Hexokinase-I directly binds to a charged membrane-buried glutamate of mitochondrial VDAC1 and VDAC2

communications biology Article https://doi.org/10.1038/s42003-025-07551-9 Hexokinase-I directly binds to a charged membrane-buried glutamate of mitochondrial VDAC1 and VDAC2 Check for updates Sebastian Bieker 1,2,6, Michael Timme1,2,6, Nils Woge1,2,DinaG.Hassan 1,2,3, Chelsea M. Brown 4, Siewert J. Marrink 4,ManuelN.Melo 5& Joost C. M. Holthuis 1,2 Binding of hexokinase HKI to mitochondrial voltage-dependent anion channels (VDACs) has farreaching physiological implications. However, the structural basis of this interaction is unclear. Combining computer simulations with experiments in cells, we here show that complex assembly relies on intimate contacts between the N-terminal α-helix of HKI and a charged membrane-buried glutamate on the outer wall of VDAC1 and VDAC2. Protonation of this residue blocks complex formation in silico while acidification of the cytosol causes a reversable release of HKI from mitochondria. Membrane insertion of HKI occurs adjacent to the bilayer-facing glutamate where a pair of polar channel residues mediates a marked thinning of the cytosolic leaflet. Disrupting the membrane thinning capacity of VDAC1 dramatically impairs its ability to bind HKI in silico and in cells. Our data reveal key topological and mechanistic insights into HKI-VDAC complex assembly that may benefit the development of therapeutics to counter pathogenic imbalances in this process. Voltage-dependent anion channels (VDACs) are abundant β-barrel proteinsintheoutermembraneofmitochondria(OMM)thatserveasthemain conduits for the large flux of ions, ATP/ADP, NAD +/NADH and Krebs’ cycle intermediates from and into mitochondria1,2.Inmammals,three isoforms exist (VDAC1-3) with non-redundant functions3,4.VDAC1and VDAC2 are the most abundantly expressed isoforms in most tissues. Besides their central role in controlling the flow of metabolites across the OMM, both isoforms act as scramblases that mediate phospholipid import into mitochondria5. Additionally, VDAC1 and VDAC2 function as dynamic translocation platforms for a variety of proteins that control the permeability of the OMM for cytochrome cto either promote or prevent mitochondrial apoptosis. VDAC binding partners include the proapoptotic Bcl-2 proteins BAX and BAK6–8, which mediate the decisive step in OMM permeabilization by which cytochrome cand other apoptogenic factors are released into the cytosol to trigger the apoptotic cascade9. Moreover, ceramides, central intermediates of sphingolipid metabolism, exert their pro-apoptotic activity, at least in part, by interacting directly with VDAC210. VDAC1 and VDAC2 also function as the physiological receptors of hexokinases (HKs). These enzymes phosphorylate glucose to generate glucose-6-phosphate (G-6-P), an ATP-dependent reaction that serves as entry point for glucose into the glycolytic pathway for energy production or, alternatively, into the pentose phosphate pathway to generate anabolic intermediates11. Elevated levels of mitochondrially bound HK isoforms HKI and HKII lead to ahigh rate of glycolysis and lactate production,a metabolic signature referred to as the Warburg effect12. This metabolic switch from oxidative to glycolytic metabolism is a central hallmark of tumor progression, allowing pre-malignant lesions to maintain a high metabolic rate in oxygen-deprived avascular environments13–15. Moreover, mitochondrially bound HKs protect cancer cells from drug-induced mitochondrial apoptosis by diminishing the propensity of VDACs to interact with proapoptotic Bcl-2 proteins BAX and BAK16–18. Conversely, a reduction in HKI concentrationin thespinalcordis thought toenhance binding ofVDAC1to specific amyothrophic lateral sclerosis type I-associated variants of superoxide dismutase 1 (SOD1), thereby promoting formation of toxic SOD1 aggregates, mitochondrial dysfunction and cell death in motor neurons19,20. The importance of HKI-VDAC interactions in carcinogenesis and neurodegenerative disease has prompted a search for small molecules and peptides capable of disrupting or stabilizing this protein-protein complex21–23. However, these efforts are hampered by a lack of structural 1Molecular Cell Biology Division, Department of Biology/Chemistry, University of Osnabrück, 49076 Osnabrück, Germany. 2Center for Cellular Nanoanalytics, Osnabrück University, Artilleriestraße 77, 49076 Osnabrück, Germany. 3Department of Environmental Medical Sciences, Faculty of Graduate Studies and Environmental Research, Ain Shams University, Cairo, Egypt. 4Groningen Biomolecular Sciences and Biotechnology Institute, University of Groningen, Nijenborgh 7, 9747 AG Groningen, The Netherlands. 5Instituto de Tecnologia Química e Biológica António Xavier, Universidade Nova de Lisboa, Av. da República, 2780-157 Oeiras, Portugal. 