DNAM-1 mediates NK-cell activation and host-pathogen interaction via direct binding to fungal cell wall proteases

Abstract

Natural killer (NK) cells play a key role in combating invasive mycoses, with surface proteins crucial for immune cell-target interactions. Here, we investigate the role of DNAX accessory molecule 1 (DNAM-1; CD226) in the recognition of Aspergillus fumigatus and Candida albicans hyphae. Super-resolution microscopy reveals a uniform distribution of DNAM-1 on naïve NK cells that is maintained upon contact with fungal hyphae. However, soluble DNAM-1 binds to the fungal cell wall and blocking DNAM-1 reduces antifungal activity in colony-forming assays. Furthermore, in silico domain–domain interaction analysis identifies several fungal proteins as putative DNAM-1 targets, including OpsB in A. fumigatus and Sap10 in C. albicans. High-affinity binding between Sap10 and DNAM-1 is predicted computationally and confirmed experimentally by co-immunoprecipitation and fluorescence correlation spectroscopy. Microscopic analysis further demonstrates that Sap10 binds to primary NK cells but not to DNAM-1-deficient NK cells. Sap10 binding induced NK-cell activation, as evidenced by increased CD69 expression and elevated perforin and CCL3 secretion. These findings identify Sap10 as a fungal ligand of DNAM-1 and reveal a mechanism by which NK cells recognize fungal pathogens.

DNAM-1, an NK-cell receptor, binds fungal hyphae and interacts with Candida albicans Sap10. This interaction enhances NK-cell activation suggesting a role for DNAM-1 in immune activation during fungal infection.

Introduction

Invasive pulmonary aspergillosis (IPA) and candidiasis are connected to high mortality in immunocompromised patients. Natural killer (NK) cells play a crucial role in the immune system’s fight against these diseases^1,2. Patients with delayed NK-cell reconstitution or reduced NK-cell counts after allogeneic stem cell transplantation (alloSCT) experience a poor prognosis for remission from IPA³. When activated, NK cells release cytotoxic granules containing granzyme and perforin, damaging Aspergillus fumigatus, the major causative agent of IPA⁴. Upon direct contact with A. fumigatus germ tubes, NK cells also release IFN-γ that is essential to stimulate additional immune cells, e.g., alveolar macrophages⁵, but also interferes with fungal metabolic activity and growth⁶.

NK cells detect fungal invaders via surface-associated proteins that mediate appropriate cell biological adaptation. The immunoglobulin-like lectin receptor NKp30 recognizes β-1,3-glucan at the surface of Candida albicans and Cryptococcus neoformans and induces granule repolarization and fungal killing⁷. NKp46 binds the fungal adhesins Epa1, Epa6, and Epa7 to mediate the fungal killing of Candida glabrata⁸. NKG2D is a lectin-like protein that functions as fungal pattern recognition receptor (PRR) for various fungal species and provokes immune cell activation and elimination of fungi⁹. Neural cell adhesion molecule 1 (NCAM-1, CD56) acts as PRR of human NK cells that detects galactosaminogalactan (GAG) in A. fumigatus¹⁰. NCAM-1 relocalizes from the NK cell surface towards the immunological synapse (IS) upon contact with the fungal hypha¹¹. Blocking of NCAM-1 is associated with reduced fungal-mediated NK-cell activation and inhibited cytokine and chemokine secretion (MIP-1α, MIP-1β, RANTES)¹¹.

The membrane associated DNAX accessory molecule 1 (DNAM-1), also known as CD226, T lineage-specific activation antigen 1 (TLisA1) or platelet and T-cell antigen 1 (PTA-1), is a 65 kDa type I transmembrane glycoprotein and member of the immunoglobulin superfamily. DNAM-1 plays an important role in NK-cell regulation in mice and human^12–14 and is considered as one of the markers of mature NK cells¹⁵. Malfunction of DNAM-1 is an assignable cause of different diseases ranging from autoimmune diseases to cancer and viral infections. DNAM-1 is involved in the recognition of several types of human tumors by NK cells, including myeloma, melanoma, and ovarian carcinoma. Along with NK cells, DNAM-1 is also found in T cells, NK-T cells, B cells, monocytes/macrophages, dendritic cells (DC), megakaryocyte/platelet lineage, and hematopoietic precursor cells¹⁴.

DNAM-1 consists of three domains. The 236 amino acid (aa) extracellular domain (ECD) includes 2 immunoglobulin V-like domains, and 8 N-linked glycoprotein sites and provides the structural basis for recognition of ligands and adhesion. The transmembrane domain ensures plasma membrane localization and a cytosolic domain allows for intracellular signaling via four tyrosine and one serine residues¹⁴. Engagement of DNAM-1 initiates a phosphorylation cascade that ultimately results in the degradation of the Forkhead Box-O-1 (FOXO1) transcription factor, a negative regulator of NK-cell activation¹⁶. Thus, DNAM-1 is integral in facilitating NK cell activation.

In humans and mice two proteins act as ligands of DNAM-1, first the poliovirus receptor (PVR, CD155), also known as Nectin-like molecule-5 (Necl-5), and second, the herpesvirus entry mediator B (HVEB)¹⁷. Both molecules are widely expressed in many tumor cells. Consequently, cytotoxic T-lymphocyte attacks can occur via DNAM-1-specific binding¹⁴. Besides that, its function is also interconnected with signaling pathways of other receptors recognizing the same ligands, namely T cell activation, increased late expression (TACTILE, CD96), T cell immunoreceptor with Ig and ITIM domains (TIGIT), and cytotoxic and regulatory T cell molecule (CRTAM)¹⁸. TACTILE and DNAM-1 molecules modulate the secretion of IFN-γ in NK cells¹⁸ and the synergistic effect of DNAM-1, TACTILE, TIGIT, and CRTAM are reported to balance the activation of NK cells in vivo^13,16.

Considering the state of the art, there is currently no report investigating the role of NK-cell DNAM-1 in the context of fungal infection. Given the importance of this receptor in the context of cancer, it is likely that DNAM-1 is also involved in the regulation of NK cell’s attack against fungal hyphae. Here, we investigate the interaction of DNAM-1 with the clinically relevant fungi A. fumigatus and C. albicans. We show that DNAM-1 directly interacts with fungal hyphae. Our bioinformatic forecast suggests fungal cell-wall-located proteases as potential interaction partners for DNAM-1. While comparing human primary natural killer (hpNK) cells with a CRISPR-Cas mediated DNAM-1-KO mutant, we prove that Sap10, a cell surface-associated secreted aspartic protease of C. albicans^19,20, interacts with DNAM-1 and influences NK-cell activity via DNAM-1.

Results

DNAM-1 is uniformly distributed on the surface of hpNK cells and binds to the fungus

DNAM-1 (CD226) is an important regulator of NK-cell activity^21–23 and is expressed in the vast majority of NK cells²⁴. Though NK cells are crucial in the fight against invading fungi in neutropenic patients and mice^1,3, the role of DNAM-1 has not yet been analyzed in the context of NK cells interacting with fungal hyphae.

We used super-resolution fluorescence microscopy to visualize the distribution of DNAM-1 on the surface of NK cells. By employing an anti-CD226 antibody labeled with Alexa647, we detected DNAM-1 on isolated human primary NK cells using either confocal laser scanning microscopy equipped with an Airyscan detector (Airyscan imaging) or direct stochastic optical reconstruction microscopy (dSTORM) (Fig. 1a, b). In addition, we visualized the distribution of NCAM-1 (CD56) with an anti-CD56 antibody using dSTORM for comparative analysis.

Fig. 1. Super-resolution microscopy analysis of DNAM-1 and NCAM-1 expression on human primary NK cells.

Visualization of the surface receptors DNAM-1 and NCAM-1 on naïve NK cells in the absence of fungi (a, b). DNAM-1 and NCAM-1 are evenly distributed on the surface of hpNK cells. hpNK ΔDNAM-1 cells exhibit similar distribution of NCAM-1 while DNAM-1 is absent. Bars represent 2 µm (a) or 1 µm (b). Images were acquired using the Airyscan SR (a) or dSTORM (b). Data representative of three independent experiments.

Overall, DNAM-1 and NCAM-1 exhibit similar distribution on the surface of naïve NK cells. Both proteins are associated with the cell surface and equally distributed without polarity alongside the spherical-shaped NK cell. Additionally, using the fine resolution of dSTORM, we did not detect any significant sub-organization; both proteins were distributed uniformly on the cell surface.

With CRISPR-Cas technology, we eliminated DNAM-1 in human primary NK cells (hpNK ΔDNAM-1) and applied the same protocol as before. As anticipated, the DNAM-1 signal was absent, while NCAM-1 displayed a distribution comparable as observed in control hpNK cells showing successful deletion of DNAM-1 expression (Fig. 1b).

Next, we challenged NK cells with A. fumigatus hyphae for 6 h to investigate how the interaction with the fungus affects the receptor distribution. Recent studies have shown that the NK-cell pattern recognition receptor NCAM-1 alters its localization from an even distribution across the cell surface to an accumulation at the immunological synapse in a time- and actin-dependent manner^11,25. In accordance, we noted a strong accumulation of NCAM-1 at the contact site between the fungus and NK cell (Fig. 2a). Similarly, NCAM-1 accumulated in the immunological synapse in hpNK ΔDNAM-1 cells, suggesting that the absence of DNAM-1 does not hinder synapse formation.

Fig. 2. Super-resolution microscopy analysis of DNAM-1 and NCAM-1 expression on human primary NK cells interacting with A. fumigatus hyphae.

