Introduction

The complement system is a key part of innate immunity, eliminating foreign material by inducing complement dependent cytotoxicity (CDC) via formation of the membrane attack complex (MAC). CDC can also be induced by antibodies, bridging innate and adaptive immunity. Upon binding to (membrane) surfaces, antibodies can hexamerize to provide a docking platform for the hexavalent complement protein C1q, activating the classical complement pathway leading to MAC formation [1]. This pathway is exploited in cancer treatment and is considered a mechanism behind the anti-tumor effects of therapeutic antibodies like the CD20-targeting antibody rituximab for B-cell chronic lymphocytic leukemia and B-cell non-Hodgkin lymphoma and the CD38-targeting antibody daratumumab for multiple myeloma [2,3,4]. To further increase therapeutic activity, Fc-domain engineering has been explored to augment antibody-mediated CDC capacity [5, 6]. An effective strategy to enhance CDC is the HexaBody® technology, whereby a single point mutation (such as E430G) in the antibody Fc-domain enhances antibody hexamerization and subsequent C1q binding [7]. A complete understanding of the antibody-induced CDC response in cancer, including potential resistance mechanisms, could enable more effective use of CDC-inducing antibodies.

Most studies exploring resistance to antibody-mediated CDC have focused on membrane regulators, specifically complement regulatory proteins (CRPs) CD46, CD55, and CD59, or disruptions in MAC assembly and stabilization [8]. This focus is driven by the prevailing belief that CDC is inevitable once enough MAC pores form on the cell surface. Intracellular events such as an increased calcium influx, enhanced expression of pro-apoptotic proteins, and mitochondrial swelling/damage, have been linked to CDC sensitivity, but their precise roles in cell death and resistance thereof remain unclear [8, 9]. Some reports suggest the occurrence of intracellular resistance to CDC [10,11,12], but the underlying mechanisms remain unidentified.

Our recent analyses of malignant cells from B-cell lymphoma patients revealed a clear heterogeneity in the CDC-susceptibility among diffuse-large B-cell lymphoma (DLBCL) samples [11]. This heterogeneity could not solely be attributed to differences in target antigen or CRP expression, indicating possibly intracellular mechanisms of CDC regulation. Understanding these mechanisms could shed light on why about one-third of DLBCL patients develop resistance to R-CHOP, the standard first-line chemoimmunotherapy, which includes the CDC-inducing antibody rituximab [13, 14].

Therefore, we aimed to elucidate the role of intracellular factors in CDC resistance in DLBCL models using the potent CDC-inducer DuoHexaBody-CD37, a bispecific antibody that harbors the E430G hexamerization-enhancing mutation and targets two non-overlapping epitopes on CD37 [11, 15].

Our findings reveal that mitochondrial damage and subsequent reactive oxygen species (ROS) production are critical components of CDC-induced cell death. Resistance was associated with reduced mitochondrial damage and ROS levels, and was linked to altered cytoskeletal dynamics and mitochondrial rearrangements. These insights suggest that intracellular adaptations can enable cancer cells to evade antibody-induced CDC, offering a new perspective on mechanisms of therapeutic resistance.

Methods

An overview of the key resources is provided in Supplemental Table 1. Please see Supplemental Materials for detailed “Methods”. All methods were performed in accordance with the relevant guidelines and regulations.

Flow cytometry and – BLI-based CDC assays

CDC assays were performed by incubating DLBCL cell lines (DSMZ) or patient lymph node samples with 20% normal human serum (pooled from 4–6 donors) and DuoHexaBody-CD37 or rituximab for 45 min at 37 °C. Pre-treatments were maintained during the assay without washing except for Mdivi-1/M1, where cells were pre-treated in flasks for 3 days before washing and plating in 96-well plates. Viability was assessed by flow cytometry using a live/dead marker, and cytotoxicity was calculated as 100%-viability. For bioluminescence-based viability, Luc-GFP+ cells were incubated with beetle luciferin (125 µg/mL) for 20 min and imaged using a Promega™ GloMax® Plate Reader.

Generation of the U2932-R cell line

U2932-WT cells were transduced with a lentiviral vector (pLM) to stably express the human CD37 gene. After transduction, cells were treated with DuoHexaBody-CD37 (10 µg/mL) and sorted via FACS to isolate viable cells with CD37 levels similar to U2932-WT, generating the CDC-resistant line U2932-R-CD37+, referred to as U2932-R. U2932-R cells were re-challenged with DuoHexaBody-CD37 (10 µg/mL) and 20% NHS every 4–6 weeks.

CRISPR-Cas9 screen in the CDC-sensitive RI-1 cell line

RI-1 cells were stably transduced with Lenti-Cas9-2A-Blast and selected with 10 µg/mL blasticidin for 7 days. The Cas9-expressing RI-1 line was then transduced in triplicate with the TKOv3 knockout library (MOI ~ 0.3), which targets 18,053 genes with 71,090 sgRNAs. Three days post-transduction, cells were selected with 2 µg/mL puromycin for 4 days, treated three times with DuoHexaBody-CD37 (at IC90) and 20% NHS, and cultured until CD37 expression reappeared (7–10 days). The resulting CD37+ cells were isolated by FACS and stored at −80 °C.

