Extracellular vesicles and PD-L1 suppress macrophages, inducing therapy resistance in TP53-deficient B-cell malignancies

The tumor microenvironment (TME) is characterized by multiple reciprocal interactions of malignant cells with nonmalignant stroma or immune cells.¹ Particularly, macrophages are at center stage in this network, determining disease progression and therapeutic response, as well as refractory niches.^2-5 In previous work, we and others could show that macrophages exert antibody-dependent cellular phagocytosis (ADCP) and represent the essential mediator of synergy in the administration of chemoimmunotherapy (CIT) of aggressive B-cell lymphoma and multiple myeloma.^2,6-9 This specific combination treatment strongly increases tumor clearance by repolarization of tumor-associated macrophages from a suppressed state to an activated phenotype.^2,10 CIT serves as a standard therapy in numerous B-cell malignancies and is mostly targeted at CD20, such as rituximab in combination with either fludarabine/cyclophosphamide or cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP).^11,12 Despite the success of CIT in the frontline setting, therapy of relapsed or refractory disease, particularly in diffuse large B-cell lymphoma (DLBCL), still imposes a major clinical challenge. Although distinction between germinal center B-cell–like and activated B-cell subtypes and further molecular subclassification opened avenues for tailored treatment strategies, limited molecular markers are available to adapt frontline therapy to overcome primary resistant lymphoma.^13,14 Immune checkpoint inhibition has revealed limited responses as monotherapy, although combination of pembrolizumab with R-CHOP in DLBCL patients was reported to show improved progression-free survival with limited toxicity.^15,16

Mutations of TP53 are present in ∼20% of DLBCL and 10% of CLL, thus representing a significant subgroup of patients.^13,17-19 Loss or mutation of TP53 has been identified as an important mediator of chemoresistance in a variety of malignant entities due to its central coordinating function in multiple cellular stress responses.²⁰ Moreover, loss of TP53 mediates protumorigenic alterations in the tumor microenvironment.^21,22 It remains to be clarified how alterations of the DNA damage pathway in malignant cells functionally affect the complex interactions and outcome of CIT.

Here, we primarily approached disruptions in the DNA damage response (DDR) cascade affecting the synergistic effects of CIT. Our data indicate that, particularly, TP53 functional status determines cellular crosstalk in the TME, phagocytic function, and macrophage-dependent therapeutic response to monoclonal antibodies through release of extracellular vesicles (EVs) and expression of the immune checkpoint PD-L1.

The hMB humanized double-hit lymphoma (DHL) and the chronic lymphocytic leukemia (CLL) Eµ-TCL1/Cd19 (Eµ-TCL1/Cd19-Tp53 wild-type (WT) (Eµ-TCL1) and Eµ-TCL1/Cd19 ^Cre/wt-Tp53 ^fl/fl (Eµ-TCL1/Tp53^fl/fl) murine models were generated as described elsewhere.^2,8,23-26 Detailed explanations can be found in the supplemental Material (available on the Blood Web site). Eµ-TCL1 mice were treated in vivo with cyclophosphamide (10 mg/kg) or phosphate-buffered saline (PBS) when reaching 8 to 12 weeks old and 30% of leukemic cells (CD45⁺CD19⁺CD5⁺ cells) in the peripheral blood.

EVs were extracted from supernatant of cells cultured for 24 hours at a concentration of 16.6 × 10⁶ cells per mL in serum-free CD293 media supplemented with Glutamax 1% (85% to 100% cell viability). Supernatants were subjected to a series of centrifugation steps (1200 rpm, 5 minutes; 2900 rpm, 10 minutes; and 3500 rpm, 20 minutes) and filtered with 0.22 µm polyethersulfone (PES) membrane filters (VWR). EVs were concentrated with the Total Exosome Isolation kit (Thermo Fisher Scientific) according to the manufacturer’s instructions and ultracentrifuged (Type 45 Ti rotor, k-Factor 133, Beckman Coulter) at 100 000 g for 90 minutes at 4°C. Pellets were resuspended in cold PBS (1 µL PBS for 1 mL supernatant), dissolved by pipetting with a 1 mL syringe (Henke-Sass Wolf, Germany) with 26G 7/8-inch needle (Terumo, Germany), and immediately used for ADCPs assays or stored at −80°C. For western blot analysis, EV pellet was directly lysed.