NK cells interacting with fungal hyphae (a, b). NK cells were cocultured with A. fumigatus hyphae for 6 h and stained for DNAM-1 and NCAM-1 as indicated. c Comparison of hpNK cells and hpNK ΔDNAM-1 cells. The immunological synapse is also formed in the absence of DNAM-1. d Distribution of DNAM-1 in comparison to NCAM-1. Note that NCAM-1 accumulates in the immunological synapse while the distribution of DNAM-1 is unaffected. Bars represent 2 µm (a) or 1 µm (b). Images were acquired using the Airyscan SR (a) or dSTORM (b). Data are representative of three independent experiments.

In contrast, after 6 h of co-incubation with A. fumigatus hyphae, DNAM-1 either does not or does only marginally relocalize to the IS. Instead, DNAM-1 remains predominantly evenly distributed across the entire surface of the hpNK cells (Fig. 2b). This observation is also supported by our flow cytometry measurements (Supplementary Fig. 1). A strong reduction in mean fluorescence intensity (MFI) is an indirect measure of IS formation as assessed by flow cytometry^11,25. However, in comparison to the NCAM-1 fluorescence signal following stimulation with A. fumigatus, the MFI of DNAM-1 only decreased marginally upon contact with A. fumigatus hyphae, even after 12 h of co-incubation. This indicates a lack of targeted relocalization of DNAM-1 towards the IS. In line with these experiments, Airyscan imaging of NK cells facing C. albicans did not yield any accumulation of DNAM-1 at the IS (Supplementary Fig. 2).

To investigate a potential direct interaction of DNAM-1 with fungal hyphae, we treated A. fumigatus and C. albicans strains with soluble human DNAM-1 protein or BSA (control) and detected the bound protein using an Alexa647-labeled anti-DNAM-1 antibody (Fig. 3). Our results confirmed direct binding of DNAM-1, but not BSA, to the cell wall of A. fumigatus ATCC 46645 and C. albicans SC5314 (Fig. 3a,b). The staining intensity varied among samples and was locally enriched along the hyphal wall, suggesting distinct regions that may contain potential fungal binding partners for DNAM-1. We further analyzed the binding of DNAM-1 protein to fungal hyphae by flow cytometry (Fig. 3c,d). A significant increase in the number of fluorescent hyphae was observed for both fungal species at a concentration of 2 µg/ml.

Fig. 3. Binding behaviour of DNAM-1 to fungal hyphae.

A. fumigatus (a, c) and C. albicans (b, d) hyphae germinated 16 or 3 h, respectively, and were incubated with soluble DNAM-1 or Bovine-serum-albumin (BSA) protein or they were left untreated (w/o), as indicated. The samples were then stained with Alexa647-tagged anti-CD226 antibody (magenta) and analyzed with Airyscan SR (a,b). Simultaneously, the autofluorescence was assessed in the GFP-channel (cyan). Note that DNAM-1 directly binds to both fungi. Bars represent 5 µm. Flow cytometric quantification (c, d) of the binding of soluble DNAM-1 (2 µg/ml) or BSA (4 µg/ml) to fungal hyphae measured as fluorescence positivity percentage. Data representative of five independent experiments. Error bars represent standard deviation. Statistical significance was determined using one-way ANOVA repeated measures.

To understand if the interaction of DNAM-1 with fungal components has immunobiological consequences, we assessed damage of C. albicans hyphae after exposure to NK cells in the presence or absence of a DNAM-1 D1 domain-blocking antibody (Fig. 4a)²⁶. Notably, the presence of the blocking antibody significantly reduced the percentage of damaged hyphae, suggesting that DNAM-1 contributes to NK-cell cytotoxicity in response to fungal hyphae.

Fig. 4. NK-cell-mediated hyphal damage—influence of DNAM-1 activity on fungal metabolism and viability.

a Hyphal metabolic activity assay was performed by co-culturing NK cells from five different donors (with or without blocking antibody (bab)) and C. albicans germ tubes (germinated for 3 h) at an effector-to-target ratio of 20:1. Hyphal damage was measured by XTT assay after 6 h and expressed as normalized to control (C. albicans only). Controls included C. albicans hyphae alone (negative control) and hyphae treated with hydrogen peroxide (positive control). b Viability (CFU) assay determining the DNAM-1-mediated NK cell antifungal activity. NK cells (n = 5) were co-incubated with or without bab and challenged with C. albicans germ tubes for 4 h (20:1 E:T value). Viability was assessed by normalizing control condition (C. albicans alone). Data are shown as box plots; black dots indicate the mean. Statistical significance was determined using a Student’s t test with paired samples.

To determine the role of co-stimulatory molecules DNAM-1, CD96 and TIGIT in antifungal activity, a CFU assay (viability test) was carried out. Co-incubation of C. albicans germ tubes with NK cells reduced fungal viability approximately by half. When NK cells were treated with blocking antibodies against DNAM-1 (Fig. 4b), CD96, or TIGIT (Supplementary Fig. 3) it had effects on the viability of C. albicans.

DNAM-1 blockade (Fig. 4b) significantly and partially reversed the reduction in viability. In contrast, TIGIT blockade and the corresponding isotype controls did not consistently affect viability (Supplementary Fig. 3). Notably, CD96 blockade also significantly increased viability, yielding an effect comparable to that observed upon DNAM-1 blockade.

Computational analysis suggests Sap10 in C. albicans/OpsB in A. fumigatus as potential interaction partners of CD226

After proving that DNAM-1 interacts with the fungal surface, we aimed at identifying possible fungal targets for DNAM-1. Therefore, we searched for potential interaction partners of DNAM-1 in the surface proteome of C. albicans and A. fumigatus. As there is no previous finding of a given interaction between orthologs in other host and pathogens, there is no result in the interlog protein network for yet unknown pathogen-host interactions²⁷. Therefore, in this study, we used a domain-pairs-based approach, where entries in the domain-domain-interaction network are compared with the proteome present in the pathogen and host, as shown in Fig. 5a. The 3did database was queried using the V-set domain of DNAM-1 to identify potential domain interactors from the fungal pathogens²⁸. A high number of fungal proteins potentially interacting with the V-set domain were obtained. These candidates were refined based on their localization (accessibility of fungal proteins to DNAM-1) and their potential to undergo pathogenicity-related processes.

Fig. 5. Domain–domain interaction analysis reveals potential interaction partners for DNAM-1 on fungal hyphae.

Workflow of domain-domain-interaction analysis (a). Protein motifs known to interact with V-set domain of DNAM-1 were searched in pathogen domain libraries and functionally filtered yielding a refined protein-protein interaction network for C. albicans (b) and A. fumigatus (c). Different colors of resulting protein IDs represent different protein domains. Details can be obtained from Supplementary data 1 and 2.

Indeed, our computational analysis of domain-domain interaction of various fungal proteins with the extracellular V-domain of DNAM-1 revealed various proteins from both C. albicans (35 proteins) and A. fumigatus (25 proteins) as depicted in Fig. 5b, c, respectively. An overview of all proteins representing potential interaction partners of DNAM-1 is provided in Supplementary data 1 (A. fumigatus) and Supplementary data 2 (C. albicans).

Among the proteins identified as potential target proteins of DNAM-1, several proteins were found to share the same domain. Some domains were present in both A. fumigatus and C. albicans, while others were exclusively observed in one of both species. Proteins containing the Asp-domain were frequent and shared by both fungi (A. fumigatus: 5, C. albicans: 11). Besides that, also proteins with alpha-amylase domain (A. fumigatus: 3, C. albicans: 1) or metallophosphatase domain (A. fumigatus: 2, C. albicans: 1) were present in both fungi. Among the proteins only present in A. fumigatus we found COesterase domain (5), peptidase domain (3), subtilase-family domain (4), and a group of different proteins that were only found once each (4 proteins). In C. albicans, suggested interacting proteins contained a Ras-domain (9), sugar-transporter domain (Sugar_tr; 3), ABC-transporter domain (2), thioredoxin (2), mitochondrial carrier protein domain (Mito_carr; 2), and others only found once each (4).

The C. albicans proteins were scored according to their localization and known involvement in pathogenicity and virulence. 5 proteins from C. albicans sharing this domain reached the highest score (1.0), namely Sap1, Sap2, Sap9, Sap10 and Pra1, all exhibiting the Asp-domain. In contrast, Mal2 exhibiting the alpha-amylase domain only reached a score of 0.14. Due to the accessibility to purified protein and to previous experimental data, we focused on the glycosylphosphatidylinositol (GPI)-anchored protease Sap10^19,20.

Sap10 (AAF66711), also known as candidapepsin-10, a 60 kDa protein consists of 453 aa and is highly N-glycosylated¹⁹ (Fig. 6a). Deletion of Sap10 leads to attenuated damage of epithelial cells by C. albicans suggesting a role of Sap10 in host-pathogen interactions¹⁹. The computational analysis also yielded a high likelihood for a homologous protein of Sap10 in A. fumigatus, aspartic type endopeptidase OpsB (KEY82700). This protease consists of 485 aa, shares >30% sequence identity with Sap10 and exhibits 8 potential N-glycosylation sites. Though this protein has not yet been analyzed in detail in A. fumigatus, phenotypic investigations of the homologous protein in Aspergillus oryzae suggest contribution of OpsB in processing cell wall synthesis-related proteins²⁹.