6 These authors contributed equally: Sebastian Bieker, Michael Timme. e-mail: [email protected];[email protected] Communications Biology | (2025) 8:212 1 1234567890():,; 1234567890():,; insights into how HKI and VDAC assemble into a complex. Like HKII, HKI contains a short N-terminal, 20-amino acid hydrophobic α-helix that enables OMM binding, presumably through its interaction with VDACs24–26. Two protein-protein docking studies reported models for complex formation based on a direct plugging of the N-terminal helices of HKI/HKII into the pore of VDAC127,28.Asignificant shortcoming of these models is that they fail to address a critical role of a membrane-buried glutamate at position 73 (E73) located on the outside wall of VDAC1 in HKI binding29,30. Moreover, a molecular docking simulation study revealed a high-affinity binding site for a peptide mimicking the N-terminusof HKI on the outside wall of VDAC1 in close proximity to the bilayer-facing E73 residue31. Another modeling study postulated that HKII initially binds the OMM through insertion of its hydrophobic N-terminus into the cytosolic leaflet and then interacts with the outer wall of VDAC1 to form a binary complex32. Whether the interaction of HKI with VDACs follows a similar scenario remains to be established. At present, the membrane topology or sidedness of VDAC channels has not been definitively assigned, with complementary experimental approaches yielding divergent and contradicting results33–35. Knowledge of the actual topography of VDACs is a prerequisite for any comprehensive analysis of their role as mitochondrial scaffolds for a broad variety of proteins. Here, we combined molecular dynamics simulations with experimental studies in cells to define the structural and topological determinants that govern HKI binding to VDAC1 and VDAC2. We find that complex assembly critically relies on direct interactions between the N-terminal αhelix of HKI and a membrane-buried, deprotonated glutamate on the outer wall of both channel isomers. Protonation of this residue abolished complex assembly in simulations. Consistent with this result, we show that VDACdependent mitochondrial translocation of a reporter carrying the N-terminal α-helix of HKI is exquisitely sensitive to fluctuations in cytosolic pH. Moreover, we find that a pair of polar channel residues flanking the membrane-buried glutamate causes a marked thinning of the cytosolic leaflet, providing a low-energy passageway for HKI to facilitate complex assembly. Taken together, our data offer fundamental mechanistic insights into HKI-VDAC complex formation and indicate that the C-termini of VDAC channels must face the intermembrane space to provide functional binding platforms for HKI. Results A membrane-buried Glu in VDACs is critical for stabilizing the mitochondrial pool of HKI The bulk of HKI normally resides on mitochondria, with VDACs serving as essential binding platforms. While VDAC1 is widely viewed as principal HKI docking site, the role of VDAC2 is less well defined. As expected, GFPtagged HKI expressed in HeLa cells extensively co-localized with the OMM marker Tom20 (Fig. 1a; Supplementary Fig. 1a). Removal of either VDAC1 orVDAC2didnotsignificantly affect mitochondrial localization of HKIGFP. However, loss of both channels abolished mitochondrial localization of the enzyme and caused its accumulation in the cytosol, even though a portion of the enzyme was found associated with the ER and plasma membrane (Fig. 1a, b; Supplementary Fig. 1a, b; Supplementary Fig. 2a). Moreover, endogenous HKI protein levels were significantly reduced in VDAC1/2 double KO cells while subcellular fractionation experiments showed that in these cells, endogenous HKI primarily resides in the cytosol (Supplementary Fig. 2c, d). Reintroducing VDAC1 or VDAC2 into VDAC1/2 double KO cells restored both mitochondrial localization and expression of HKI (Fig. 1a, b; Supplementary Fig. 2e). These data indicate that VDAC1 and VDAC2 each contribute to stabilizing the mitochondrial pool of HKI. Both VDAC1 and VDAC2 harbor a uniquely positioned glutamate (Glu)inthetransmembraneregionofβ-strand 4 –Glu73 in VDAC1 and Glu84 in VDAC2 –that faces the bilayer’s hydrophobic core. Prior work revealed that Glu73 in VDAC1 is required for HKI binding30. Consistent with this, substitution of Gln for Glu73 in VDAC1 abolished its ability to restore mitochondrial localization and expression of HKI in VDAC1/2 double KO cells (Fig. 1a, b; Supplementary Fig. 2e). Likewise, a VDAC2 mutant in which Gln was substituted for Glu84 failed to stabilize the mitochondrial HKI pool. In contrast, substitution of Asp for Glu73 in VDAC1 or Glu84 in VDAC2 yielded a channel that supported mitochondrial recruitment of HKI to a level beyond that observed for its wild type counterpart (Fig. 1a, b). Taken together, these results suggest that HKI binding to VDAC1 and VDAC2 critically relies on a negatively-charged, membrane-buried Glu residue on the outer channel wall. Mitochondrial recruitment of HKI is mediated by its N-terminal α-helix HKIcontainsanN-terminalα-helixof20-aminoacids(HKI-N)thatenables binding to the OMM24, presumably by interacting directly with VDACs. As expected, a truncated HKI variant lacking this region (HKIΔ2-14) failed to localize to mitochondria and displayed a cytosolic distribution (Fig. 1c). To confirm that HKI-N alone is sufficient for mitochondrial localization, we fused the 17 N-terminal amino acids of HKI to a HaLo-Tag and expressed the construct in HeLa cells. In wild-type cells, HKI-N extensively colocalized with OMM marker Tom20. In contrast, when expressed in VDAC1/2 double KO cells, HKI-N failed to target mitochondria and localized to the cytosol (Fig. 1d). This indicates that HKI binds VDACs primarily via its N-terminal helix, possibly involving direct contact with the bilayer-facing Glu (Fig. 1e). The N-terminal helix of HKI has been shown to bind membranes even in the absence of VDACs, presumably