To dissect the mechanism of the synergistic interaction between alkylating chemotherapy and monoclonal antibodies,^2,6,24 we disrupted key components of the DDR pathway in 2 different aggressive B-cell lymphoma models: the hMB humanized DHL model²⁴ and the Myd88 p.L252P-driven DLBCL model.²⁷ In both cases, we generated short hairpin RNA (shRNA)-mediated knockdowns of genes that code for the key DDR mediators: checkpoint kinase 1 (CHK1), CHK2, DNA-dependent protein kinase (PRKDC), MAPK-activated protein kinase 2 (MAPKAPK2), p38-α, ATM, and TP53 (supplemental Figure 1A-B). We exposed shRNA-defined DDR component-deficient cells to mafosfamide, an in vitro alkylating agent, as surrogate for the effect of cyclophosphamide. We then determined the ADCP rate using alemtuzumab or 18B12 anti-CD20 antibody in coculture assays with various sources of macrophages (Figure 1A). We could observe that chemotherapy pretreatment of hMB lymphoma B-cells significantly enhanced their susceptibility for phagocytic engulfment as compared with untreated shRNA-CTRL (shCTRL)-hMB cells (Figure 1B). However, disruption of the DDR led to an abrogation of increased phagocytosis compared with chemotherapy-treated shCTRL-hMB cells (Figure 1B). To employ an independent model of aggressive lymphoma, we used the Myd88 p.L252P-driven DLBCL model, where we confirmed that combination treatment of mafosfamide with 18B12 improved the phagocytosis of shCTRL Myd88 p.L252P cells by primary peritoneal and bone marrow–derived macrophages, as well as by the macrophage cell line J774A.1 (supplemental Figure 1C). In addition, Chk1-, Prkdc-, and MAPKAPK2-deficient lymphoma cells were sensitized to mafosfamide treatment (supplemental Figure 1D). However, Tp53-deficient cells displayed the highest resistance toward CIT (supplemental Figure 1D).

We speculated that the disruption of CIT synergy response when altering DDR components could be a result of insufficient p53 activation. Thus, we used nutlin-3A as an indirect activator of p53 and observed that nutlin-3A similarly modified the ADCP of control, CHK1-, CHK2-, p38a-, and MAPKAPK2-deficient hMB cells but left TP53-deficient lymphoma cells unaffected, which exhibited a significantly reduced phagocytosis level (Figure 1C). Furthermore, nutlin-3A treatment of PRKDC- and ATM-deficient cells did not alter macrophage phagocytic capacity, which correlates with the absence of p53 induction by nutlin-3A (supplemental Figure 1E-F). Altogether, these results indicate that p53-axis functions as a key node in lymphoma cells to influence the response of macrophages to CIT.

To further confirm our observation, we used primary CLL patient samples defined by their TP53 status (Figure 1D; supplemental Figure 1E; supplemental Table 1.3). Notably, macrophages cocultured in the presence of monoclonal antibodies alemtuzumab or rituximab with TP53-mutant CLL cells showed significantly lower ADCP than macrophages cocultured with TP53-WT cells, irrespective of TP53 mutant clone size (Figure 1D; supplemental Figure 1H). Likewise, the effect of TP53 on tumor cell–macrophage crosstalk was investigated using primary malignant MM patient cells cocultured with autologous macrophages in the presence of daratumumab (Figure 1E).²⁸ Macrophages were able to phagocytose ∼50% of MM cells from patients with normal karyotype and del13q14-aberration but displayed a significantly reduced phagocytosis of TP53-deficient MM cells (Figure 1E) that was particularly lower in the higher ratios (10:1 ratio, del13q14DEL vs TP53DEL, P = .025).²⁹ The relevance of TP53 as a central regulator of microenvironment-dependent treatment response in vivo was validated by utilizing the Eµ-TCL1 mouse model of CLL. Here, leukemic Eµ-TCL1/Cd19-Tp53 WT (Eµ-TCL1) and Eµ-TCL1/Cd19^Cre/wt-Tp53 ^fl/fl (Eµ-TCL1/Tp53^fl/fl) were treated with cyclophosphamide. After 24 hours, leukemic cells were obtained for ex vivo ADCP assessment targeted by the antimurine CD20 antibody. In line with our previous observations, in vivo chemotherapy treatment significantly improved the ADCP of Eµ-TCL1–derived leukemia cells, whereas no phagocytic improvement was found in Eµ-TCL1/Tp53^fl/fl–derived cells (Figure 1F). These results indicate that loss of TP53 switches malignant B-cells toward a refractory state against CIT. To verify our hypothesis, we studied the in vivo response to CIT with the humanized mouse model of DHL. Specifically, shTP53-transduced hMB alongside shCTRL cells were transplanted into immunodeficient NSG recipient mice and CIT treatment (CA) was initiated upon disease onset (Figure 1G). CIT resulted in a significantly increased survival of mice transplanted with shCTRL-hMB cells (CA median, 41.5 days vs PBS median, 28 days; P = .0075). However, the response of mice bearing the shTP53-hMB cells to the combination therapy showed a significantly shorter overall survival (CA median, 33.5 days; P = .044) compared with shCTRL-hMB cells (Figure 1G). In summary, TP53 expression serves as a central regulatory node for interactions with macrophages in the TME, particularly in determining response to CIT in B-cell malignancies.