Fig. 6. Analysis of the interaction between soluble Sap10 and DNAM-1 protein.

a Protein model for the interaction between Sap10 and the V-domain of DNAM-1 and protein docking calculated with Zdock Server. b Co-immunoprecipitation assay (Co-IP) assay analyzed by flow cytometry. Anti-DNAM-1 antibodies were immobilized on beads and incubated with soluble Sap10 (ATTO643-labeled) in the presence or absence of DNAM-1 protein as indicated. The percentage of Sap10-positive events was measured by flow cytometry. Data are representative of five independent experiments. Error bars represent standard deviation. Significance tests were performed using a standard student t test. Fluorescence correlation spectroscopy (FCS) analysis. c Normalized correlation function G(τ) of the fluorescence fluctuations of N molecules freely diffusing through the excitation/detection volume for the mixed protein and protein reference samples as indicated. d Saturation curve showing the characteristic diffusion time τ_d of Sap10 for different ratios of DNAM-1 and Sap10 (ATTO643-labeled) as indicated. A Hill function was used to obtain the ratio at which half of the Sap10 proteins were bound to DNAM-1. Data representative of three independent experiments. e Microscopical analysis. Airyscan images of hpNK and hpNK ΔDNAM-1 cells as indicated. NK cells were incubated with fluorescently labeled SAP10 (5 µg/ml; Alexa594), fixed and stained with anti-DNAM-1 antibody (Alexa647). Bars represent 2 µm. Data are representative of three independent experiments.

Using AlphaFold 3 prediction³⁰ for the interaction of human DNAM-1 and C. albicans, Sap10 revealed a highly interconnecting protein assembly, suggesting a high affinity binding of Sap10 to DNAM-1. To analyze this interaction in more detail, we used Zdock server for protein-protein docking³¹ and analyzed the interaction between DNAM-1 (crystallographic structure PDB 6ISB) and truncated Sap10 (AlphaFold monomer v2.0 prediction³²; truncation visualized by scissors in Supplementary Fig. 4). Zdock analysis revealed strong interaction between the V-like domain and Sap10, as indicated by the high docking score of 2667.3. The intense interaction between the two proteins is also observable from the interconnected protein structural models (Fig. 6a).

We further analyzed the interaction of the interface residues and the hydrogen bonds by means of DBePISA³³. In the binding complex, 52 residues from Sap10 and 47 residues from DNAM-1 participate as active interface residues in the complex formation. From Sap10 the amino acids Thr71, Ser73, Tyr74, Asp75, Trp77, Ser134, Ile135, Lys163, Gly180, Ser240, Gly344, Gly346, and Gly347 proposed to play a major role in the stabilization of the complex structure due to their central location within the dimeric interface. Only the domain D1 of DNAM-1 is strongly involved in the interaction with the Glu18, Tyr21, Pro22, Ser23, Gly25, Ile26, Phe45, Pro47, Gly50, Leu64, Ser66, Thr67, Met68, Phe173, Pro174, and Thr195, being the presumptive candidates for interaction. Seven hydrogen bonds were noted between Sap10 and DNAM-1 (Supplementary Fig. 5). PDBePISA calculated the percentage of area accessible to solvent to only 9.1% for Sap10 and 13% for DNAM-1. Thus, the high number of amino acids not accessible to the solvent in the interface of Sap10 (1349 Ų) and DNAM-1(1574 Ų) significantly contributes to this complex formation.

We also investigated the influence of glycosylation of Sap10 on the stability of the complex. The Sap10 protein was first scanned through ScanProsite³⁴, to identify the peptidase domain, that ranges from aa52 to aa372. Next, we aimed at identifying the potential N-glycosylation sites in the protein using NetNglyc³⁵ and GlycoEP³⁶ servers. Generally, glycan chains are added at the Asn residues at Asn-X-Ser/Thr motifs and indeed such common N-glycosylation sites are present in positions 128, 208, 211, and 398. Subsequently, we used CHARMM-GUI³⁷ to model Sap10 with glycan chain and the glycosylated protein was subjected to docking with DNAM-1 protein in a further in silico study. We calculated the docking scores and binding energies for the interaction of DNAM-1 with Sap10 with and without glycosylation. The Docking score increased from 1086.52 (non-glycosylated Sap10) to 2667.3 (glycosylated Sap10), signifying that glycosylation enhances the stability of the Sap10-DNAM-1 complex. The increase in stability is also reflected by a reduction of binding energy from −11.55 kcal/mol or −24.76 kcal/mol for unglycosylated or glycosylated Sap10, respectively.

DNAM-1, TIGIT and TACTILE share the same ligand, CD155, in their interaction with cancer cells¹⁸. Therefore, we checked the potential interaction of Sap10 with TIGIT and TACTILE. Indeed, both proteins were predicted to interact with Sap10, as indicated by binding energies for non-glycosylated protein of −19.76 kcal/mol (TIGIT) and −13.48 kcal/mol (TACTILE). The calculated differences in binding energy of the non-glycosylated proteins align with the known binding affinities for CD155 (TIGIT > TACTILE > DNAM-1)^38–40. However, weaker binding of the glycosylated Sap10 protein to TIGIT (−13.80 kcal/mol) and TACTILE (−11.10 kcal/mol) was calculated, which may have immunobiological implications.

DNAM-1 interacts with Sap10

Following the domain-domain interaction forecast, we experimentally investigated a potential interaction between DNAM-1 and the fungal protease Sap10, using a truncated excreted version of Sap10 (kindly provided by the Lab of Professor Bernard Hube) heterologously expressed in P. pastoris and purified from the culture medium¹⁹.

First, we used a modified immunoprecipitation approach (Fig. 6b), to analyze the interaction between soluble DNAM-1 and fluorescently labeled Sap10 (Atto643). DNAM-1 antibody was immobilized on Aminolink^TM plus coupling resin agarose beads and then incubated with ATTO643-labeled Sap10 protein in the presence or absence of DNAM-1 protein. After elution and stabilization, the beads were subsequently analyzed by flow cytometry to determine the percentage of Sap10-positive events. In the presence of DNAM-1, the percentage of Sap10-positive events significantly increased to 71.2%, compared to 22.5% in the absence of DNAM-1. This clearly indicates the high affinity of Sap10 for DNAM-1, as efficient binding was observed only in the presence of DNAM-1. Positive events observed in the control group are expected due to the presence of unbound, labeled Sap10 protein, which contributes to background fluorescence. This was confirmed by fluorescence microscopy of the beads.

We then investigated the dose-dependency of DNAM-1 with fluorescently labeled Sap10 (Atto643) by fluorescence correlation spectroscopy (FCS, Fig. 6d). While the dye alone shows a diffusion time of 0.538 ± 0.006 ms, the labeled Sap10 protein exhibits a diffusion time of 1.07 ± 0.09 ms. Adding unlabeled soluble DNAM-1 to the labeled Sap10 protein further increases the diffusion time up to 2.05 ± 0.03 ms under saturating conditions in a ratio dependent manner, yielding 50% saturation (Kd value) at a ratio of 2.0 ± 1.8 DNAM-1 molecules per Sap10 molecule (Fig. 6d). As Sap10 concentration was 1 nM throughout the experiments, this result confers to a concentration of 2 ± 1.8 nM, reflecting a high-affinity binding of Sap10 to DNAM-1. Adding soluble DNAM-1 to the sole Atto643 dye did not result in any change of the diffusion time (Supplementary Fig. 6). Taking all the data into account, this is a clear indication that Sap10 is an interaction partner of DNAM-1, binding with high affinity.

Next, we challenged hpNK cells with purified Sap10 protein labeled with Alexa594 (Fig. 6e) and visualized DNAM-1 with an antiCD226 antibody. The distribution of the Sap10 signal showed a high similarity with the DNAM-1 signal that was ubiquitously distributed on the cell surface. Furthermore, in some cells, we also observed intracellular signals that might be due to the uptake of Sap10 into vesicles. In contrast, the fluorescence signal of Sap10 was completely absent on the cell surface in the hpNK ΔDNAM-1 cells, again indicating that DNAM-1 specifically promotes the binding of Sap10 to the cell surface. However, the fluorescent intracellular vesicles were also observed in ΔDNAM-1 cells supporting the hypothesis of non-targeted uptake of Sap10. Of note, Cos7-cells, that are known to lack DNAM-1⁴¹, did not exhibit any fluorescence on the surface after treatment with fluorescent Sap10 (Supplementary Fig. 7).

Sap10 leads to activation of NK cells in a dose-dependent manner

After proving the specific binding of Sap10 to DNAM-1, we investigated whether Sap10 has the capacity to modulate NK-cell activity. To examine the potential of Sap10 to induce NK-cell response, we stimulated NK cells with purified soluble Sap10 in various concentrations and at different time points. Using CD69, a membrane-bound type II C-lectin receptor and a classical early marker of lymphocyte activation⁴², we observed a significant and dose-dependent upregulation of CD69 surface expression, indicating NK-cell activation (Fig. 7a). As positive control cells, NK cells were treated with Phorbol 12-myristate 13-acetate (PMA) and Ionomycin (IO), which was shown previously to lead to a highly activated state of lymphocytes⁴³. The increase in CD69 MFI for PMA/IO-treated NK cells was more than one magnitude higher compared to untreated control cells (~30 fold). For Sap10 concentrations of 2 µg/ml and higher, CD69 MFI slightly increased in comparison to untreated cells and showed highest MFI at 8 µg/ml. Treatment with the blocking antibody alone resulted in only weak, non-significant changes in CD69 MFI compared to the unstimulated control, indicating that the blocking antibody itself does marginally affect baseline CD69 expression in the absence of Sap10. In addition, NK-cell activation increased over time following Sap10 stimulation (Fig. 7a). Highest activation was observed with 8 µg/ml and 24 h of treatment (increment of CD69 expression by a factor of 2.97 ± 1.87). Of note, in flow cytometry experiments, we observed similar numbers of both DNAM-1 and NCAM-1 positive cells and comparable MFI of DNAM-1 in Sap10-treated cells as in untreated hpNK cells (Supplementary Fig. 8a,b).