owing to its partially hydrophobic nature36. This implies that mitochondrial recruitment of HKI involves two consecutive steps, namely insertion of its N-terminal helix in the cytosolic leaflet of the OMM followed by VDAC binding to form a binary complex. HeliQuest analysis37 revealed that HKI-N forms an α-helix with an apolar face composed mostly of non-polar and hydrophobic residues and a polar face primarily containing hydrophilic and charged residues (Fig. 2a, b). The amphipathic nature of HKI-N predicts a membrane binding mode whereby its apolar face engages with the hydrophobic membrane core and the polar face with the lipid head groups (Fig. 2c). To investigate the membrane binding affinity of the HKI-N, we performed coarse-grain molecular dynamics (CG-MD) simulations using the Martini3 forcefield38,39. A bilayer mimicking the OMM was constructed40 and an α-helical peptide comprising HKI-N with an additional Gln at its Cterminus (corresponding to Gln18 in HKI) was restrained onto the cytosolic membrane surface. After lifting the restraints, the desorption of the peptide into the aqueous phase was monitored over time41.TheHKI-Npeptide remained membrane-bound, with its apolar face buried into the hydrophobic membrane core and with residence times of >5000 ns. Leu7 is a key component of the membrane-oriented HKI-N apolar face, sitting at its very center (Fig. 2b), and thus likely in constant contact with the hydrophobic membrane core. Substitution of Gln for Leu7 shortened the HKI-N membrane residence time to ~350 ns (Fig. 2d), supporting a critical role of the apolar face in membrane binding. Moreover, substitution of Gln for Leu7 in GFP-HKI abolished its mitochondrial localization in HeLa cells (Fig. 2e, f). Together, these data suggest that membrane insertion of its N-terminal αhelixisaprerequisiteforHKIbindingtoVDACintheOMM. HKI-N binding to VDACs is directly controlled by protonation of the membrane-buried Glu To elucidate the structural basis of HKI-VDAC complex formation, we next performed CG-MD simulationsof HKI-N binding to VDAC1 and VDAC2. Astheforegoing experimentssuggestedthatHKI-VDACcomplexassembly requires a negatively-charged, membrane-buried Glu residue on the outer channelwall(Fig.1a, b), we first set out to estimate the pKa values of the correspondingGlu residuesin VDAC1and VDAC2using titratableMartini simulations42. This revealedthat the pKa valueof Glu73inVDAC1isshifted compared to a free glutamate in solution43 butbylessthanoneunit,i.e.from 4.3 to ~4.8 (Fig. 3b, c). For Glu84 of VDAC2, the estimated pKa value is ~5.1 (Supplementary Fig 3a, b). This indicates that at neutral pH, both Glu73 in VDAC1and Glu84in VDAC2are intheir deprotonated,negativelycharged https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 2 Fig. 1 | Mitochondrial localization of HKI relies on its N-terminal α-helix and a membrane-buried Glu in VDACs. a Fluorescence images of wild-type (WT) and VDAC1/2-DKO HeLa cells expressing EGFP-tagged HKI (green) alone or in combination with HA-tagged VDAC1, VDAC1E73Q, VDAC1E73D, VDAC2, VDAC2E84Q orVDAC2E84D,fixedandthenstainedwithDAPI (blue) andan antibody against Tom20 (magenta). Line scans showing degree of overlap between HKI and Tom20 signals along the path of the arrow shown in the zoom-in. Scale bar, 10 μm. bPearson’s correlation co-efficient analysis between HKI and Tom20 signals in cells as in (a). For each violin plot, the middle line denotes the median, and the top and bottom lines indicate the 75th and 25th percentile. From left to right, n= 20 (WT), 20 (VDAC1/2-DKO), 20 (VDAC1/2-DKO +VDAC1), 20 (VDAC1/2DKO +VDAC1E73Q), 20 (VDAC1/2-DKO +VDAC2), 20 (VDAC1/2DKO +VDAC2E84Q), 46 (WT), 20 (VDAC1/2-DKO), 20 (VDAC1/2-DKO + VDAC1), 23 (VDAC1/2-DKO +VDAC1E73D), 20 (VDAC1/2-DKO +VDAC2) and 20 cells (VDAC1/2-DKO +VDAC2E84D) over at least 2 independent experiments. pvalues were calculated by unpaired two-tailed ttest. cFluorescence images of WT HeLa cells expressing EGFP-tagged HKI or N-terminal truncation mutant HKI Δ2-14 ,fixed and then stained with DAPI (blue) and an antibody against Tom20 (magenta). Line scans showing degree of overlap between HKI and Tom20 signals along the path of the arrow shown in the zoom-in. Scale bar, 10 μm. dFluorescence images of live WT and VDAC1/2-DKO HeLa cells co-expressing EGFP-tagged Tom20 (magenta) and Halo-tagged HKI-N (N-terminal HKI residues 1-17, green). Line scans showing degree of overlap between HKI-N and Tom20 signals along the path of the arrow shown in the zoom-in. Scale bar, 10 μm. eModels of complexes formed between HKI, Halo-tagged HKI-N and VDAC1/2. The membrane-buried Glu is marked in red. https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 3 state even when residing in the hydrophobic membrane interior. Consequently, we performed CG-MD simulations of HKI-N binding to VDAC1 andVDAC2 withthebilayer-facingGluin thedeprotonated(charged)state. CG-MD-simulations of channels with protonated (neutral) Glu residues served as control to verify the importance of having Glu in its negatively charged form for HKI-N binding. As the membrane topology of VDACs is not known, each channel was also simulated in two orientations, namely with its C-terminus facing the cytosol—where HKI-N was present–or the inter-membrane space (IMS). Main simulations were performed in an OMM-mimicking bilayer with an aggregate time of 1.41 ms (Supplementary Table 1)—only attainable