To identify potential p53-dependent pathway alterations induced between malignant B-cells and the TME, we leveraged CD19⁺ cell–depleted TME cells from both Eµ-TCL1 and T Eµ-TCL1/Tp53^fl/fl models under in vivo chemotherapeutic treatment and subjected them to single-cell RNA sequencing (scRNA-seq) analysis. We obtained 21 782 cells across conditions (Figure 2A; supplemental Figure 2A), with 4 distinct macrophage clusters, which were comprised of the largest TME compartment (Figure 2B, left; supplemental Figure 2B) and sharing a largely similar expression of macrophage marker genes across the 4 clusters (Figure 2B, right; supplemental Figure 2C). We therefore decided to combine data from all macrophage clusters together followed by GO enrichment analysis stratified by treatment and genotype to identify the key pathways regulating macrophage function. (Figure 2C). Here, we identified that of the 980 significantly regulated genes in the dataset, 477 were Eµ-TCL1 PBS, 178 Eµ-TCL1 CTX, 165 Eµ-TCL1/Tp53^fl/fl PBS, and 160 Eµ-TCL1/Tp53^fl/fl CTX (Figure 2D, left) and that the GO terms across process/function/component were all largely dominated by Eµ-TCL1 PBS (Figure 2D, right). However, we observed an enrichment for chemotaxis/phagocytic/myeloid GO process terms that were driven by overexpression under Eµ-TCL1 PBS and Eµ-TCL1/Tp53^fl/fl CTX treatment (Figure 2E). Under closer inspection, we observed both reduced expression and percentage of positive cells in Eµ-TCL1 PBS and Eµ-TCL1CTX treatment groups, whereas we observed the opposite under Eµ-TCL1/Tp53^fl/fl PBS and Eµ-TCL1/Tp53^fl/fl CTX treatment (Figure 2F, middle). Furthermore, after CTX/PBS gene expression calculation for the original 4 macrophage clusters in Eµ-TCL1 and Eµ-TCL1/Tp53^fl/fl, respectively, we further observed in the context of Tp53 loss that these chemotaxis/phagocytic/myeloid genes are significantly downregulated in relation to Eµ-TCL1 TME cells, such as Ccr1/2 (Q, 0.001905; Q, 0.032622), Mif/Pycard (Q, 0.009330; Q, 0.004890), and Csfr1/Cd36 (Q, 0.009330; Q, 0.000632) (Figure 2F, right). In all, this scRNA-seq analysis suggests that loss of Tp53 on malignant B-cells under in vivo chemotherapy treatment drives fundamental changes to the macrophage TME, namely loss of phagocytic function and downregulation of chemotaxis.