Fig. 7. Sap10 treatment leads to increased activation and enhanced relative secretion of perforin and CCL3 by NK cells.

a, b Flow cytometry analysis of hpNK cells from at least five different donors. a Relative mean fluorescence intensity of CD69 for untreated (control) cells and cells treated with different concentrations of Sap10 for 12, 18 or 24 h as indicated. b Relative MFI of CD69 of hpNK cells exposed to 4 µg/ml Sap10 for 24 h and treated with either isotype or anti-DNAM-1 blocking antibody as indicated. c, d Relative secretions of perforin (c) and CCL3 (d) by hpNK cells treated with Sap10 (8–0.5 µg/ml) in the presence of blocking antibody or corresponding isotype control. Measurements were normalized to the data for hpNK cells without Sap10. Both perforin and CCL3 secretion increased in a dose-dependent manner with Sap10 treatment, and this effect was reduced by DNAM-1 blockade. Error bars represent standard deviation. Statistical significances were determined using repeated measures ANOVA (a, c, d) or paired student’s t test (b).

To further confirm that Sap10-induced NK-cell activation was mediated by DNAM-1, we used an anti-CD226 blocking antibody prior to incubation of NK cells with Sap10. We compared Sap10-dependent changes of CD69 MFI of cells treated with the blocking antibody and untreated cells (Fig. 7b). A significant reduction of MFI of the cells treated with the blocking antibody (1.2 ± 0.18) in comparison to cells treated with an isotype antibody (1.4 ± 0.2) was observed, suggesting that Sap10 is recognized by DNAM-1 on the NK-cell surface leading to NK-cell activation. A similar but less pronounced effect was also observed at lower Sap10 concentrations (Supplementary Fig. 8d).

As we aimed in further analyzing the Sap10 dependent activation of NK cells, we assessed the secretion of cytotoxic effector molecule perforin (Fig. 7c, and Supplementary Fig. 9a) and the chemokine CCL3 (Fig. 7d and Supplementary Fig. 9b) using ELISA. Indeed, Sap10 exposure triggered robust secretion of perforin and chemokine CCL3 after 24 h. In the NK-Sap-10 co-culture, the concentration of both perforin and CCL3 increased with increasing concentrations of Sap10. The concentration of perforin increased only slightly but significantly from 2.63 ± 0.35 ng/ml (untreated control cells) to 4.3 ± 0.58 ng/ml after 24 h treatment (Supplementary Fig. 9a). The relative amount of perforin released by Sap10-treated NK cells was significantly reduced when the blocking antibody was used instead of an isotype control, further indicating that NK-cell activation is mediated by Sap10 binding to DNAM-1 (Fig. 7c).

In contrast, CCL3 secretion was significantly and strongly increased after treatment of NK cells with Sap10 (Supplementary Fig. 9b). While the CCL3 concentration was in the range of 0.034 ± 0.02 ng/mL treatment with 0.5 µg/ml Sap10 already yielded a slightly elevated CCL3 concentration (0.11 ± 0.03 ng/mL). Maximal CCL3 release was observed after stimulation with 4 µg/ml Sap10 (0.36 ± 0.05 ng/mL). This is remarkable as CCL3 also functions to attract additional NK cells to the place of infection⁴⁴. As observed with perforin, also the relative amount of CCL3 released by Sap10-treated NK cells was significantly reduced when the blocking antibody was used instead of an isotype control, once more indicating that the activation of NK cells occurs via Sap10 binding to DNAM-1 (Fig. 7d).

Discussion

NK cells play a crucial role in combating fungal infections in neutropenic patients suffering from invasive mycoses^5,6. Therefore, it is critical to understand how the NK cells perceive and orchestrate the signals released by the invading fungus. NK cells possess an arsenal of different receptors capable of recognizing pathogens or cancer cells^13,45. DNAM-1 (CD226) is well known as a key protein regulating NK-cell activity and has been investigated in the interactions of NK cells with viruses and cancer cells^15,24,46,47. However, there are no reports about the role of DNAM-1 in the context of fungi. Here, we investigate the role of DNAM-1 in the interaction of NK cells with two different fungal species A. fumigatus and C. albicans that are of high clinical relevance, increasing incidence, and capacity to evade host immune responses. While our study yielded only limited information about DNAM-1-A. fumigatus interaction, for C. albicans, we show that fungal proteases are targets for the DNAM-1 protein and capable of modulating the NK cell reaction.

Other receptors are known to be involved in the detection of fungi and regulation of immune response of NK cells. NKp30 and NKp46 are NK-cell receptors involved in the detection of fungal β-1,3 glucan and adhesins, respectively^7,8. NKG2D is a lectin-like protein that functions as fungal PRR of various fungal species and provokes immune cell activation and elimination of fungi⁹. Besides that, NKG2D and NCAM-1 (CD56) act as pattern recognition receptors of fungi^9,11. The latter engages with the fungal polysaccharide GAG in the fungal cell wall, thereby activating the NK-cell’s response against A. fumigatus in vitro^10,11,25. NCAM-1 is evenly distributed on the cell surface of naïve NK cells. Upon contact with GAG on A. fumigatus hyphae, it relocalizes in an actin-dependent manner towards the NK cell-hyphal interaction site, resulting in a notable reduction in NCAM-1 fluorescence intensity in flow cytometry measurements^9,11,25.

Having this phenomenon in mind, we investigated whether the presence of A. fumigatus hyphae also induces DNAM-1 relocalization in hpNK cells. While our microscopical analysis revealed homogenous distribution of DNAM-1 along the membrane of hpNK cells (Fig. 1), neither microscopy nor flow cytometry (Fig. 2 and Supplementary Fig. 1) provided evidence of a significant displacement of DNAM-1 receptor proteins toward the fungus. Only, after prolonged incubation periods (up to 24 h), we observed a slight enrichment in the IS in some experiments. Also, NK cells interacting with C. albicans hyphae did not exhibit any relocalization towards the IS (Supplementary Fig. 2). This is remarkable, as DNAM-1 molecules localize in the IS during interactions with cancer cells, specifically K562 cell line¹². DNAM-1 associates with lipid rafts at the immunological synapse, a process that requires serine phosphorylation at residue 329^48,49. The activating synapse triggers DNAM-1-mediated signaling, which promotes the accumulation of LFA-1⁵⁰. This close association between LFA-1 and DNAM-1 initiates a downstream signaling cascade, which in turn leads to NK-cell activation (expression of genes involved in cytotoxic granule release, cytokine secretion, and calcium mobilization)^16,23,50. However, despite the absence of DNAM-1 reorganization, both hpNK cells and hpNK ΔDNAM-1 cells exhibited relocalization of NCAM-1 towards the IS, indicating the normal function of these cells. We may interpret the absence of DNAM-1-relocalization as indication that the binding and activation of DNAM-1 occurs not only at the fungal cell surface in the IS region but also by a soluble fraction of fungal proteins. An alternative explanation could reflect shorter DNAM-1 interaction times and reliance on other adhesion molecules such as LFA-1. Additionally, DNAM-1 may play an upstream role in early recognition and NK-cell priming without remaining enriched at the fully formed IS. Nevertheless, both cell sorting and microscopic analysis of A. fumigatus and C. albicans hyphae revealed the direct binding of DNAM-1 to the fungal hyphae (Fig. 2).

At the same time, we could show that blocking of the active domain of DNAM-1 led to a reduced metabolic activity as analyzed by an XTT-based assay (Fig. 4a). Consistently, blocking DNAM-1 significantly restored fungal viability as demonstrated by CFU assay in comparison to the isotype control, confirming the functional contribution of DNAM-1 in NK-cell-mediated antifungal activity (Fig. 4b).

The bioinformatical domain-domain interaction approach forecasted several fungal surface proteins as potential interaction partners of DNAM-1. Among them, proteins containing the V-set domain showed the highest score considering various parameters that are of importance in pathogenicity. Candida albicans Sap1, Sap2, Sap9, Sap10, and PRA1 reached the maximal score. As Sap10 was previously investigated in detail^19,20 and available as soluble protein, we focused on Sap10 in our experiments (Fig. 5, Supplementary data 1 and 2).

Sap10 is constitutively expressed in hyphae and membrane-bound via a GPI-anchor. When the GPI-anchor is cleaved, Sap10 is transferred to the cell wall and thus accessible in the cell surface of C. albicans^19,20. In our experiments, DNAM-1 was not homogenously distributed on the fungal surface; instead, it accumulated locally without any obvious morphological pattern, suggesting varying accessibility to the fungal binding partner. Though this hints toward clustering, electron microscopy images of the Sap10 fungal ligand heterologously expressed in Pichia pastoris do not reveal a clustered appearance¹⁹. A patch-like distribution was however, reported for C. albicans, where similar patterns with patches of exposed peptides have also been described during nanobody-mediated visualization of extracellular candidalysin on the surface of invading hyphae⁵¹. As Sap10 is involved in biofilm formation^52,53, these accumulations may also represent anchors for biofilm formation. However, cell sorting analysis revealed a clear and significant binding of DNAM-1 towards the surfaces of both fungi.