Genomic DNA from these cells was amplified, indexed, and sequenced on a HiSeq 4000. Reads were mapped to the library, and analysis was performed with DrugZ to rank genes by median sgRNA counts across replicates. To focus on CDC resistance, the top 15 gene hits from the CD37− population were excluded. See Supplemental Methods for further details.

Flow cytometry-based mt-Keima assay

U2932-WT and U2932-R cells were transduced with mt-Keima-expressing lentivirus. Mt-Keima expressing cells were treated with DuoHexaBody-CD37 (1 or 10 µg/mL) for specified times in CDC assays. After washing, cells were stained with a live/dead marker and analyzed by flow cytometry. Mitophagy was detected by the red-to-green signal ratio (550 nm/440 nm) in cells with mitolysosomes.

Mitotracker dyes

Cells were washed and incubated for 20 min with MitoTracker™ Green FM, MitoTracker™ Deep Red FM, and MitoSOX Red in serum-free medium, following the manufacturer’s protocol, in duplicates or triplicates. After washing, cells were incubated with a live/dead marker and, if applicable, experiment-specific cell markers, followed by flow cytometry analysis.

CENCAT

CENCAT (CEllular eNergy metabolism by protein synthesis via Click chemistry and Translation inhibition) was performed as described [16]. After CDC assays, 0.5 × 10⁶ cells/well were plated in 96-well plates and treated for 15 min at 37 °C with DMSO (control), 2-DG (100 mM), oligomycin (1 µM), or both (DGO). β-ethynylserine (βES, 500 µM) [17] was added for 30–60 min, with a βES-negative control included. Cells were stained with a fixable live/dead marker, fixed, permeabilized (eBioscience™ Foxp3 Fix/Perm kit), and incubated with AFDye 405 Azide Plus click reagent (30 min, RT; see Supplemental Table 1). Samples were washed and analyzed by flow cytometry.

Wide-field microscopy to analyze mitochondrial morphology

U2932 cells were seeded on poly-L-lysine-coated 8-well plates (Ibidi) and incubated overnight at 37°C. After washing, cells were treated with 80 nM CMXRos probe for 30 min, fixed in 3.7% paraformaldehyde, and mounted in Mowiol. Mitochondrial morphology was visualized with a Nikon ECLIPSE Ti2 microscope (Apo λ 100× Oil; NA = 1.45, WD = 130 µm) and categorized as fragmented, mixed, or tubular [18]. At least 30 cells per slide were analyzed in duplicate across two independent experiments.

Patient samples

Lymph node biopsies of DLBCL patients and corresponding clinical data were collected after obtaining informed consent and in accordance with the medical ethical committee of Amsterdam UMC and the code of conduct for medical research developed by the Council of the Federation of Medical Scientific Societies (FEDERA).

Lymph node biopsies were mechanically disrupted into small fragments and the supernatant was filtered using a 70 µM Easy Strainer. The diffused cells, including mostly (malignant) B-cells and T-cells, were cryopreserved in liquid nitrogen until use. Prior to experimental assays, cells were thawed and metabolically activated by 24-h incubation with soluble human recombinant MEGACD40L® Protein. For transcriptomic analysis, DLBCL patient samples were FACS-sorted following metabolic activation to isolate CD19−, CD3⁺, and 7-AAD− (viable) cells.

RNA-sequencing

Total RNA was isolated from U2932-WT, U2932-post 1× CDC treatment, U2932-R cells, and FACS-isolated CD19+CD3− B-cells from DLBCL patients activated with CD40L. mRNA libraries were prepared using the KAPA mRNA HyperPrep Kit and sequenced on the Illumina NovaSeq 6000. Reads were processed with fastp for trimming and filtering [19]. Alignment and quantification to the GRCh38 reference genome using ENSEMBL release 110 gene annotations was performed using STAR v2.7.8a with default parameters [20]. The R package DESeq2 (v1.42.0) was used to determine differentially expressed genes (DEGs; adjusted for p-value of <0.05) [21]. DEGs were analyzed for enrichment in Gene Ontology (GO) biological processes and molecular functions [22,23,24], using GO terms containing 15–500 genes. Gene set enrichment analysis (GSEA) was performed using the fgsea R package (v1.28.0) with GO Biological Process and Molecular Function gene sets from GSEA MsigDB [25, 26].

Gene expression data R-CHOP treated DLBCL patients

Raw.CEL files and clinical data from Lenz et al. [27] (GEO: GSE10846) and processed data from Chiche et al. [28] (Mendeley data repository [28]), were analyzed, focusing on R-CHOP treated patient samples (n = 233). Differential probeset testing was performed using the R package limma [29], with multiple testing correction by Benjamini-Hochberg [30]. Functional pathway analysis of DEGs was conducted using GO enrichment analysis powered by Panther [22,23,24], including GO terms with 15–500 genes. Probesets without gene ID annotation were excluded.