We identified TP53 not to specifically alter cytokine mediated activations of macrophages (supplemental Figure 3A); however, significant effects by conditioned media of TP53-deficient lymphomas indicated a functional role of soluble crosstalk (supplemental Figure 3B). To identify the mechanism of this specific interaction, we assessed shCTRL and shTP53 tumor B-cells’ proteomic landscape after mafosfamide treatment (1 µM, 6 hours) by a mass spectrometry label-free protein quantification approach (Figure 3A). In total, >4000 proteins were detected in every condition (supplemental Table 3.1). Hierarchical agglomerative clustering clearly differentiated the shCTRL from the shTP53 proteome (supplemental Figure 3A). Our results showed that TP53 depletion profoundly influenced the proteomic profile of lymphoma B-cells, showing a downregulation of 36 proteins and the upregulation of another 60, compared with the control cells (Figure 3A). Interestingly, we observed that mafosfamide treatment differently modified the proteomic profile of TP53-deficient and control cells (supplemental Figure 3B) and produced a shift in the number of up- and downregulated proteins controlled by TP53 (Figure 3A; supplemental Figure 3D-F). Analyzing GO terms of significantly (P-adjust ≤0.1) regulated proteins (supplemental Table 3.2), we observed that loss of TP53 expression mainly correlated to changes on actin cytoskeleton, extracellular plasma membrane, and vesicle-related terms, composed of proteins such as CD9, vacuolar protein sorting 33A (VPS33A), and sorting nexin 9 (SNX9) (Figure 3B-C). Similarly, mafosfamide treatment of shTP53 cells differently regulated proteins associated to actin cytoskeleton and cell adherens junction processes, displaying altered expression levels of EV-associated proteins, such as annexins 1 and 6 (ANXA1 and ANXA6) and the tetraspanin CD81 (Figure 3B-D). Therefore, proteomic analysis indicated that TP53 loss in lymphoma B-cells involves actin cytoskeleton reorganization and alterations in the formation of EVs, which we hypothesize could be responsible for the communication between the malignant B-cells and the TME macrophages.

EVs have emerged as an important mechanism of cellular communication between tumor and stroma cells.^30-32 To assess the relevance of the proteomic findings, we sought to characterize EVs produced by control and TP53-deficient lymphoma B-cells. EVs isolated from lymphoma B-cell supernatants showed typical cup-shaped morphology with a size ranging from 50 to 200 nm (Figure 4A). Loss of TP53 did not alter vesicle size (Figure 4B) or induce significant changes in microRNA (miRNA) abundance (supplemental Table 4.1). However, NTA showed that TP53-deficient cells released significantly higher numbers of EVs than the control cells (Figure 4B). We characterized EV-protein composition by immunoblot, detecting an enriched expression of the EV markers CD81, CD9, CD63, and syntenin compared with the respective cell source, whereas calnexin, a negative marker for EVs, was only present on the cell lysate samples (Figure 4C).³⁰ To determine whether TP53-deficient cell-derived EVs could affect macrophage properties, we exposed different types of macrophages to EVs isolated from shCTRL (shCTRL-EVs) or shTP53 lymphoma B-cells (shTP53-EVs). Our results showed that shTP53-EVs were engulfed by macrophages in a similar way as the shCTRL-EVs (Figure 4D; supplemental Figure 4A). However, macrophages exposed to shTP53-EVs during the coculture exhibited a significantly reduced phagocytic capacity of lymphoma B-cells, whereas they were not significantly affected by shCTRL-EVs (Figure 4E). TP53-dependent inhibition on the ADCP persisted even when the concentration of shTP53-EVs added in the coculture was reduced by 90% (supplemental Figure 4B). In parallel, we excluded EV-mediated effects on lymphoma B-cell viability (supplemental Figure 4C). To clarify if shTP53-EVs were affecting general macrophage phagocytic capacity, we performed a bead-based phagocytosis assay, where macrophages were exposed to shCTRL- or shTP53-EVs 16 hours prior to adding DyLight680-labeled latex beads (supplemental Figure 4D). Bead phagocytosis was unaffected by the presence of shTP53-EVs (supplemental Figure 4D). Thus, shTP53-EVs specifically abrogated Fc receptor– dependent macrophage antitumor effects, such as ADCP. Next, to address the functional role of EVs on the lymphoma B-cell-macrophage crosstalk, we generated lymphoma B-cells unable to release EVs by depleting RAB27A expression via the CRISPR/Cas9 approach (Figure 4F).³³ We proved that RAB27A knockout cells (RAB27A-KO) showed a significantly reduced ability to produce EVs (Figure 4G).