According to the domain-domain-interaction analysis, Sap10 is not an exclusive fungal interaction partner of DNAM-1. This finding is experimentally supported by the fact that soluble DNAM-1 also bound to C. albicans sap10Δ/Δ strain with similar intensity as the wild type strain (Supplementary Fig. 10). Though Sap10 is associated with the fungal cell wall, it is likely that other fungal secreted proteins of C. albicans like Sap1-9⁵⁴ or from A. fumigatus like pepA and pep2, may also interact with DNAM-1⁵⁵. These proteins share the Asp domain with Sap10 and have been suggested as potential interaction partners based on domain-domain-interaction analysis (Fig. 5a, Supplementary data 1 and 2). Due to the high similarity between Sap10 and OpsB from A. fumigatus, it is likely that OpsB undergoes similar interaction with DNAM-1 as Sap10. However, there is currently very limited data available for OpsB, beyond reports describing it as SAP-like protease expressed in A. fumigatus. This limitation compelled us to focus experimental work on Sap10, which was readily available in a purified form and has been extensively studied in the literature.

RNAseq data show that opsB (AFUA_6G05350) and pepA (AFUA_5G13300) of A. fumigatus are expressed in hyphae in the host environment. mRNA-levels of both proteins were found to be in a similar range for hyphae grown in standard media and blood (GEO/NCBI (accession number GSE70227)⁵⁶, a prerequisite for a possible interaction of these proteins with NK-cell surface receptors. The role of these proteins in the interaction with DNAM-1 will be the task of future investigations requiring purified proteins and/or mutant strains for detailed analyses. Furthermore, there are various other domains expected to bind to the V-set domain of DNAM-1 (Supplementary data 1 and 2). Further detailed investigation is required to assess whether there is direct interaction between these proteins and DNAM-1 and to evaluate their immunological relevance.

We could clearly identify C. albicans Sap10 as an important fungal ligand of DNAM-1 that, upon binding, can provoke a specific immune response in hpNK cells. Binding of Sap10 to the V-set domain of DNAM-1 was calculated by the PDBePISA algorithm³³, suggesting strong binding via hydrogen bonds. Sap10, as well as the A. fumigatus homolog, exhibits 4 N-glycosylation sites and glycosylation of Sap10 was experimentally proven by treatment with N-glycosidase F¹⁹. The glycosylation of Sap10 is predicted to enhance the binding to DNAM-1, a phenomenon that is well-documented in other host–pathogen interactions^57,58. Our in-silico study comparing the binding of Sap10 to DNAM-1 domain D1 with and without glycosylation revealed importance of glycosylation for the stabilization of the complex structure. The docking score increased from 1086.52 for Sap10-DNAM-1 protein complex to 2667.3 for glycosylated Sap10-DNAM-1. In accordance, a recent investigation highlights the importance of glycosylation of the mammalian ligand CD155 in the interaction with DNAM-1⁵⁹.

We experimentally demonstrated a strong interaction between DNAM-1 and Sap10 using immunoprecipitation and FCS (Fig. 6b–d). Following the diffusion of a truncated version of Sap10 labeled with Atto643 in the presence of varying concentrations of soluble human DNAM-1, we observed a significant increase of the perfusion time with increasing DNAM-1 concentrations. Half-maximal binding saturation at a ratio of 2.0 ± 1.8 DNAM-1 molecules per Sap10 molecule suggests high-affinity binding. This was further supported by microscopical analysis, which revealed binding of fluorescent Sap10 to hpNK cells but not to DNAM-1 deleted cells (Fig. 6e) or Cos7 cells (Supplementary Fig. 7), suggesting a specific binding of Sap10 to DNAM-1 in NK cells.

As demonstrated by cell sorting and ELISA experiments, the binding of Sap10 to DNAM-1 has important immunological consequences. Using CD69 expression as a measure of NK-cell activity⁴², it is obvious that hpNK cells’ activity is significantly augmented with increasing doses of Sap10 and longer incubation times (Fig. 7a). In contrast, when DNAM-1 accessibility is prevented by a DNAM-1-specific antibody, hpNK activation is significantly reduced, further demonstrating the specificity of Sap10 binding to DNAM-1 (Fig. 7b). The remaining activation of hpNK cells observed after treatment with the blocking antibody may be provoked by the anti-DNAM-1 antibody itself. This phenomenon is similar to what has been described for the anti-DNAM-1 antibody NewE, which activates the DNAM-1 and enhances subsequent NK-cell activation⁶⁰ thereby increasing the killing activity of NK cells against tumor cells⁶¹.

The elevated activation level of NK cells was also accompanied by a significantly increased release of perforin and CCL3 after treatment with 1 µg/ml Sap10 or higher (Supplementary Fig. 9a,b). This effect could be partly but significantly abolished by the addition of anti DNAM-1 blocking antibody (Fig. 7c,d). Perforin is a key effector molecule in the NK cell-mediated antifungal response and has been demonstrated to be effective against various fungi, including Aspergillus, Rhizopus, Candida, and Cryptococcus species⁶². While the increase in perforin release was moderate, the strong Sap10-induced increase in CCL3 release is noteworthy. CCL3 plays a multifaceted role in NK cell biology, including regulating their trafficking, enhancing antitumor responses, recruiting other immune cells, and participating in inflammatory processes. Thus, it is suggested that Sap10 binding to DNAM-1 leads to the activation of NK cells and, in addition to perforin-mediated killing activity, provokes the release of a chemokine that will attract additional NK cells^44,63. It should be emphasized that perforin-mediated killing activity depends on different NK-cell subsets. While NCAM-1^dim population of NK cells, that makes up about 90% of circulating NK cells and is considered more mature, is primarily responsible for cytotoxic functions through IS formation, NCAM-1^bright NK-cell population is less cytotoxic but a major cytokine producer⁶⁴. In contrast to NCAM-1^dim NK-cell population, NCAM-1^bright cells typically exhibit high DNAM-1 expression level¹⁵. Thus, our findings are consistent with a model in which DNAM-1 expression appears to bias NK-cell responses toward cytokine secretion, with a comparatively limited impact on direct cytotoxicity.

Based on our current data, NK-cell activation is mediated by Sap10 acting as a ligand for DNAM-1 rather than through its protease activity. In our experiments, NK-cell activation was induced by purified, soluble Sap10 under conditions where the proteolytic activity toward the host is unlikely to be required. Moreover, the observed DNAM-1-dependent binding and activation occurred independently of detectable protease-mediated cleavage events. While we cannot exclude a contributory role for Sap10 enzymatic activity under physiological conditions, our results support a model in which Sap10 primarily functions as a DNAM-1–binding ligand in the context of NK-cell recognition.

In summary, our investigation reveals the involvement of DNAM-1 in the interaction with a fungal pathogen. Though our study focused on the two fungal pathogens A. fumigatus and C. albicans, it is likely that our results can also be transferred to other clinically relevant fungi sharing conserved domains acting as potential binding partners to DNAM-1. Also, we found that Sap10 is a direct interaction partner that provokes activation of NK cells in a dose dependent manner via DNAM-1. However, many open questions remain regarding the specific interaction of NK cells with fungal hyphae.

Maintenance of isolated NK cells requires addition of IL-2, one of the important cytokines involved in the differentiation of hematopoietic stem cells into natural killer cells⁶⁵. IL-2 was shown to increase the proliferation rate without leading to hyperactivity, using Interferon gamma (IFNγ) and macrophage inflammatory protein-1 alpha (MIP-1α; CCL3) as measures⁶⁶. However, IL-2 treatment provokes phosphorylation of the target of DNAM-1, the transcription factor FOXO1, a negative regulator of NK-cell maturation and function⁶⁷. Thus, the effect of the absence or presence of DNAM-1 might - at least partly - be masked by the addition of IL-2 to the medium, which is required for NK-cell proliferation⁶⁶.

In vivo studies using murine infection models by comparing wild-type and DNAM-1–deficient mice, may help to address many of the remaining questions by enabling dose-dependent infection experiments under physiological conditions. In particular, such studies could clarify the role of DNAM-1 in fungal cytotoxicity, including IFN-γ and perforin secretion, as well as its costimulatory nature with TIGIT and CD96. Additionally, several potential fungal interactors identified in the manuscript could be further characterized through in vivo studies. However, interpretation of such experiments must account for species-specific differences, as DNAM-1 is expressed on only a subset of murine NK cells but on the majority of human NK cells⁶⁸, and also compensatory signaling through TIGIT and CD96 may differ between host species.

In human NK cells, the response of DNAM-1 is highly interconnected with other receptors that detect the same ligands. It would be of interest to understand how the fungal ligands orchestrate these different receptors, and which consequences are to be expected. DNAM-1 signaling has been shown to depend on the integrin LFA-1⁵⁰, making it necessary to examine whether LFA-1-DNAM-1 association influences not only NK-cell activation but also cytotoxicity toward fungi. A recent study also revealed interplay between DNAM-1 and NKG2D, another major NK cell receptor and panfungal PRR. NKG2D engagement was found to impair DNAM-1-mediated signaling while simultaneously increasing TIGIT expression⁶⁹.