Western blot analysis

Cell pellets, obtained from cell lines either at baseline or following conventional 45-min CDC assays, were washed in ice-cold PBS and fractionated into mitochondrial and cytosolic components using the Mitochondria/Cytosol Fractionation Kit with sonication for homogenization [31]. Protein concentration was measured with the Pierce™ BCA Protein Assay Kit. Samples were denatured with β-mercaptoethanol, boiled, and equal protein amounts were loaded onto Mini-PROTEAN TGX Gels (Bio-Rad). After electrophoresis, proteins were transferred to 0.45 µm Immobilon-FL PVDF membranes, blocked, and incubated overnight with primary mouse antibodies at 4 °C. After washing, membranes were incubated with IRDye-conjugated secondary antibodies, scanned using UVITC Q9 Alliance software, and quantified with ImageJ, normalizing to cytosolic tubulin.

Results

CDC is mediated via mitochondrial damage and ROS

To identify the intracellular mechanisms driving sensitivity to antibody-mediated CDC, we performed a genome wide CRISPR-Cas9 library screen in the DLBCL cell line RI-1, which is highly susceptible to DuoHexaBody-CD37-mediated CDC. Following library transduction, cells were exposed repeatedly to DuoHexaBody-CD37 in CDC assays at 90% of maximal inhibitory concentration (IC90), resulting in the near-complete elimination of cells (Fig. 1A). As anticipated, 80–92% of the surviving cells exhibited no or low expression of CD37, indicating that sgRNA targeting of CD37 was the main contributor to CDC resistance (Supplemental Table 3). We focused on the remaining CD37+ resistant cells and identified the top 15 sgRNAs enriched exclusively in the CD37+ fraction. These hits included genes associated with oxidative stress (KEAP1), epigenetic regulation (EP300, TAF5L, TAF6L, MTA2, SUPT20H), and key regulators of cellular functions, such as the transcription factors FOXK1 and FOXO1, and the tumor suppressor PTEN (Fig. 1B).

Fig. 1: CRISPR-Cas9 library screening identifies a role for FOXK1, FOXO1, KEAP1, and PTEN genes, and mitochondrial damage and ROS production in CDC.
figure 1

A Schematic overview of the CRISPR-library screen, performed in triplicate using TKOv3-in-pLCKO lentiviral library in the CDC-sensitive, Cas9-expressing DLBCL cell line RI-1. Post-screening, CD37+ cells were sorted and processed for gDNA isolation, sgRNA-enrichment, and sequencing. B Top 15 gene hits from the screen in CD37+ cells, identified by DrugZ-analysis and ranked by median count per gene. C CD37 surface expression (median fluorescence intensity; MFI) and D DuoHexaBody-CD37-induced CDC (%; at 0.5 µg/mL) in RI-1 cells and the CRISPR-targeted FOXK1, FOXO1, KEAP1, and PTEN variants (n ≥ 4; data shown as mean ± SD). E Detection of dysfunctional mitochondria in RI-1 cells after 5 min of DuoHexaBody-CD37 (0.5 µg/mL) CDC induction, defined by maintained mitochondrial mass (MitoTracker Green; MTG) and decreased membrane potential (MitoTracker DeepRed; MTDR), compared to no antibody control. Representative dot-plots are shown (n = 3). F Fold-change mitochondrial ROS (mtROS) levels upon 5 min of incubation with DuoHexaBody-CD37 (0.5 µg/mL) in CDC-assays in RI-1 cells, measured by MitoSOX (MFI) in functional and dysfunctional mitochondria, and normalized to no antibody control (n = 3; mean ± SD). G DuoHexaBody-CD37-induced CDC (%; at 0.5 µg/mL) in RI-1 cells after 2-h pre-treatment with ROS-scavenger n-acetylcysteine (NAC; 10 and 20 µM) (n = 4; mean ± SD). H Detection of dysfunctional mitochondria and I fold-change mtROS induction in U2932 cells upon 5 min CDC assay with 1 µg/mL DuoHexaBody-CD37 (n = 3). Representative dot-plot of dysfunctional mitochondria detection is shown. J Effect of NAC pre-treatment on DuoHexaBody-37-induced CDC in U2932 cells (n = 3). K Fold-change in mtROS levels in dysfunctional mitochondria upon 5 min CDC assay with DuoHexaBody-CD37 (0.5 µg/mL) in RI-1 cells and gene-targeted variants (n = 3). Data are shown in a Box and Whiskers plot. N indicates independent experiments, each performed in replicate or as triplicates; the average value per experiment was used for analysis. Statistical analyses were performed using a one-way ANOVA with post-hoc Tukey’s multiple comparison test.