Next, we proved that ADCP targeting shTP53/RAB27A-KO tumor cells was significantly higher than targeting shTP53/WT controls (Figure 4H). To validate our EV-related results in vivo, we used the hMB humanized mouse model of DHL, transplanting mice with shTP53/RAB27A-KO or shTP53/RAB27A-WT hMB cells and using CIT or PBS (control) as treatment. We observed that mice transplanted with lymphoma B-cells unable to release EVs displayed significantly longer overall survival after CIT (CA⁺ shTP53/RAB27A-KO median equals 21 plus or minus 0.48 days vs CA⁺ shTP53/RAB27A-WT median equals 17 plus or minus 0.47 days; P ≤ .0001) (Figure 4I). To further support our hypothesis, we also confirmed a significant increase in total number of EVs produced by TP53-deficient primary murine CLL (Eµ-TCL1) cells (Figure 4J) and an upward trend in TP53-mutant primary CLL cells vs WT cells (Figure 4K). In conclusion, we show that EV formation is enhanced by loss of TP53 and leads to suppression of macrophage effector function toward malignant cells.

Given the diminished phagocytic capacity of macrophages in the presence of EVs produced by TP53-deficient malignant B-cells, we considered the possibility that the EVs secreted by shCTRL and shTP53 cells displayed different expression levels of “dońt eat me” and related immune checkpoint molecules. Our data confirmed the absence of CD47 and the low expression of CD200, respectively, but the presence of PD-L1 on leukemic B cells-EVs (Figure 5A). Next, by using the anti–PD-L1 antibody atezolizumab, PD-L1 was blocked on the EVs before their addition to the coculture of macrophages and leukemic B-cells (Figure 5B). Remarkably, neutralizing PD-L1 on the shTP53-EVs, we could restore macrophage antitumor capacity assessed by ADCP. This observation was confirmed by using EVs obtained from PD-L1–knockout (PD-L1–KO) hMB cells, which do not impair macrophage antitumor function (Figure 5C-D). Thus, PD-L1 expression on shTP53-EVs mediate suppression of macrophage phagocytic activity through PD1 expressed on macrophages (supplemental Figure 5A-B).

Likewise, shTP53-dependent PD-L1 cell surface expression was addressed in the context of alkylating treatment in vitro (Figure 5E). We observed significantly higher PD-L1 protein expression in shTP53 cells than in control cells, with PD-L1 upregulation even slightly increased after chemotherapy exposure (Figure 5E). However, CD47 and CD200 expression showed unaltered expression in all conditions (supplemental Figure 5A). We could corroborate that primary TP53-deficient malignant B-cells also expressed higher PD-L1 protein levels than the respective controls, shown by analyzing samples from Eµ-TCL1 mice and from CLL patients (supplemental Figure 5D-E).

This observation prompted us to evaluate the relevance of the PD1/PD-L1 axis in the resistance to CIT. We performed an ADCP assay in the presence of anti-PD1 antibody in combination with chemotherapy (Figure 5F). The addition of anti-PD1 antibody to the coculture significantly improved the ADCP of both shCTRL and shTP53 hMB cells. This finding was confirmed by using the Myd88 p.L252P-driven DLBCL model, where PD1/PD-L1 inhibition restored the resistance to chemotherapy observed on the ADCP of TP53-deficient tumor cells (supplemental Figure 5C). In addition, the use of the triple therapy approach (anti-PD1 antibody, cyclophosphamide, and alemtuzumab) in the humanized mouse model of DHL with TP53-deficient lymphoma cells revealed a significantly increased overall survival of the mice compared with cyclophosphamide/alemtuzumab combination therapy (CA median equals 11 plus or minus 0.54 days vs CA plus anti-PD1 median equals 13 plus or minus 1.40 days; P = .023) (Figure 5G). To clarify the role of PD1/PD-L1 axis in the abrogation of macrophage antitumor functions on TP53-deficient lymphomas, we analyzed the ADCP of shTP53/PD-L1–KO hMB cells (Figure 5H-I). We demonstrated that PD-L1 deletion favored the ADCP of TP53-defcient cells in vitro (Figure 5H), showing a higher phagocytic response of macrophages upon CIT (Figure 5I). Finally, we evaluated the role of PD-L1 in the resistance to chemotherapy in vivo by injecting shTP53/PD-L1–KO hMB cells into NSG mice (Figure 5J). Our results showed a significantly prolonged survival of shTP53/PD-L1–KO vs shTP53/PD-L1–WT lymphoma-bearing mice (PBS⁺PD-L1–KO median equals 10 plus or minus 0.55 days vs PBS⁺PD-L1–WT median equals 8 plus or minus 0.61 days; P = .007). Importantly, CIT response was highly effective in mice injected with shTP53/PD-L1–KO tumor cells (Figure 5J) (PBS⁺PD-L1–WT median equals 10 plus or minus 0.58 vs CA⁺PD-L1–KO median equals 14 plus or minus 1.67 days; P = .005). Thus, CIT treatment in combination with PD1/PD-L1 inhibition improved TP53-induced therapy resistance. In conclusion, the upregulation of PD-L1 on TP53-deficient tumor cells appears to be a resistance mechanism amenable for therapeutic targeting in B-cell malignancies in addition to CIT.