using CG-MD. Strikingly, HKI-N formed stable contacts with both VDAC1 and VDAC2 provided that the channel’sC-terminus faced the IMS and the membrane-buried Glu was deprotonated (Fig. 3d–f; Supplementary Fig. 3c, d). When these conditions were met, the N-terminal half of HKI-N was observed to insert vertically into the cytosolic membrane leaflet along one side of the channel wall and bind directly to Glu73–in VDAC1 and Glu84–in VDAC2 (Supplementary Movies 1 and 2). HKI-N residues most frequently in direct contact with the deprotonated Glu were Met1, Ala4 and Gln5, all situated on the same side along the axis of the α-helix (Fig. 3g; Supplementary Fig. 3f), with Met1-Glu73/Glu84 contacts occurring for 17.20 ± 3.11% and 15.68 ± 5.83% of the aggregate simulation time for VDAC1 and VDAC2, respectively (errors indicate SEMs over 3 replicates). Thesebindingeventswereoftenobservedmultipletimespersimulationand reachedμs durations(Fig.3e;SupplementaryFig.3d; SupplementaryFig. 4). Protonation of the membrane-buried Glu severely reduced the contact prevalence to 0.31 ± 0.25% and 2.14 ± 0.74% for VDAC1 and VDAC2, respectively. Flipping the membrane orientation of the channel in each case completely abolished complex formation. Under these conditions, HKI-N failed to insert into the cytosolic leaflet and no interaction with the bilayerfacing Glu occurred. Instead, contacts with VDAC1 and VDAC2 became random and short-lived (<10 ns), involving channel residues facing the cytosol (Fig. 3f, g; Supplementary Fig. 3e, f). Consistent with the localization studies of GFP-tagged HKI in HeLa cells (Fig. 1a, b), VDAC channels with a Glu-to-Gln substitution lacked affinityforHKI-Nin simulations,regardless oftheirtransbilayerorientation (0.06 ± 0.05% and 1.67 ± 0.31% contact prevalence for VDAC1 and VDAC2, respectively). On the other hand, VDAC channels with a Glu-toAsp substitution retained the ability to bind HKI-N, provided that the Asp was deprotonated and the channel’sC-terminus faced the IMS (17.39 ± 4.50% and 36.30 ± 13.26% contact prevalence for VDAC1 and VDAC2, respectively; Supplementary Fig. 5). Collectively, these results indicate that HKI-VDAC binding critically relies on both the membrane topology of VDACs and the protonation state of the bilayer-facing Glu. Acidification of cytosolic pH triggers dissociation of HKI-N from mitochondria To challenge the idea that HKI-VDAC complex formation is controlled by the protonation state of the bilayer-facing Glu, we next investigated the impact of cytosolic acidification on the subcellular distribution of Halotagged HKI-N in HeLa cells. Cytosolic pH was adjusted by incubating cells inabufferwiththedesiredpHinthepresenceofH+/K+ionophore nigericin p < 0.0001pp < 0.0001 Fig. 2 | HKI-N binding to membranes. a Atomic model of HKI (PDB: 1BG3, cyan) with the N-terminal α-helix (HKI-N) highlighted in residue-type coloring. bHeliQuest analysis and VDW/Dynamic Bonds representation of the coarsegrained HKI-N backbone reveals an α-helix with a polar and apolar face. cModel predicting that the apolar face of HKI-N mediates membrane binding, with the first half of the α-helix protruding deeper into the membrane bilayer. dMembrane residence time analysis of HKI-N and HKI-NL7Q using CG-MD simulations. Helices were bound to OMM-mimicking membranes following the restraining protocol described in Methods. After lifting the restraints, the distance of the helix residue closest to the membrane’s top leaflet was measured, until it surpassed 1.4 nm. Stills show representative configurations from each condition. Plots represent the time progression of the helix-membrane distances for six independent replicas per condition, as overlaid semitransparent traces; vertical rises correspond to each trace’s membrane-leaving event, from which point that trace is no longer drawn. eFluorescence images of WT HeLa cells expressing EGFP-tagged HKI or HKIL7Q (green), fixed and then stained with DAPI (blue) and an antibody against Tom20 (magenta). Line scans showing degree of overlap between HKI and Tom20 signals along the path of the arrow shown in the zoom-in. Scale bar, 10 μm. fPearson’s correlation co-efficient analysis between HKI and Tom20 signals in cells as in (e). n= 20 (HKI) and 28 cells (HKIL7Q) over three independent experiments. pvalues were calculated by unpaired two-tailed ttest. https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 4 (Fig. 4a; Supplementary Fig. 6a). Equilibration of cytosolic pH with the pH of the external buffer was quantitatively assessed with the intracellular pH indicator pHrodoTM RedAM(SupplementaryFig.6b).Tomonitoradrop in cytosolic pH in real time, we took advantage of the fact that the fluorophore of EGFP is more sensitive to acidic pH when compared to mCherry44 and HaloTag Ligand JF646. Thus, in cells expressing Tom20-EGFP, JF646Fig. 3 | HKI-N binding to VDAC1 is directly controlled by the protonation state of the membrane-buried Glu. a Atomic model of HKI (cyan, with residue-type colored HKI-N) bound to VDAC1 (yellow) with the membrane-buried Glu (E73) marked in red.bStill from a titratable MD simulation of VDAC1 (yellow) to evaluate the protonation state of E73 (red) at pH 5.0. PO 4 groups in the POPC-based bilayer are marked in orange and protons are marked in blue.cTitration curve showing the degree of deprotonation of E73 in VDAC1, simulated at a pH range of 3–8. dStills from an MD simulation