Full size image

While mitochondrial damage and ROS have been implicated in CDC, whether they are key factors for cell death has not been demonstrated. The identification of KEAP1, which regulates the antioxidant response through the Nrf2 pathway [32, 33], as top hit implies that oxidative stress is indeed central to CDC. We validated the role of KEAP1 by demonstrating that its targeting did not alter CD37 expression (Fig. 1C) but significantly reduced CDC sensitivity (Fig. 1D). Furthermore, using MitoTracker dyes, we monitored mitochondrial health and ROS production upon DuoHexaBody-CD37-induced complement activation (hereafter referred to as “antibody-induced complement activation” or AICA) in parental RI-1 cells. Within the first 5 min of AICA, dysfunctional mitochondria were detected in the viable cell fraction in both cell lines, characterized by retained mitochondrial mass but a loss of mitochondrial membrane potential (Fig. 1E). Cells exhibiting dysfunctional mitochondria produced significantly higher ROS levels, detected by MitoSOX Red in RI-1 cells (Fig. 1F), and scavenging ROS with N-acetylcysteine significantly inhibited CDC (Fig. 1G). Similar results were observed for another DLBCL cell line U2932 (Fig. 1H–J). As anticipated, KEAP1-targeted RI-1 cells exhibited a visible, albeit non-significant, reduction in the ROS burst within AICA-induced dysfunctional mitochondria (Fig. 1K). These findings highlight the critical role of mitochondrial dysfunction and ROS production in driving CDC. Consistent with this, mitochondrial dysfunction was also observed following AICA-induced by the clinically used CD20-targeting antibody rituximab in two CD20-expressing DLBCL cell lines (LY1 and LY18), indicating that this mechanism is a general feature of AICA (Supplemental Fig. 1).

Beyond KEAP1, we also identified FOXK1, FOXO1, and PTEN as candidate regulators of CDC due to their known roles in metabolic pathways. Similar to KEAP1, targeting FOXK1 and PTEN significantly reduced CDC without impacting CD37 surface expression (Fig. 1C, D). Notably, CDC resistance was most pronounced at lower antibody concentrations and could be (partially) overcome by increasing antibody levels or by pre-incubating cells with antibodies prior to adding NHS, indicating that resistance was not absolute (Supplemental Fig. 2A). Targeting FOXK1 and PTEN also reduced ROS production during AICA (Fig. 1K). Collectively, these findings indicate that mitochondrial damage and ROS are key factors in CDC and link ROS modulation with CDC resistance.

Comparison of the transcriptome of WT vs resistant U2932 cells shows a role for cytoskeletal dynamics in CDC resistance

We next aimed to understand the role of the mitochondria and ROS in cells that exhibit inherent CDC resistance. To this end, we generated a fully CDC-resistant DLBCL cell line, U2932-R, from the partially (~ 40–50%) CDC-resistant U2932 (-WT) cell line, by repeated AICA exposure. CD37 expression levels were sustained by lentiviral-mediated CD37 gene transduction (Fig. 2A). Despite comparable levels of MAC formation, U2932-R cells remained viable, indicating intracellular resistance mechanisms (Fig. 2B, C). While U2932-R cells exhibited mitochondrial damage, they did not mount a detectable ROS burst, and prolonging AICA to 45 min eliminated the dysfunctional mitochondria, suggesting efficient mitochondrial quality control likely through enhanced repair or removal (Fig. 2D, E). These findings align with our CRISPR-Cas9 screen in RI-1, linking CDC resistance to ROS neutralization via mitochondrial repair. Mitochondrial quality control encompasses continuous cycles of fission and fusion, and mitophagy. Fission fragments (damaged) mitochondria, fusion restores function by merging healthy or partially damaged mitochondria, and mitophagy removes dysfunctional mitochondria in a fission-dependent manner [34].

Fig. 2: Intrinsic CDC resistance of U2932-R is associated with alterations in mitochondrial phenotype and an altered response to therapy-induced mitochondrial damage.
figure 2

A CD37 expression in U2932-WT and U2932-R cells, quantified as antibody binding molecules per cell using the QIFIKIT (n = 3). B DuoHexaBody-CD37-induced CDC (%; at 10 µg/mL) in U2932-WT versus U2932-R in conventional CDC assay (n = 3). C Maximal membrane attack complex (MAC) formation detected by C5b-9 binding in U2932-WT (left) and U2932-R (right) cells after incubation with 10 µg/mL DuoHexaBody-CD37. Peak C5b-9 levels were observed after 5 min in U2932-WT cells, reflecting rapid CDC and subsequent loss of viable cells. In contrast, U2932-R cells retained viability and C5b-9 signal at 45 min. Representative histograms are shown (n = 3). D Detection of dysfunctional mitochondria in U2932-R cells after 5 min and 45 min of DuoHexaBody-CD37 (1 µg/mL) CDC induction, compared to no antibody control. Representative dot-plots are shown (n = 3). E Fold-change mitochondrial ROS (mtROS) induction in U2932-WT cells following 5 min CDC assay with DuoHexaBody-CD37 (1 µg/mL), measured separately in functional and dysfunctional mitochondria, normalized to no antibody control (n = 3). N indicates independent experiments, each performed in replicate or as triplicates; the average value per experiment was used for analysis. Data are shown as mean ± SD. Statistical analysis was performed using a paired t-test for (A, B) and a one-way ANOVA with post-hoc Tukey’s multiple comparison test for (E).