NK cells constitutively express TIGIT, but expression is upregulated in tumor environment. TIGIT inhibits NK-cell cytotoxicity in both humans and mice and helps in maintaining self-tolerance under conditions of acute inflammation. High TIGIT expression on intratumoral NK cells was associated with tumor progression in patients with colon cancer²³. It is likely that TIGIT and CD96 have a similar influence in the interaction of NK cells with fungal pathogens. While TIGIT/DNAM-1 ligands such as CD155 and CD112 are restricted to mammalian cells and are absent in fungi, TIGIT has been shown to be stimulated by members of the Agglutinin-like protein family of C. albicans⁷⁰. Computational modeling of protein-protein interactions between Sap10 and TIGIT or TACTILE, performed using the Zdock server, yielded high docking scores of 1426.60 and 1277.11, respectively, indicating strong interactions. Interestingly, glycosylation of Sap10 was predicted to enhance its binding to DNAM-1, while concurrently weakening its binding to TIGIT and TACTILE. As a potential consequence, Sap10 glycosylation may induce a functional shift toward NK-cell activation. The receptor-ligand recognition by NK cells is often glycan-dependent. Glycosylation has been shown to affect ligand binding and functional activity of NK cell receptors, such as 2B4⁷¹, and to fine-tune the balance of signaling within the DNAM-1/CD96/TIGIT axis interacting with the CD155/CD112 ligand set⁷². However, in CFU assay, blocking of inhibitory receptors CD96 and TIGIT, as well as corresponding isotype controls did not consistently change fungal viability (Supplementary Fig. 3), suggesting a modulatory but non-dominant role of these receptors in antifungal activity. Still the observed low viability after blocking TIGIT is in line with its function as inhibitory receptor.

Finally, our understanding of NK cell biology is evolving. Different subsets of NK cells have distinct function during infections. For example, in tuberculosis patients, DNAM-1⁺ subsets produced more IFN-γ and CD107a than DNAM-1^- subsets. Additionally, the proportions of DNAM-1⁺ and DNAM-1^- cell subsets differ between healthy individuals and tuberculosis patients highlighting the importance of DNAM-1 in NK cells in the immune surveillance against tuberculosis⁷³. The recent newly introduced definition of various subpopulations of NK cells with distinct functions in immunology will require reassignment of experiments regarding the interaction of NK cell and fungi to the new subsets⁷⁴.

The heterogeneity in NK-cell populations helps to explain our main observations. The observed cell behavior - upregulation of CD69 induced by Sap10-DNAM-1 engagement together with perforin and CCL3 secretion–, matches the characteristics of NK1-like (CD56^dimDNAM-1^pos) subsets, with strong cytotoxic and chemokine-producing features. Subset heterogeneity also clarifies the donor-dependent variations observed in Sap10-driven responses in the hyphal damage and ELISA assays.

As NK-cell subsets differ in adhesion and integrin profiles (e.g., LFA-1) and in cytoskeletal organization, these features may also influence synapse stabilization and thereby explain the limited or absent accumulation of DNAM-1 towards IS. For example, DNAM-1⁺ NK cells may show weaker adhesion, whereas DNAM-1-/TIGIT^high (NK3-like) subsets, may display a more diffuse DNAM-1 signaling, sufficient for activation but insufficient for visible receptor clustering.

Moreover, subsets with high levels of inhibitory receptors (e.g., TIGIT) may counteract DNAM-1 signaling, thereby reducing activation despite ligand binding. This may explain why DNAM-1 blockade or knockout affected only selected experiments, likely due to compensatory activation via other activating receptors or variation in TIGIT expressions, both of which can modulate DNAM-1 responses.

Further investigations will be required to understand which NK cell subsets are involved in combating fungal infections. One may speculate that different subsets bind to the fungus or release cytokines to alert neighbored immune cells upon binding of soluble fungal components.

Methods

Fungal culture

Fungal culture was performed under sterile conditions. A. fumigatus colonies (ATCC reference strain #46645) were grown in a ventilated T75 cell culture flask filled with 25 ml potato glucose agar (PDA, Sigma-Aldrich, #102282563) or in standard Petri dishes on beer wort agar (Oxoid, PO5055A) and incubated at 35 °C until conidiophores developed. Conidia suspensions were prepared by rinsing the plates with pure water and filtering the mixture through a 40-µm cell strainer. Resting spores were counted (2–8 × 10⁸ cells/ml) and preserved at 4 °C until further use.

For immunoassays, conidia were germinated in Roswell Park Memorial Institute medium (RPMI, Thermo Fischer Scientific, #R7509-500ML) medium (20 ml in 50 ml tubes) at 25 °C with shaking at 200 rpm until small germ tubes were visible. The germ tubes were harvested by centrifugation (5000 × g, 10 min) and resuspended in RPMI medium at a defined concentration of 4 × 10⁶ germ tubes/ml.

For microscopic observations, spores were transferred to KOH-treated⁷⁵ 8-well cover-glass chambers, incubated overnight at 34 °C in a humid condition, resulting in small germ tubes for fungus only or co-culture experiments.

For experiments with C. albicans, the strain SC5314 (ATCC, #ATCC-MYA-2876) was used. YPD broth (Yeast peptone dextrose broth; Sigma-Aldrich, #Y1375) was inoculated directly and incubated for 16 h at 30 °C, 120 rpm in a thermal-shaker. The overnight culture was harvested and washed by centrifugation at 4500 rpm for 5 min, resuspended in 1x PBS (Sigma-Aldrich, #D8537-500ML), cell number was counted and diluted in RPMI colorless medium. The cells were then transferred to KOH-treated 8-well chambers and incubated at 37 °C and 5% CO₂ for 2–3 h for hyphal formation.

Isolation of primary NK cells and cell culture

The collection and processing of human peripheral venous blood from healthy adult donors was approved by the Ethics Committee of the University Hospital Würzburg (#302/12). Peripheral venous blood was collected from donors after obtaining informed consent. All procedures involving human participants were conducted in accordance with the relevant ethical regulations and guidelines. Human peripheral blood mononuclear cells (hPBMCs) were isolated from leukoreduction system (LRS) chambers from plateletpheresis donations from healthy individuals as described in detail before^10,75. From hPBMCs, human primary natural killer (hpNK) cells were isolated using the MACS NK negative selection kit (Miltenyi Biotech, #130-092-657) according to the manufacturer’s protocol. These NK cells were cultured in RPMI medium (10⁶ cells/ml) supplemented with 10% fetal bovine serum (FBS; Sigma-Aldrich, #F7524-500ML) and 120 µg/ml gentamicin (Sigma-Aldrich, #G1397) at 37 °C and 5% CO₂. Prior to stimulation, NK cells were incubated overnight with 1000 U/ml IL-2 (Novartis). NK cells were cryopreserved by centrifugation (300 × g, 10 min) followed by resuspension in ice-cold freezing medium (90% fetal calf serum and 10% dimethyl sulfoxide) at a final concentration of 3 × 10⁶ cells/ml (DMSO, Roth, #A994.2)⁷⁵. Aliquots were transferred to cryotubes and placed in a controlled-rate freezing container (Mr. Frosty) before cooling to −80 °C for 24 h. Cryotubes were subsequently transferred to liquid nitrogen for long-term storage until use.

COS7 cells (from African Green Monkey, Cytion, #605470) were cultured in DMEM (Dulbecco’s Modified Eagle’s Medium; F-12 Ham, Sigma-Aldrich, #D8062-500ML) supplemented with 10% FBS and 1% P/S (Penicillin-Streptomycin, P/S, Sigma-Aldrich, #P4333-100ML) and incubated at 37 °C and 5% CO₂ until 80–90% confluency was gained. For the experiments, approximately 2 × 10⁴ COS-7 cells were seeded per well after splitting for further processing.

Binding of DNAM-1 to C. albicans and A. fumigatus germ tubes

C. albicans or A. fumigatus germ tubes were incubated with either 2 µg/ml of soluble DNAM-1 (27.9 kDa; Human DNAM-1/CD226 Protein, His-tag; Acrobiosystems, DN1-H52H6-100UG) or BSA (Bovine Serum albumin; Sigma-Aldrich, #A7030) or untreated for 2 h in RPMI medium (colorless) at 37 °C and 5% CO₂. After washing in RPMI medium (colorless) two times, samples were fixed with 1% formaldehyde (FA, methanol-free, 16%, Thermo Fischer, #28906) in RPMI medium for 10 min at room temperature (RT). After washing and blocking (with 2.5% BSA in Phosphate-buffered saline, 1× PBS), the samples were stained with mouse anti-human CD226-Alexa647 antibody (5 µg/ml in blocking solution; BD Biosciences, #564797) for 1 h at RT. Samples were washed in PBS, covered with PBS and stored at 4 °C until further use.

Generation of DNAM-1 knock-out (ΔDNAM-1) NK cells

Upon isolation, primary NK cells were cultured in NK MACS basal medium (10⁶ cells/ml) supplemented with 1% NK MACS Supplement (NK MACS medium with supplements, Miltenyi Biotech, #130114429), 5% human serum (Sigma-Aldrich, #H4522-100ML) and 500 U/ml recombinant IL-2 (PeproTech, #20002-50UG) at 37 °C and 5% CO₂. Primary NK cells were washed twice with PBS and suspended at 5–10 × 10⁷ cells/ ml in T Buffer (Neon kit, Thermo Fischer, #N10096). Cells were electroporated in the presence of pre-assembled Cas9 protein and DNAM-1 guide RNAs (IDT) using two pulses (2050 V/20 ms and 500 V/100 ms). Pre-warmed medium (37 °C) was added to the cells, followed by 500 U/ml IL-2 after 30 min. Seven days after electroporation, DNAM-1 deletion was confirmed by flow cytometry

NK-cell immobilization on Poly-D-lysine-coated glass

hpNK cells and ΔDNAM-1 were immobilized in 8-well cover glass chambers (coated with 0.05% poly-D-lysine; PDL, Sigma-Aldrich, #P6407-5MG) in a density of around 200,000–250,000 cells per well. For recovery and stability, NK cells were incubated for at least 3 h at 37 °C, 5% CO₂ before immune staining and fixation.