showing HKI-N bound to VDAC1 with a deprotonated E73 (red) and IMS-facing C-terminus. Glycerol groups in the OMM-mimicking bilayer are marked in cyan.eStills from an MD simulation, showing the approach and binding of HKI-N to VDAC1 with a deprotonated E73 (red) and IMS-facing Cterminus. fRelative duration of contacts between HKI-Met1 and specific residues of VDAC1 with a protonated or deprotonated E73 and cytosolor IMS-facing Cterminus simulated in an OMM-mimicking bilayer. Shown are the combined data of three individual replicas with a total simulation time between 169 μs and 211 μs per condition.gRelativeduration ofcontactsbetweenVDAC1-E73 andspecificresidues of HKI-N under the same conditions as in (f). https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 5 labeled Tom20-Halo and mCherry fused to the OMM anchor of AKAP1 (OMM-mCherry), a shift in cytosolic pH from 7.4 to 6.0 strongly reduced EGFP fluorescence without affecting the other two fluorophores (Fig. 4b, c; Supplementary Fig. 6c, d). Strikingly, acidification of the cytosol readily triggered the translocation of JF646-labeled HKI-N-Halo from mitochondria into the cytosol. Dissociation of HKI-N-Halo from mitochondria was already measurable when lowering the cytosolic pH to 6.8 and gradually progressed with increased acidification so that at pH 6.0 the bulk of HKI-N-Halo resided in the cytosol (Fig. 4b, d; Supplementary Fig. 6c, d). Raising the cytosolic pH from 6.0 back to 7.4 restored the mitochondrial https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 6 localization of HKI-N-Halo (Fig. 4b). Consistent with the CG-MD simulations, these results support the notion that HKI-VDAC binding is controlled by the protonation state of the bilayer-facing Glu even though we cannot exclude that protonation of additional acidic residues also play a role. HKI-VDAC binding critically relies on an asymmetric positioning of the membrane-buried Glu The foregoing CG-MD simulations revealed that the transbilayer orientation of VDACs is a critical determinant of HKI binding (Fig. 3f, g; Supplementary Fig. 3e, f). Interestingly, we noticed that the membrane-buried Glu in VDACs is asymmetrically positioned a few Å away from the bilayer center and resides in the cytosolic leaflet when the channel’sC-terminus faces the IMS, the orientation compatible with HKI binding (Fig. 5a). We therefore hypothesized that channels with the opposite topology may fail to bind HKI because the membrane-buried Glu in that orientation lies too deep in the lipid bilayer for the enzyme’sN-terminal α-helix to make stable contacts. To verify this idea, we substituted Phe for Glu73 and Glu for Phe71 in VDAC1, effectively creating a channel in which the asymmetric position of the membrane-buried Glu is flipped across the bilayer center (Fig. 5a). Next, we performed CG-MD simulations to probe HKI-N binding to the VDAC1E73F/F71E mutant channel in both membrane orientations and with a deprotonatedGlu.UnlikeVDAC1, theVDAC1E73F/F71E variantwasunable to form stable contacts with HKI-N irrespective of its transbilayer orientation (Fig. 5b). Moreover, unlike VDAC1, the VDAC1E73F/F71E variant completely failed to restore mitochondrial localization of GFP-HKI in VDAC1/2double KO cells (Fig. 5c). These results indicate that bilayer depth of the charged Glu on the outer channel wall, although critical, is not the sole Fig. 4 | Cytosolic pH controls mitochondrial association of HKI-N. a Schematic outline of experimental strategy to determine the impact of cytosolic acidification on mitochondrial association of HKI-N. bFluorescence images of live HeLa cells coexpressing EGFP-tagged Tom20 (green), OMM-mCherry (cyan)andHalo-taggedHKIN(magenta) grown in Optimem (top), treated with 10 μM nigericin in pH 6.0 buffer for 5 min (middle) and then with 10 μM nigericin in pH 7.4 buffer for 5 min (bottom). Line scans showing degree of overlap between OMM and HKI-N signals along the path of the arrow shown in the zoom-in. Scale bar, 10 μm. cFluorescence images of live HeLa cells co-expressing EGFP-tagged Tom20 (green), OMM-anchored mCherry (cyan)andHalotagged Tom20 (magenta) treated as in (b). Line scans showing degree of overlap between OMM and Tom20-Halo signals along the path of the arrow shown in the zoom-in. Scale bar, 10 μm. dQuantitative assessment of mitochondria-associated levels of OMMmCherry, Tom20-EGFP, Tom20-Halo and HKI-N-Halo in live HeLa cells after treatment with nigericin in buffer at indicated pH for 5 min. Fluorescence values in corresponding pH buffer were set relative to values of same cell in Opti-MEM. Data are means ± SD, n= 6 cells per condition over four independent experiments. pvalues were calculated by unpaired two-tailed ttest. Fig. 5 | HKI-VDAC binding critically relies on an asymmetric positioning of the membrane-buried Glu. a Stills from MD simulations of VDAC1 and VDAC1E73F/F71E withthe membrane-facing Glu and Phe residues at positions 71 and 73 represented as red and white balls, respectively. The graphs show the position of Glu73 in VDAC1 and Glu71 in VDAC1E73F/F71E relative to the membrane center (dashed line) over the course ofa simulation. bRelative durationof contactsbetweenHKI-Met1andspecific residues of VDAC1or VDAC1E73F/F71E withcytosol-orIMS-facing C-termini. Data for VDAC1 are taken from Fig. 3f and shown for comparison. For VDAC1E73F/F71E data of three individual simulationswere combined with