Full size image

To investigate the differential regulation of U2932-R cells in countering mitochondrial damage compared to U2932-WT cells, we analyzed their transcriptomes. Among the >3000 DEGs (p < 0.05) (Fig. 3A and Supplemental Table 4), the top 50 DEGs, all downregulated in U2932-R cells relative to U2932-WT, included genes not associated with mitochondria but involved in calcium signaling (e.g., NCALD, UNC13C, CACNA2D3, CADPS, STC2)—a pathway implicated in CDC—and importantly, cytoskeleton regulation (e.g., MYO6, PDLIM3, MYO1B, TNNI, GPHN, RAP2B) (Fig. 3B). Notably, MYO6 (Myosin VI), a cytoskeleton motor protein, promotes actin cage assembly around damaged mitochondria, facilitating mitophagy and preventing reintegration of damaged mitochondria into the mitochondrial network [35,36,37,38,39], linking cytoskeleton to mitophagy regulation. GO enrichment analysis of 621 DEGs with log2FoldChange < −1.5 in U2932-R cells indicated significant enrichment in actin-binding-, actin filament binding-, and microfilament motor activity genes (Fig. 3C), while no significant GO terms were enriched among the 144 genes upregulated in U2932-R cells. GSEA further revealed significant negative enrichment of the actin-related pathway GOBP_ACTIN_FILAMENT_ORGANIZATION (NES = −1.40, FDR = 0.037), confirming coordinated downregulation of actin filament organization genes in U2932-R cells compared to U2932-WT.

Fig. 3: CDC-resistant U2932-R show decreased cytoskeleton and specifically actin-binding genes, impacting CDC.
figure 3

A Volcano plot highlighting differentially expressed genes (DEGs; red) between U2932-WT and U2932-R cells. B Heapmap of the top 50 DEGs between U2932-WT and U2932-R ranked by log₂ fold change. C GO-term enrichment analysis of 621 DEGs (adjusted p < 0.05, log₂ fold change < −1.5) showing the top 10 enriched pathways for biological process (blue) and molecular function (red), sorted by false discovery rate (FDR).

Full size image

Together, our data indicate that transcriptional reprogramming of actin dynamics is a key feature of CDC resistance, and provide a link between cytoskeletal changes and mitochondrial quality control.

Targeting actin alters the outcome of CDC

To determine if differences in actin-associated gene expression translated to altered actin protein levels, we performed Western blot analysis and found significantly lower actin levels in U2932-R cells compared to U2932-WT cells, relative to tubulin (Fig. 4A, B). This protein reduction may reflect the shift in cytoskeletal dynamics indicated by our transcriptomic data. We then investigated the link between actin downregulation and mitochondria by analyzing actin levels in cytosolic fractions and mitochondria-enriched fractions following AICA treatment. In both U2932-WT cells and U2932-R cells, AICA specifically reduced actin levels in the mitochondrial fraction, without affecting tubulin (Fig. 4A, C, D), suggesting a targeted effect on mitochondrial actin. To validate the functional role of actin in CDC, we pre-treated U2932-WT and U2932-R cells with jasplakinolide, an inducer that stabilizes and promotes actin polymerization. This treatment enhanced CDC in U2932-WT cells and, importantly, re-sensitized U2932-R cells to CDC (Fig. 4E, F). Conversely, pre-treatment of CDC-sensitive cell lines LY18 and LY1 cells with cytochalasin-D, an inhibitor of actin polymerization, significantly inhibited rituximab-mediated CDC (Fig. 4G), indicating that a dynamic, polymerized actin network is critical for CDC. The effects of actin modulation were not due to changes in target antigen expression (Supplemental Fig. 3). Together, these findings support that sufficient and dynamic mitochondrial actin polymerization facilitates CDC.

Fig. 4: Antibody-induced complement activation leads to mitochondrial actin downregulation, impacting CDC.
figure 4

A Western blot analysis of actin in (C) cytosolic and (M) mitochondrial-enriched fractions in U2932-WT and U2932-R cells, at baseline and upon DuoHexaBody-CD37-induced (1 µg/mL) CDC. Tubulin is shown as cytosolic loading control. Representative blot shown (n = 3). B–D Quantification of actin band intensity normalized to cytosolic tubulin and expressed relative to cytosolic U2932-WT (set to 1). Shown are baseline levels (left), post-CDC levels in U2932-WT (middle), and post-CDC levels in U2932-R (right). E Cytotoxicity (%) achieved by DuoHexaBody-CD37 1 µg/mL (left) or 10 µg/mL (right) in U2932-WT in conventional CDC assays with or without 1-h jasplakinolide (8 µM) pre-treatmen (n ≥ 3). F Cytotoxicity (%) achieved by DuoHexaBody-CD37 (10 µg/mL) in U2932-R in conventional CDC assays with or without 1-h jasplakinolide (8 µM) pre-treatment (n ≥ 3). G Cytotoxicity (%) achieved by rituximab (RTX; 10 µg/mL) in LY18 (left) and LY1 (right) with or without a 30-min Cytochalasin D (5 µM) treatment (n ≥ 3). N indicates independent experiments, each performed in replicate or as triplicates; the average value per experiment was used for analysis. Data are shown as mean ± SD and statistical analyses were performed using one-way ANOVA with post-hoc Tukey’s multiple comparison test.