Co-culture of NK cells and A. fumigatus

hpNK cells and A. fumigatus germ tubes were co-incubated in RPMI-1640 medium supplemented with 10% FBS in 8-well cover glass chambers at 37 °C, 5% CO₂ for 6 h in all the experiments. The multiplicity of infection (MOI) was maintained at 0.5 for the co-culture in all cases. For co-cultures with ΔDNAM-1 NK cells, the same method was used.

Staining of NK-cell surface proteins

Immobilized NK cells or NK cell—A. fumigatus co-cultures were blocked with 2.5% BSA (in PBS) for 30 min. After blocking any unspecific binding, the samples were stained with respective antibodies in the concentration of 5 µg/ml in blocking solution (DNAM-1: Alexa647 anti-human CD226; NCAM-1: Alexa488 and Alexa647 anti-human CD56; both from Biolegend, #318311 and #318313). Staining was performed at 4 °C on ice for 1 h. Thereafter, samples were washed with blocking solution. After washing, fixation was done with 1% FA in PBS for 20 min at RT and washed in 1x PBS afterwards. The cell wall of the fungus was stained with calcofluor white (Fluorescent Brightener 28; Sigma-Aldrich, #F3543) in a concentration of 5 µg/ml in PBS for 30 min at 4 °C. After 30 min, the solution was pulled off and replaced with PBS and stored at 4 °C until the imaging was done.

Hyphal metabolic activity (XTT) assay

A colorimetric assay was used to assess the role of DNAM-1 in fungal damage, as described by Schmidt et al.⁴ with some modifications. Germ tubes of C. albicans were prepared as described above, but in 96-well plates. In certain conditions, DNAM-1 accessibility was inhibited using a blocking antibody (purified anti-human DNAM-1, clone 11A8, Biolegend, #338302). Isolated, IL-2 stimulated primary NK cells were co-incubated with or without blocking antibody (10 µg/ml) and added to the germ tubes at an effector-to-target (E:T) ratio of 20:1, followed by incubation for 6 h at 37 °C with 5% CO₂. Germ tubes or hyphae alone, without effector cells, served as the negative control. As a positive control, germ tubes were incubated for 2 h with 2.4 mM of H₂0₂ (30%, Merck, #1.07209), as described by Ruf et al.⁷⁶. Absorbance was measured spectrophotometrically at 450 nm with a reference at 650 nm using a Tecan Infinite M200 plate reader, and hyphal damage was calculated as a percentage relative to the negative control, according to the formula described by Schmidt et al.⁴.

NK cell–mediated antifungal activity (CFU assay)

A CFU (colony forming unit) assay was established to determine NK cell–mediated antifungal activity via the receptors DNAM-1, TIGIT, and CD96. NK cells (effector) and C. albicans (target) were prepared as described above. NK cells were pre-incubated with blocking antibodies against DNAM-1, TIGIT (mouse antihuman TIGIT, VSTM3; Biolegend, #372702), or CD96 (mouse anti human CD96; Biolegend, #338402), or with their corresponding isotype controls (10 µg/ml), for 30 min at 37 °C and 5% CO₂. NK cells were then added to C. albicans hyphae at an E:T ratio of 20:1 and coincubated for 12 h under the same conditions. A control condition without NK cells was included. After coincubation, NK cells were lysed and fungal clumping reduced by adding 0.05% Triton X-100 (Triton^TM X-100, 10%, Thermo Fischer, #28314) and incubating for 2–3 min at RT. Following gentle mixing, serial dilutions (10⁻¹–10⁻⁴) were prepared in PBS. From each dilution, 100 µl was plated onto YPD agar plates (in technical duplicates, spread evenly with a sterile spreader). Plates were allowed to absorb for 30 min, then incubated at 30 °C for 24–48 h in an inverted position. Colonies were counted, and CFU values were calculated using:

$$CFU/ml=\frac{NumberofcoloniesxDilutionfactor}{Volumeplated\left(ml\right)}$$ 1

Where the plated volume refers to 0.1 ml.

Antifungal activity (killing) was calculated by comparing wells containing NK cells and C. albicans with wells containing C. albicans alone, using:

$$\%Killing=\frac{CAonly-CA+NK}{CAonly}$$ 2

Statistical significance was then assessed by comparing wells treated with blocking antibodies to their respective isotype controls.

Treatment of NK cells with Sap10

Soluble, truncated Sap10 protein was heterologously expressed in Pichia pastoris and purified following the protocol described by Albrecht et al.¹⁹. After transformation and production of recombinant enzymes, culture supernatants were harvested and dialyzed against 20 mM sodium citrate buffer (pH 6.5) containing NaCl (buffer A). The proteases were subsequently purified using an Acta explorer FPLC system (Amersham Biosciences) on a column equilibrated with buffer A. Protein-containing fractions were collected, and purified Sap10 was used for downstream experiments.

For microscopical analysis, immobilized hpNK cells, ΔDNAM-1 NK cells, or COS7 cells (which lack DNAM-1 on their surface⁴¹), were blocked with 2.5% BSA in 1× PBS and co-incubated (on ice, 30 min) with Sap10 protein (5 µg/ml, conjugated with Alexa594; Alexa Fluor 594 NHS-ester, Thermo Fischer Scientific, #A20004). As labeled Sap10 protein tends to assemble into clusters, experiments with Sap10 had to be done with freshly labeled and purified protein only. After treatment with Sap10, cells were carefully (to avoid cell loss) washed with 1× PBS. DNAM-1 surface protein was stained with an anti-CD226 antibody (Alexa647, 5 µg/ml in blocking solution) at 4 °C for 1 h and washed in PBS after that. Then, cells were fixed (1% FA in PBS, 30 min, RT) and washed and stored in PBS at 4 °C until imaging was done.

NK cells used in ELISA were co-incubated with Sap10 in a dose-dependent manner (0.5-4 µg/ml) for 24 h at 37 °C. As a control, several stimulating agents, specifically Phorbol 12-myristate 13-acetate (25 ng/ml; PMA; Sigma-Aldrich, #P1585) and ionomycin (1 µg/ml; Sigma-Aldrich, I0634) were used instead of Sap10. Sap10, PMA and ionomycin were diluted in RPMI + 10% FBS media. To analyze perforin and chemokine release from stimulated NK cells, cell samples were harvested and centrifuged at 300 × g for 10 min. Culture supernatants were cryopreserved at −80 °C until further analysis.

NK cells used in flow cytometry were treated as described for ELISA. In addition, in some experiments, accessibility of DNAM-1 was blocked by means of a blocking antibody. Cells were treated with purified anti-human CD226 (DNAM-1) antibody or the corresponding isotype control for 30 min before staining with fluorescent antibodies (see section “Flow cytometry”).

Co-Immunoprecipitation assay for DNAM-1 and Sap10 protein interaction analysis

Pierce^TM Co-Immunoprecipitation kit (Co-IP; Thermo Fischer, #26149) was used to determine the interactions between DNAM-1 and Sap10, as per the manufacturer’s protocol with appropriate modifications. For antibody coupling, purified anti-human CD226 (DNAM-1) was coupled according to the protocol, using a concentration of 10 µg/ml. After antibody immobilization, Co-IP was performed with Atto643-labeled Sap10 (see Sap10-dye conjugation for experimental details) and DNAM-1 (his-tagged) at a final concentration of 100 µg. As a control, immobilized antibody was incubated with only Sap10-Atto643. To avoid disruption of the complex by detergents, 1x PBS was used in place of the IP Lysis/wash Buffer specified in the protocol. Afterward, elution of the Co-IP was executed, and 10x PBS (50 µl per sample) was added to the eluates to stabilize the samples. Flow cytometry was used as the read-out system. To exchange the buffer for flow cytometry, 7 K MW desalting columns (Zeba spin; Thermo Fischer, #89883) were used. The flow-through was collected in 50 µl of 1x PBS. The protein mixture was then kept at 4 °C until analysis by flow cytometry.

Fluorescence microscopy

A confocal laser scanning microscope was used (CLSM, Zeiss LSM700), equipped with a plan- apochromat 63×/1.40 oil immersion objective. The size of all the images was 71.35 × 71.35 μm with a laser power of 8% and 650 master gain, pinhole 1 AU (53.7). Z-stacks of the images were taken with an interval of 1.73 μm.

Along with the CLSM, Airyscan (Zeiss LSM 900) was used in super-resolution mode, with a plan-apochromat 63×/1.40 oil immersion objective. A standard imaging setup with a scan speed of 1 μs was used. For most of the experiments, laser power used was 0.3–0.5% (for all channels) with a master gain of 850 V. Z-stacks of the images were taken with an interval of 0.17 μm, with 40–70 slices, and the range between 7 and 10 μm.

Sectioning and analyzing were done by using ZEN software (Zeiss Zen 3.10) and Image J (1.54f, Wayne Rasband). For visualization, maximum intensity projections of z-stacks were used in both cases.

dSTORM imaging was performed for both naïve NK cells and in co-culture with A. fumigatus. A photo-switching buffer (100 mM Cysteamine hydrochloride; Sigma-Aldrich, #M6500-25G; in PBS, pH 7.4) was used for the imaging of the samples. 2D measurements were accomplished on an inverted wide-field fluorescence microscope (IX – 71; Olympus), as described previously by Ziegler et al.¹¹. Approximately 15,000 to 20,000 images were captured for each measurement, with an exposure time of 20 ms and an irradiation intensity of ~ 7 kW/cm², using highly inclined and laminated optical sheet (HILO) illumination. The open-source software rapidSTORM 3.3 (http://www.super-resolution.biozentrum.uni-wuerzburg.de/) was used to reconstruct the final images of all 2D dSTORM measurements.