a total simulation time between 148 μs and 162 μs per condition. cFluorescence images of VDAC1/2-DKO HeLa cells coexpressing EGFP-tagged HKI (green) and HA-tagged VDAC1 or VDAC1E73F/F71E, fixed and then stained with DAPI (blue) and antibodies against the HA-epitope (red) and Tom20 (magenta). Line scans showing degree of overlap between HKI and Tom20 signals along the path of the arrow shown in the zoom-in. Scale bar, 10 μm. https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 7 determinant of HKI binding and that other unique features on the membrane-facing surface of VDACs also play a role. VDAC channels cause thinning of the lipid monolayer proximal to the membrane-buried Glu VDACs experience a global hydrophobic mismatch with the lipid bilayer in whichtheyareinsertedin—havinga hydrophobicinterfaceofonly~2.4 nm, which is significantly less than that of biological membranes (~4 nm)45. Cumulatively with this overall mismatch, previous MD simulations of VDAC1 revealed additional membrane thinning and water defects near the outward-facing Glu5,46. By extending these studies to VDAC1 in its HKI binding-competent orientation (with the channel’sC-terminus facing the IMS), we found that this localized membrane thinning is mainly confined to the cytosolic leaflet, adjacent to the negatively charged Glu (E73–;Fig.6a). Here, the cytosolic leaflet reached a minimal thickness of 0.71 ± 0.04 nm, which was considerably thinner than the average thicknessnear the channel wall outside of this region (1.49 ± 0.01 nm) or in the absence of protein (1.94 ± 0.001 nm; Supplementary Fig. 7a, b). In the defect region, we also Fig. 6 | VDAC channels cause lipid packing defects and membrane leaflet thinning proximal to the bilayer-facing Glu. (a) Leaflet-specific membrane thinning graphs of VDAC1 and VDAC2 simulated in a POPC bilayer with C-termini facing the IMS leaflet. Gray spheres indicate the VDAC backbone and the position of the bilayer facing Glu is marked by an arrow. Membrane thinning was calculated as the average distance of the lipid backbone phosphates to the global membrane center. (b) Leaflet-specific water defect graphs of VDAC1 and VDAC2 simulated as in (a). Water defects were calculated as the amount of water molecules detected within a z-distance of 1.5 nm to the global membrane center. (c) Occupancies of lipid PO 4 groups in simulations of VDAC1 and VDAC2 as in (a). Occupancy surfaces enclose volumes with average occupancy of 0.5% or greater. The position of the bilayerfacing Glu is marked. (d) Cytosolic leaflet thinning and water defect graphs of VDAC1, VDAC1E73Q and VDAC1E73D simulated in a POPC bilayer with C-termini facing the IMS (bottom) leaflet. The bilayer-facing acidic residues were protonated or deprotonated, as indicated. Analysis was done as in (a) and (b). https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 8 observed a large degree of water penetration (Fig. 6b). Simulations of VDAC2 revealed a similar thinning of the cytosolic leaflet along with water defectsnearthechargedGlu(E73 –;Fig.6a, b), with a minimal thickness of 0.52 ± 0.05 nm and an average thickness of 1.46 ± 0.01 nm near the channel wall outside of the defect region (Supplementary Fig. 7a, b). When mapping the occupancy of the lipid phosphates or plotting the thickness of VDACsurrounding lipids, we observed that the region of membrane thinning did not perfectly overlap with the position of the charged Glu (Fig. 6c; Supplementary Fig. 7b). This suggested that membrane thinning may not rely on a charged Glu but rather on outward-facing polar residues in its vicinity. Indeed, protonation of the bilayer-facing Glu or its substitution by Gln in VDAC1 or VDAC2 greatly diminished the water defects in either case, but had little impact on the membrane thinning capacity of the channels (Fig. 6d; Supplementary Fig. 8). When the bilayer-facing Glu was replaced by a deprotonated Asp (D73–), membrane thinning and water defects were retained (Fig. 6d; Supplementary Fig. 8). Hence, while the negatively charged Glu creates conditions that facilitate the penetration of water, it appears that the membrane thinning capacity of VDACs is mediated by other residues on the outer channel wall. Polar residues proximal to the membrane-buried Glu provide a gateway for HKI-VDAC binding We considered that thinning of the cytosolic leaflet near the membraneburied Glu of VDACs may provide a low-energy passageway for the Nterminalα-helix of HKIto facilitate HKI-VDAC binding. A close inspection of the outer wall of VDAC1 in areas exhibiting the highest degree of membrane thinning revealed two polar residues, Thr77 and Ser101, which are positioned within close range of the membrane-buried Glu (Fig. 7a). In CG-MD simulations, substitution of Leu for Thr77 or Ser101 in each case led to a localized but marked reduction in the membrane thinning capacity of VDAC1 carrying a charged Glu (Fig. 7a, b; Supplementary Fig. 7c). The areas occupied by Thr77 and Ser101 each had its own local minimum of leafletthicknessthatwasselectively abolishedby mutation.Thethicknesson the S101 side displayed the lowest local minimum and corresponds to the global minimum (0.71 ± 0.04 nm). For that reason, the T77L mutation did not affect the global minimum (0.65 ± 0.01 nm). Upon introducing the S101L mutation, the minimum on the T77 side becomes the new global minimum with a higher thickness value (0.90 ± 0.01 nm). When