Full size image

Mitochondrial rearrangements in U2932-R cells

Cytoskeletal dynamics, essential for various cellular processes, play a significant role in mitochondrial rearrangements; actin in particular is critical for mitochondrial fission and mitophagy and encages damaged particles for mitophagy [37, 38, 40]. Given the reduced mitochondrial actin levels in U2932-R cells, we examined how these changes might affect mitochondrial behavior. Wide-field microscopy using MitoTracker CMXROS revealed significantly more elongated mitochondria in U2932-R compared to U2932-WT cells, indicative of increased mitochondrial fusion (Fig. 5A–C). To assess mitophagy, we used mt-Keima, a pH-sensitive fluorescent protein that shifts signal upon lysosomal acidification, marking mitolysosome formation [41]. Brief AICA treatment induced mitolysosome formation in mt-Keima-expressing U2932-WT cells but not in U2932-R cells (Fig. 5D, E), indicating impaired mitophagy in resistant cells. Flow cytometric analysis further confirmed that U2932-R cells maintain significantly higher mitochondrial mass than U2932-WT cells (Fig. 5F), a phenotype also suggested in gene-targeted RI-1 variants (FOXK1, KEAP1, PTEN; Supplemental Fig. 2B). Notably, mitochondrial mass increases over time in U2932-WT cells that survive a single CDC treatment (Supplemental Fig. 4A), suggesting this remodelling occurs early during resistance acquisition. Together, these findings establish a link between reduced actin dynamics, decreased fission and mitophagy, increased fusion, and CDC resistance. To functionally validate this, we pharmacologically inhibited mitophagy in RI-1 cells using a combination of the fusion promoter M1 and the fission inhibitor Mdivi-1. This intervention significantly reduced CDC sensitivity without affecting target expression (Supplemental Fig. 5), supporting a functional link between mitophagy and CDC. Interestingly, while U2932-WT cells showed increased mitochondrial mass following CDC, their mitochondrial activity—measured as membrane potential relative to mass—initially increased but subsequently declined over time and was significantly lower in U2932-R cells (Supplemental Fig. 4B). In parallel, CENCAT analysis demonstrated that U2932-WT cells displayed marked dependence on OXPHOS upon CDC exposure, whereas U2932-R showed no significant metabolic shift relative to U2932-WT (Fig. 5G). These findings support a model in which acute OXPHOS activation is part of the early cellular response to CDC, potentially facilitating mitochondrial fusion. In contrast, chronic CDC resistance maintained elevated mitochondrial fusion and mass alongside reduced mitophagy, despite the absence of sustained OXPHOS activity. This suggests a decoupling of mitochondrial structure from energy production, with fusion and mass accumulation serving protective, rather than bioenergetic, functions in CDC resistance.

Fig. 5: CDC-resistant U2932-R cells exhibit reduced CDC-induced mitophagy and increased baseline mitochondrial elongation compared to U2932-WT cells.
figure 5

A, B Baseline mitochondrial morphology in U2932-WT (top) and U2932-R (bottom) cells stained with CMX-ROS MitoTracker and visualized by wide-field microscopy: A overview images and B enlarged views. C Quantification of mitochondrial morphology in U2932-WT and U2932-R cells, categorized manually into fragmented, mixed, or tubular. Data represent two independent experiments, with the average from two different slides per experiment (30 cells per slide). Statistical analysis was performed using a two-way ANOVA. D, E Percentage of cells with mitolysosomes upon 5, 15, and 45 min of exposure to DuoHexaBody-CD37 (1 and 10 µg/mL) in CDC-assays in mt-Keima-expressing (D) U2932-WT and (E) U2932-R. Dot-plots below each graph illustrate the gating strategy based on the ratio of lysosomal to cytosolic mitochondrial fluorescence (n = 3). F Fold change in mitochondrial mass (MitoTracker Green signal) in live U2932-WT and U2932-R cells upon 5 and 15 min CDC-assays with DuoHexaBody-CD37 (1 µg/mL), normalized to no antibody control in U2932-WT (n ≥ 3). Statistical analysis was performed using a one-way ANOVA with post-hoc Tukey’s multiple comparison test. G CENCAT analysis of metabolic dependencies in U2932 cells. (Left) U2932-WT cells treated with control antibody or DuoHexaBody-CD37 (10 µg/mL) in the CDC assay (n = 3). (Right) Baseline comparison of U2932-WT and U2932-R cells (n = 4). Separate paired t-tests were performed for each parameter. N indicates independent experiments, each performed in replicate or as triplicates; the average value per experiment was used for analysis. Data are shown as mean ± SD.

