thorsten.zenz@usz.ch; Wolfgang Huber, European Molecular Biology Laboratory, Meyerhofstr 1, 69117 Heidelberg, Germany; email: wolfgang.huber@embl.org; and Junyan Lu, Medical Faculty Heidelberg, Heidelberg University, Im Neuenheimer Feld 130.3, 69120 Heidelberg, Germany; email: junyan.lu@uni-heidelberg.de.