the two mutations were combined, leaflet thinning in the region proximal to the charged Glu was further reduced (1.11 ± 0.01 nm), essentiallyabolishing the leaflet thinning specifictothisregion(Fig.7a, b; Supplementary Fig. 7c). This was accompanied by a substantial reduction in water defects (Fig. 7b). These results indicate that Thr77 and Ser101 each contribute to a local distortion of the cytosolic membrane leaflet, possibly facilitating access of HKI to the charged, membrane-buried Glu. Consistent with this idea, CGMD simulations revealed that substitution of Leu for Thr77 or Ser101 in VDAC1 diminished contacts between HKI-N and the charged Glu (from 17.20 ± 3.11% to 10.77 ± 4.75% and 11.11 ± 4.08% contact prevalence for T77L and S101L, respectively; Fig. 7c, d). Combining these substitutions further reduced HKI-N binding to a 4.57 ± 2.20% contact prevalence. Looking at contact lifetime distributions, it can be seen that these mutations affect binding by reducing the on-rate rather than off-rate of the binding process (Supplementary Fig. 4), indicating that the polar face surrounding the Glu acts indeed as an access pathway. Importantly, the diminished capacity of the mutant channels to bind HKI-N in silico strongly correlated with an impaired ability of these channels to restore mitochondrial recruitment of HKI in VDAC1/2-double KO cells (Fig. 7e, f). Collectively, these results indicate that Thr77 and Ser101 are core components of a membrane thinning pathway by which the N-terminal α-helix of HKI gains access tothe membrane-buriedGluof VDACs,therebyprovidinga gateway for HKI-VDAC binding. Discussion While binding of HKI to mitochondrial VDACs is crucial for cell growth and survival, the structural basis of HKI-VDAC complex assembly is not known. Using a CG-MD simulations approach complemented with functional studies in cells, we identified core structural and physicochemical features that govern binding of HKI to VDAC1 and VDAC2. As schematically outlined in Fig. 8, our results indicate that a bilayer-facing negatively charged Glu on the outer channel wall plays a crucial role in HKI binding by promoting stable contacts between the channel and the enzyme’samphipathicN-terminal αhelix (HKI-N). Protonation of the Glu residue abolishes HKI-N binding in simulations while transient acidification of the cytosol causes a reversable release of HKI-N from mitochondria. Membrane insertion of HKI occurs adjacent to the charged Glu where a pair of polar channel residues causes a marked thinning of the cytosolic membrane leaflet, creating a funnel that likely serves as low-energy passageway for the enzyme’sN-terminal α-helix to facilitate complex assembly. Consistent with this model, we found that disruptingthe membranethinningcapacity ofVDAC1 significantlyimpaired its ability to bind HKI both in silico and in cells. In line with previous work47, we demonstrate that HKI-N is essential and sufficient for VDAC binding. However, HKI-N can also bind membranes independently of VDACs36. Breaking the apolar face of HKI-N by a single point mutation significantly weakened membrane binding in silico and abolished mitochondrial localization of HKI in VDAC1/2-expressing cells. From this we infer that membrane partitioning of HKI-N is a prerequisite for VDAC binding. Our findings are hard to reconcile with a previous model of HKI-VDAC complex formation that is based on direct plugging of HKI-N into the channel’scentralpore 27. Instead, our data indicate that HKI-VDAC complex assembly is a multistep process whereby HKI initially binds the OMM through membrane adsorption involving the apolar interface of HKI-N. We envision that thinning of the cytosolic membrane leaflet by a pair of polar channel residues, Thr77 and Ser101 in VDAC1,createsafunnelthatservesasthermodynamictrapforHKI binding by enabling the enzyme’sN-terminal α-helix to tilt and insert at the VDAC/membrane interface to become aligned for stable interactions with the charged Glu on the outer channel wall. Additionally, our data provide important clues regarding the transbilayer orientationofVDACchannelsintheOMM.Thesidednessoftheseβ-barrel proteins has been probed with various approaches without reaching general consensus. For instance, studies onhuman VDAC1 carrying a cleavage site for cytosolic caspases indicate that the channel’sC-terminus faces the IMS34.In contrast, a split-NeonGreen complementation study suggests that the C-terminus of human VDAC2 faces the cytosol35. Based on packing analysis of murine VDAC1 crystals in a lipidic environment, Ujwal et al48.proposedthat VDACs are dual topology membrane proteins that may achieve anti-parallel arrangements in the OMM. However, our MD simulations clearly indicate that HKI-VDAC complex formation is only possible with channels in one orientation, namely whereby their C-termini face the IMS. It is only in this orientation that the polar channel residues critical for membrane-thinning are positioned accurately to establish a passageway for cytosolic HKI to reach the bilayer-facing Glu and form a stable complex. While our findings do not rule out the possibility of a dual topology of VDAC channels, they clearly indicate thatonlyoneofthetwopossibletransbilayerorientationsprovidesafunctional binding platform for HKI. Titratable MD simulations of VDAC1 and VDAC2 revealed that at neutral pH, the bilayer-facing Glu is predominantly in its deprotonated, fully