Enzyme-linked immunosorbent assay (ELISA)

Concentrations of CCL3 (MIP-1α; R&D Systems, #DY270-05) and perforin (Abcam, #ab83709) were determined using ELISA kits according to the manufacturer’s instructions with minor modifications as described below. ELISAs were performed in 96-well half-area, high-binding plates with one-fourth of the volume recommended by the manufacturer. Technical duplicates were used for each condition and three independent donors were analyzed. Absorbance was measured with a NanoQuant Infinite 200 M Pro microplate reader (Tecan, Maennedorf, Switzerland).

Flow cytometry

NK cell activation was determined by flow cytometry. NK-cell cultures were centrifuged for 10 min at 300 × g at RT, and cell pellets were resuspended in 100 µl of HBSS (Hanks’ balanced salt solution; Thermo Fischer Scientific, #14175095) supplemented with 2 mM EDTA (Sigma-Aldrich, #ED2P-100G) and 1% FBS. Cells were then extracellularly stained with fluorescent antibodies and incubated in the dark at 4 °C. Thereafter, cells were washed with 2 ml HBSS supplemented with 2 mM EDTA and 1% FBS, resuspended in 150 µl of 4% paraformaldehyde (Sigma-Aldrich, #47608), and incubated for 30 min at RT.

For analysis of surface expression in ΔDNAM-1 NK cells and proof of DNAM-1 deletion, cells were treated with anti-CD56 FITC and anti-CD226 APC antibodies (Miltenyi Biotech, #130-114-549 and #130-117-490). Briefly, NK-cells were centrifuged for 10 min at 300 × g at RT and cell pellets were resuspended in 100 μL HBSS supplemented with 2 mM EDTA and 0.5% FBS (FACS buffer). Cells were stained in the dark for 15 min at 4 °C, washed with FACS buffer, resuspended in 150 μl of FACS buffer and analyzed.

Flow cytometric analysis was performed with a CytoFLEX (Beckman Coulter Life Sciences; isolation of ΔDNAM-1 NK cells and DNAM-1 fluorescence intensity after exposure to A. fumigatus or Sap10 at different time points, dose-dependent Sap10 stimulation) or a FACS Calibur (Treestar/Becton & Dickinson, BD; Sap10 stimulation and effect of DNAM-1 blocking antibody on Sap10-mediated NK-cell activation) or BD FACS Aria III (BD; co-immunoprecipitation assay and DNAM-1 binding to fungal germ tubes). Downstream data analysis was performed using FlowJo 10 (Treestar/Becton & Dickinson, BD) or Kaluza v.2.1 (Beckman Coulter, Brea, CA, USA). The gating strategies used in the flow cytometry experiments are shown in Supplementary Figs. 11–15.

Sap10-dye conjugation

To check the interactions between Sap10 and DNAM-1 by Co-IP, fluorescence correlation spectroscopy (FCS), and fluorescence microscopy, purified Sap10 protein was conjugated with specific dyes. The conjugation process utilized 0.5 ml spin desalting columns (Zeba spin desalting columns) following the manufacturer’s standard protocol. In brief, 100 µl Sap10 was reconstituted in labeling buffer (100 mM NaHCO_3; Fischer Scientific, #1545049). Subsequently, respective dyes (in 5 M excess) were added to the protein and incubated in the dark for 2 h at RT. ATTO643 (ATTO643-NHS-ester; Atto-Tec, #AD 643) and Alexa594 NHS ester dyes were used for FCS, Co-IP; and binding experiments, respectively. Protein conjugates were purified and washed again using the 7 K MWCO columns in PBS to remove untethered dyes. Finally, protein concentration and the degree of labeling (DOL) were determined by UV–vis absorption spectrometry (Implen, NanoPhotometer P-Class).

Fluorescence correlation spectroscopy (FCS)

The Atto643-labeled Sap10 samples were always purified prior to measurements using spin-desalting columns (7 K MWCO) in PBS (pH 7.4) to remove protein aggregates. The purified solution was then diluted to 1 nM in PBS. For samples with dye-labeled Sap10 and DNAM-1 (His-Tag; Acrobiosystems, #DN1-H52H6-100UG), the required DNAM-1 protein was added and incubated on ice for 30 min. The samples were then transferred onto a microscope slide and covered with a coverslip.

FCS measurements were conducted on a custom-built inverted confocal microscope (Zeiss Axiovert 100 TV) with a 640 nm diode laser (Coherent Cube). The excitation beam was coupled into an oil-immersion objective (Zeiss Plan Apochromat, 63×, NA 1.4) by a dichroic beam splitter (Omega Optics, #645DLRP). The average laser power set to 400 µW. The fluorescence signal was collected by the same objective and filtered by a band-pass filter (Omega Optics, #700DF75) before being split onto two avalanche photodiodes (PerkinElmer, APD, SPCM-AQRH-14-FC), using a 50:50 non-polarizing beam splitter cube (Thorlabs), coupled to two multi-mode optical fibers with an entrance diameter of 100 μm that served as a confocal pinhole. The sample temperature was maintained at 25 °C, and at least five autocorrelation functions were recorded for each sample over a minimum of 5 min.

The interaction between Sap10 and DNAM-1 was analyzed using the software OriginPro 2023b (OriginLab Corporation) by comparing the characteristic diffusion time τ_D of mixed protein samples. The correlation function G(τ) of fluorescence fluctuations from N diffusing molecules was fitted using:

$$G_D\left(\tau \right)=N^{-1}{\left(1-\frac{\tau }{{\tau }_D}\right)}^{-1}$$ 3

Where N is the average number of molecules in the detection volume and τ_D = ω_xy²/4D the average diffusion time (lateral radius of ω_xy²) and the diffusion coefficient D given by the Stoke–Einstein equation^77,78. In cases of additional intensity fluctuations due to dye interactions, a multiexponential equation was applied:

$$G\left(\tau \right)=G_D\left(\tau \right)*\left(1+{\sum }_n{a}_n{e}^{\left(-\frac{\tau }{{\tau }_n}\right)}\right)$$ 4

Where a_n and τ_n represent the amplitude and time constant of each relaxation, respectively.

Computational analysis of protein–protein interaction

Identification of potential protein–protein interactions of DNAM-1 with C. albicans and A. fumigatus was performed by domain-pair-based inference. The domain–domain interaction of the immunoglobulin V-set domain of DNAM-1 was obtained from the three-interacting domains (3did, Patrick Aloy) database²⁸ and all interacting domains were mapped to fungal proteins using UniProt (EMBL-EBl) domain annotation⁷⁹. To focus on proteins with the potential to interact in vivo and to reduce false-positive predictions, the network was refined by filtering the fungal interactors for cell-surface location using Gene Ontology (GO; GO Consortium) functional annotation⁸⁰. The remaining interactions were then manually screened for candidates with the potential to undergo pathogenicity-related processes based on characteristics like the description, function, mutant phenotype, and gene ontology obtained from the UniProt database.

The truncated sequence of Sap-10 was modeled through homology modeling, followed by the N-glycosylation sites were predicted by using online tool GlycoEP server and verified using NetNGlyc 1.0 (DTU Health Tech). Next, D-Mannose residues were added to the predicted sites (residues 128, 208, 211, 398) using CharmmGui Glycan Reader and Modeler module (Dr. Im’s research team). After visualization of the glycosylated Sap-10 using CharmmGui PDB Reader, the structure was processed by energy minimization and structural optimization. The optimized structures were subjected to blind docking with DNAM-1 protein (CD-226, PDB ID:6ISA) using Z-dock server (Weng Lab). To calculate the binding energy of the complex, the docked complex structures were subjected to PRODIGY. The interactions were analyzed by visualizing the structures through Pymol (Schrödinger, LLC) and the interactive surfaces of the protein–protein complex were analyzed using PDBePISA tool (EMBL-EBl).

Statistics and reproducibility

Data compilation, analysis, and visualization were conducted using Microsoft Excel 365, OriginPro 2021b, and GraphPad Prism version 10. To assess statistical significance, we employed various methods after approving normality by means of a Shapiro-Wilk test: two-way analysis of variance (ANOVA) for multi-group comparisons and repeated measures one-way ANOVA for matched multi-group comparisons and student’s t tests for connected samples (standard and paired). For post-hoc analysis in multi-group comparisons, Dunnett’s test was applied when comparing against a single control group, while Tukey’s test was used for all pairwise comparisons. Detailed descriptions of the statistical tests used are provided in the figure legends. Significance was determined with adjusted p-values less than 0.05.

All experiments were conducted at least three times (n = 3) to ensure statistical relevance and reproducibility. Sample sizes and technical replicates varied depending on the method used and are specified in the corresponding figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization, F.N. and U.T.; methodology, F.N., L.H., D.A.H., C.L., J.S., B.K., M.D., L.K.; and C.S. investigation, F.N., L.H., D.A.H., C.L., J.S., B.K., M.D., L.K.; and C.S., writing—original draft, U.T., and F.N.; writing—review & editing, all the co-authors; funding acquisition, M.S., J.L. and U.T.; resources, B.H., M.Di., T.D., M.S., J.L., and U.T.; supervision, M.Di., T.D., M.S., J.L., and U.T.

Peer review

Peer review information

Communications Biology thanks Ofer Mandelboim and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Si Ming Man, George Inglis, and David Favero. A peer review file is available.

Data availability

The materials that support the findings of this study are available upon request from the corresponding author. All relevant data are included in the manuscript and the supplementary information. Image stacks and raw data used in Figs. 1–7, and Supplementary Figs. 1–3, 6–10 can be downloaded from Figshare (10.6084/m9.figshare.c.8210864).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-10056-8.