Full size image

Antibody-induced CDC in DLBCL patient samples is associated with altered cytoskeleton-dynamics

To investigate and confirm the occurrence of intracellular mechanisms of CDC resistance in tumor samples from DLBCL patients (n = 20, see Table 1), we evaluated ex vivo CDC activity mediated by DuoHexaBody-CD37. Consistent with our previous findings [11], we observed a heterogeneous CDC response (Fig. 6A), with 6/20 samples exhibiting low-CDC sensitivity (< 50% lysis). This low sensitivity could not be explained by reduced target antigen (CD37) or elevated levels of CRPs, CD46, CD55, and CD59. In fact, CRP expression levels tended to be lower in the low-responding samples, indicating CRP-independent, intrinsic resistance mechanisms (Fig. 6B). Notably, among the 16 high-CDC-sensitive B-cell samples, only one sample was from a relapse patient, while the others were derived from newly diagnosed (ND) patients. In contrast, the 6 low-CDC-sensitive samples comprised an equal distribution of relapse (n = 3) and ND (n = 3) cases, indicating a trend toward lower CDC sensitivity in relapse samples. However, the limited sample size precludes strong conclusions. To further explore the molecular basis of CDC resistance, we performed bulk RNA sequencing on malignant cells isolated from 4 high-CDC-sensitive and 3 low-CDC-sensitive B-cell samples, achieving >95% tumor cell purity.

Table 1 Clinical characteristics of DLBCL patients included in ex vivo CDC assays.
Full size table
Fig. 6: Cytoskeleton- and specifically actin-binding genes are positively associated with ex vivo CDC response to DuoHexaBody-CD37 and with survival in R-CHOP–treated DLBCL patients.
figure 6

A Dose-response curves of DuoHexaBody-CD37 in CDC assays on B-cells isolated from lymph node suspension cells of DLBCL patients (n = 20). B Maximal CDC response in 16 DLBCL samples stratified into high or low-CDC response, based on correlation between CD37 expression and maximal CDC (%), as well as the expression of total CRP (red color intensity). Asterisks indicate high (n = 4) and low (n = 3) CDC responders selected for transcriptomic analysis. C Principal component analysis (PCA) of RNA-seq data from isolated B-cells of high versus low-CDC responders. D Volcano plot of DEGs (n = 429) between high and low-CDC-sensitive patient samples. E Heat-map showing the top 50 DEGs ranked by log2 fold change between high and low responders. F Top 10 enriched GO terms for biological process (left) and molecular function (right) based on DEGs (log2 fold change < −1 or > 1; n = 411), sorted by false discovery rate (FDR). G Volcano plot showing DEGs (n = 2600) between surviving and deceased DLBCL patients treated with R-CHOP and H top 10 enriched GO terms for biological process (top) and molecular function (bottom) of genes more highly expressed in R-CHOP-treated surviving DLBCL patients (n = 1047), sorted by FDR.

Full size image

Principal component analysis revealed distinct genetic profiles between the two groups, with 429 DEGs identified (Fig. 6C, D and Supplemental Table 5). Similar to the RNA-seq results of the U2932 cell line models, the top 50 DEGs based on fold-change were several genes associated with cytoskeletal function, including the actin-binding gene CNN3 and two genes more related to microtubule and mitochondrial dynamics, SYBU and SPART (Fig. 6E). GO enrichment analysis of DEGs revealed enrichment of cytoskeletal-binding and -motor proteins, in addition to cell cycle encoding genes (Fig. 6F). Hence, these data, collected from primary DLBCL patient samples, extends our findings in cell line models and attribute a role for cytoskeletal dynamics in sensitivity to CDC.

The role of actin cytoskeleton in rituximab-induced CDC

R-CHOP, the standard therapy for DLBCL, includes the CDC-inducing antibody rituximab. To explore the clinical relevance of cytoskeletal dynamics as a response parameter to CDC-inducing therapies, we analyzed a gene expression profile dataset from 233 newly diagnosed DLBCL cases curated by Lenz et al. [27], and further annotated by Chiche et al. [28]. Differential expression analysis of patient survival following R-CHOP therapy identified 2600 DEGs (p < 0.05), adjusting for age, sex, and years post-diagnosis (Fig. 6G and Supplemental Table 6). We found that genes encoding CD20 (p < 0.01) and complement protein C3 (p < 0.01) were linked to improved survival, while complement-regulatory proteins CD55 and CD59 (both p < 0.05) were associated with poor outcomes (Supplemental Table 6), suggesting that indeed rituximab-mediated CDC contributes to antitumor activity. GO-term analysis revealed that genes upregulated in patients with cancer-related mortality were enriched in oxidoreductase pathways involved in ROS generation (Supplemental Fig. 6), while genes upregulated in surviving patients were enriched for actin-binding and ion-binding genes (Fig. 6H). These findings confirm a positive association between actin-binding genes, antibody-induced cytotoxicity, and improved patient survival.

Discussion

CDC is a key effector mechanism of therapeutic antibodies in cancer treatment. While resistance to CDC is typically attributed to membrane-associated regulatory factors, intracellular resistance mechanisms have been implicated. Understanding these mechanisms could greatly improve the effectiveness of current and future antibody therapies.