The concept of p53 as the central node in the DDR and “guardian of the genome” was recently expanded toward cell-nonautonomous effects by modulation of the TME by p53-deficient cancer cells.^34,35 The functional impact of p53 on the TME has been shown in several tumor models, with the recovery of its function leading to the restoration of immune surveillance in liver cancer.^36,37 Here, we show that in B-cell lymphoma cells, TP53 serves as a regulatory switch of the function of macrophages in the lymphoma TME.

In previous work, we could determine the essential role of cytokine release from lymphoma cells upon alkylating chemotherapy for sensitization of macrophage effector functions as the synergistic mechanism in CIT. ^2,6 Now we show that an intact DDR is required to activate this mechanism of synergy and that p53 function serves as a central node of the DDR to control resistance against CIT. Corresponding to this observation, loss or mutation of TP53 is the most important resistance mechanism to CIT in CLL.^12,38,39

The regulatory effect of TP53-deficient lymphoma cells on macrophage effector function was caused by a profoundly elevated secretion of EVs and an increased PD-L1 expression. Interestingly, we identified a central connection of a p53-dependent release of EVs and the expression of PD-L1 on cells and functional impact of PD-L1 carried by EVs

Multiple mechanisms of controlling PD-L1 expression have been described in malignant cells. Likewise, TP53-dependent control of PD-L1 expression has been described in small-cell lung cancer cells, involving miRNA-34a regulation of the PD-L1 3′UTR.^40,41 Furthermore, TP53 status has been hypothesized to predict response to checkpoint inhibitor therapy in cancer.^42,43 The functional relevance of immune oncology drugs targeting the PD-1/PD-L1 axis have so far mainly been attributed to T-cell activation and adaptive immunity.^44,45 However, there is similar evidence that the PD1/PD-L1 checkpoint regulates macrophages in the TME.⁴⁶ We have now identified PD-L1 to inhibit macrophages to effectively engulf lymphoma cells in the therapeutic setting, pointing to PD-L1 as a novel resistance mechanism for CIT regimens.

In this line, expression of PD-L1 was described as an integral component in a biomarker score indicating poor response to CIT of lymphoma.⁴⁷ In DLBCL, patients with high expression of PD-1/PD-L1 on T cells and macrophages had significantly poorer survival after R-CHOP.⁴⁸ In line with our data, it has been recently shown that loss of TP53 increases PD-L1 in a murine lymphoma model of DLBCL, inducing immune evasion, which could be overcome with PD1-blockade.⁴⁹ 

Importantly, our data indicates that PD-L1 dependent immune suppression may occur independently of direct cell-cell interactions. We observe the increase of released EVs exposing PD-L1 to mediate the suppression of phagocytic function of effector macrophages. Notably, we see this novel phenomenon both in DLBCL and CLL murine models with genetic dysfunction of TP53, as well as in CLL patient samples. Using genetic targeting of PD-L1 on EVs, applying EV-deficient cells and immune checkpoint inhibitor blockade of EVs, we provide direct evidence of the decisive functional role of PD-L1⁺ EVs in regulating macrophage effector function and treatment response to CIT.

TP53 has been described to regulate EV secretion.⁵⁰ More specifically, it has been proposed that mutant TP53 colon cancer cells reprogram macrophages toward tumor-supporting macrophages via miRNA (miR)-1246 delivered by EVs.²² Colorectal cancer−derived EVs were described to regulate polarization of tumor-associated macrophages by the miR-145, which is transcriptionally induced by p53.^51,52 However, we do not see significant alterations in the miRNA profiles due to TP53 loss.

As for the context of chemotherapy and DNA damage, EV-based metabolic reprogramming by tissue-infiltrating macrophages underlines the role of EVs in cellular crosstalk.⁵³ Also, an increase of EV-shedding was demonstrated to promote melanoma growth after chemotherapy.⁵⁴ Melanoma-derived exosomes were identified to carry the immune checkpoint ligand PD-L1 and thereby contribute to immunosuppression.⁵⁵ Recently, follicular lymphoma–derived EVs have been identified to promote polarization of the bone marrow stromal cell niche and modulate the TME toward tumor supportive function.⁵⁶ So far, no specific therapeutic concept has been identified that counteracts EV-mediated effects. However, immune checkpoint inhibitors may offer an opportunity to target EV-mediated checkpoint control. In line with our hypothesis, we could effectively improve CIT by using an anti–PD-L1 antibody, atezolizumab, thereby significantly enhancing macrophage-mediated phagocytosis of lymphoma cells.

The conclusions from our findings are manifold. First, the results support the clinical evaluation of CIT with checkpoint inhibitors in order to overcome PD-L1–associated resistance. This strategy seems promising, as a combination of chemotherapy with immune checkpoint inhibitors was recently reported to be safe and to yield complete response rates of 77% in a phase 2 trial combining pembrolizumab with R-CHOP.¹⁶ Moreover, the loss of p53 function should call for innovative, intensified therapeutic strategies in lymphoma, including immunotherapies and immune checkpoint inhibitors. Particularly, combinations of targeted therapies such as immune modulators or BTK- or BCL2 inhibitors should be examined with regard to EV and PD-L1–mediated therapeutic responses.^8,29,57 Finally, the release of immunosuppressive EVs can be recognized as a novel resistance mechanism in TP53-mutated lymphoma that may need to be monitored in future clinical studies.

The authors are indebted to their patients who contributed tissue and blood samples to this study. The authors are grateful for technical assistance from Yvonne Meyer and Sophie Neuber, support from Katrin Reiners, and the CECAD imaging and animal facilities.

This work was supported by the German research foundation (DFG) KFO 286 and CRC1530. C.P.P. was supported by the ‘Foerderprogramm Nachwuchsforschungsgruppen NRW 2015–2021, CAP Program of the Center for Molecular Medicine Cologne, and a research grant by Gilead Sciences. H.C.R. was supported by the German-Israeli Foundation for Research and Development (I-65-412.20-2016), the Deutsche Forschungsgemeinschaft (KFO-286-RP2), the Deutsche Jose Carreras Leukämie Stiftung (R12/08), the Else Kröner-Fresenius Stiftung (EKFS-2014-A06 to H.C.R., 2016_Kolleg.19), the Deutsche Krebshilfe (1117240 and 70113041), and the German Ministry of Education and Research (BMBF e:Med 01ZX1303A). H.B. was supported by Wilhelm-Sander Foundation and by the German Research Foundation (DFG BR 4775/2-1).

Contribution: E.I. and C.P.P. contributed to study design; E.I., D.V., B.S., J.L.N., V.B., N.N., J.S., M.F., T.G., S.B., D.B., L.L., M.M., R.B., O.M., R.N., M.K., G.K., H.B., N.L., E.M., and N.N. contributed to data analysis and acquisition; B.C., J.N., S.B., M.N., M.P., R.R.-R., and M.K. contributed to bioinformatic analyses; H.C.R., E.M., L.F., and G.K. contributed analytical tools; C.P.P., B.C., N.N., H.B., and M.H. contributed clinical samples and annotation; C.P.P. contributed to study supervision and funding; and E.I., D.V., S.B., and C.P.P. contributed to manuscript preparation.

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

Correspondence: Christian P. Pallasch, Department I of Internal Medicine, Center for Integrated Oncology (CIO), University Hospital Cologne, Kerpener Str 62, 50937 Cologne, Germany; e-mail: christian.pallasch@uk-koeln.de.