Using a CRISPR-Cas9 screen in RI-1 cells, we identified mitochondrial damage and ROS production as key determinants of CDC sensitivity. Gene-targeted variants of KEAP1, FOXK1, and PTEN exhibited reduced ROS levels and decreased sensitivity to CDC. Similarly, the CDC-resistant U2932-R cell line, despite forming MACs like U2932-WT, mitigated ROS and overcame mitochondrial damage. These findings demonstrate that resistance to CDC is linked ROS neutralization and mitochondrial health.

Transcriptomic and proteomic analyses revealed reduced expression of actin and actin-binding genes in U2932-R cells compared to the sensitive U2932-WT cells, indicating a shift towards a less dynamic cytoskeleton. While actin is known to influence CDC in epithelial cells via plasma membrane modulation [42], our findings highlight a distinct mechanism in lymphoma, wherein cytoskeletal remodelling impacts mitochondrial dynamics. Specifically, AICA treatment reduced actin levels in mitochondrial fractions, and MYO6—an actin-binding gene essential for mitophagy [36]—was among the most downregulated in U2932-R cells. In resistant cells, reduced actin dynamics correlated with impaired mitophagy, increased mitochondrial fusion, and elevated mitochondrial mass—features associated with ROS mitigation [43]. These mitochondrial changes in U2932-R cells were uncoupled from energy production, suggesting a protective rather than bioenergetic function. Functional inhibition of mitophagy further reduced CDC sensitivity, establishing a causal link between mitochondrial quality control and CDC response. Although mitophagy generally supports mitochondrial health, impaired mitophagy upon excessive damage may promote cell death [44,45,46]. Moreover, many cancer cells display mitophagy-inhibitory mechanisms, indicating that suppressed mitophagy may provide a survival advantage [47,48,49]. Thus, reduced mitophagy, paired with enhanced fusion, may offer a survival advantage under complement attack. Collectively, our data identify cytoskeletal remodelling—specifically reduced actin dynamics—as a key driver of CDC resistance through mitochondrial remodelling, allowing reduced mitophagy paired with enhanced fusion to counter AICA-induced oxidative stress.

While our findings emphasize the role of actin in mitochondrial rearrangements in CDC resistance, we acknowledge the broader and more complex functions of actin cytoskeleton. Beyond its involvement in mitophagy, actin is known to regulate mitochondrial oxidative stress through KEAP1 [50] and modulate ion channels [51, 52]. Exploring the interplay between ion channel regulation, mitochondrial rearrangements, and cytoskeletal dynamics may provide further insights into the mechanisms of CDC resistance.

To assess the clinical relevance, we analyzed CDC responses in DLBCL patient samples. Our results showed that low-CDC sensitivity in these samples could not be attributed to CD37 expression or CRP levels, indicating the involvement of intrinsic mechanisms. Transcriptomic analysis of these patient samples revealed reduced expression of cytoskeleton-associated genes, aligning with our in vitro findings and suggesting that cytoskeletal dynamics play a critical role in CDC resistance. These results represent the first demonstration of CDC resistance linked to intrinsic genetic programs beyond CRPs. Moreover, actin-binding gene expression in rituximab-treated (R-CHOP) patients correlated with improved survival. Supporting our findings, a previous CRISPR screen by Thomsen et al. identified actin-associated genes DYNLL1 and KPTN among the top hits for rituximab-induced CDC, underscoring the clinical significance of actin dynamics. Notably, these data are presented in their publication’s supplementary material (Appendix S1, Fig. 2D; Mol Oncol, 2020) [53]. Although pharmacologically targeting actin in vivo poses specificity challenges, actin-related gene expression profiles may serve as predictive biomarkers for CDC responsiveness. In parallel, our CRISPR screen identified epigenetic regulators potentially involved in mitochondrial and cytoskeletal control, offering new therapeutic angles.

Future studies should investigate whether these findings extend to other tumor types and ROS-inducing therapies. We hypothesize that the impact of actin dynamics may vary based on the kinetics and intensity of ROS generation, particularly differentiating between therapies that induce a gradual ROS response and those that trigger a rapid ROS burst, as observed during CDC. Furthermore, we observed that (1) reduced CDC in FOXK1-, KEAP1-, and PTEN-targeted variants occurs only at lower antibody doses, and (2) U2932-R cells required repeated treatments to maintain resistance, which suggest that a strategy favoring high antibody dosing with drug holidays may be preferable to repeated lower dosing.

In summary, we identify mitochondrial damage and ROS regulation as key intracellular factors in CDC resistance. We demonstrate that decreased actin dynamics contribute to resistance by promoting mitochondrial fusion and impeding mitophagy. Our findings suggest that cytoskeletal dynamics, or using actin-related gene expression profiles as biomarkers, could enhance the efficacy of CDC-based therapies. Future studies could focus on regulators of these pathways as potential therapeutic targets to overcome CDC resistance in clinical settings.