negatively-charged state. Although it is energetically unfavorable for a charged residue to be exposed to the hydrophobic membrane interior, membrane thinning imposed by polar residues in close proximity of the bilayer-facing Glu may explain why its pKa value is shifted by less than one unit in comparison to a free Glu. Converging lines of evidence indicate that the protonation status of the bilayer-facing Glu is a key determinant of HKI binding. To begin with, protonation of this Glu in VDAC1 and VDAC2 in each case proved sufficient to abrogate HKI-N binding in simulations. Replacing Glu with the non-titratable Gln abolished HKI-N binding to VDAC channels in simulations and disrupted VDAC-dependent mitochondrial localization of HKI in cells. Conversely, replacing Glu for titratable Asp promoted complex formation both in silico and in cells. Mild acidification of the cytosol from pH 7.4 to 6.0 instantly dissociated HKI-N https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 9 37. Gautier, R., Douguet, D., Antonny, B. & Drin, G. HELIQUEST: A web server to screen sequences with specificα-helical properties. Bioinformatics 24, 2101–2102 (2008). 38. Souza, P. C. T. et al. Martini 3: a general purpose force field for coarse-grained molecular dynamics. Nat. Methods 18, 382–388 (2021). 39. Marrink, S. J. et al. Two decades of Martini: Better beads, broader scope. Wiley Interdiscip. Rev.: Comput. Mol. Sci. 13, e1620 (2023). 40. Horvath, S. E. & Daum, G. Lipids of mitochondria. Prog. Lipid Res. 52, 590–614 (2013). 41. Melo, M. N. Coarse-grain simulations of membrane-adsorbed helical peptides. Methods Mol. Biol. 2405, 137–150 (2022). 42. Chiariello, M. G., Grünewald, F., Zarmiento-Garcia, R. & Marrink, S. J. pH-dependent conformational switch impacts stability of the PsbS dimer. J. Phys. Chem. Lett. 14, 905–911 (2023). 43. Isom, D. G., Castañeda, C. A., Cannon, B. R., Velu, P. D. & GarcíaMoreno, E. B. Charges in the hydrophobic interior of proteins. Proc. Natl Acad. Sci. Usa. 107, 16096–16100 (2010). 44. Doherty, G. P., Bailey, K. & Lewis, P. J. Stage-specificfluorescence intensity of GFP and mCherry during sporulation in Bacillus Subtilis. BMC Res. Notes 3, 303 (2010). 45. Lafargue, E. et al. Lipid composition of the membrane governs the oligomeric organization of VDAC1. https://doi.org/10.1101/2024.06. 26.597124 (2024). 46. Villinger, S. et al. Functional dynamics in the voltage-dependent anion channel. Proc. Natl Acad. Sci. Usa. 107, 22546–22551 (2010). 47. Rossi, A. et al. Defective mitochondrial pyruvate flux affects cell bioenergetics in Alzheimer’s disease-related models. Cell Rep. 30, 2332–2348 (2020). 48. Ujwal, R., Cascio, D., Chaptal, V., Ping, P. & Abramson, J. Crystal packing analysis of murine VDAC1 crystals in a lipidic environment reveals novel insights on oligomerization and orientation. Channels (Austin). 3, 167–170 (2009). 49. Quach, C. H. T. et al. Mild alkalization acutely triggers the Warburg effect by enhancing hexokinase activity via voltage-dependent anion channel binding. PLoS ONE 11, e0159529 (2016). 50. Bergdoll, L. A. et al. Protonation state of glutamate 73 regulates the formation of a specific dimeric association of mVDAC1. Proc. Natl Acad. Sci. Usa. 115, E172–E179 (2017). 51. Sun, L., Shukair, S., Naik, T. J., Moazed, F. & Ardehali, H. Glucose phosphorylation and mitochondrial binding are required for the protective effects of hexokinases I and II. Mol. Cell. Biol. 28, 1007–1017 (2008). 52. Lauterwasser, J. et al. Hexokinases inhibit death receptor-dependent apoptosis on the mitochondria. Proc. Natl Acad. Sci. USA 118, e2021175118 (2021). 53. Corbet, C. & Feron, O. Tumour acidosis: from the passenger to the driver’s seat. Nat. Rev. Cancer 17, 577–593 (2017). 54. Reshkin, S. J. et al. Na+/H+exchanger‐dependent intracellular alkalinization is an early event in malignant transformation and plays an essential role in the development of subsequent transformation‐ associated phenotypes. FASEB J. 14, 2185–2197 (2000). 55. Grillo-Hill, B. K., Choi, C., Jimenez-Vidal, M. & Barber, D. L. Increased H+efflux is sufficient to induce dysplasia and necessary for viability with oncogene expression. Elife 4, e03270 (2015). 56. Dadsena, S., Hassan, D. G. & Holthuis, J. C. M. Unraveling the molecular principles by which ceramides commit cells to death. Cell Stress 3, 280–283 (2019). 57. Patwardhan, G. A., Beverly, L. J. & Siskind, L. J. Sphingolipids and mitochondrial apoptosis. J. Bioenerg. Biomembr. 48,153–168 (2016). 58. Ogretmen, B. Sphingolipid metabolism in cancer signalling and therapy. Nat. Rev. Cancer 18,33–50 (2018). 59. Galluzzi, L., Kepp, O., Tajeddine, N. & Kroemer, G. Disruption of the hexokinase-VDAC complex for tumor therapy. Oncogene 27, 4633–4635 (2008). 60. Smilansky, A. et al. The voltage-dependent anion channel 1 mediates amyloid βtoxicity and represents a potential target for Alzheimer disease therapy. J. Biol. Chem. 290, 30670–30683 (2015). 61. Magri, A. & Messina, A. Interactions of VDAC with proteins involved in neurodegenerative aggregation: an opportunity for advancement on therapeutic molecules. Curr. Med. Chem. 24, 4470–4487 (2017). 62. Shoshan-Barmatz, V., Zakar, M., Rosenthal, K. & Abu-Hamad, S. Key regions of VDAC1 functioning in apoptosis induction and regulation by hexokinase. Biochim. Biophys. Acta—Bioenerg. 1787, 421–430 (2009). 63. Jain, A., Beutel, O., Ebell, K., Korneev, S. & Holthuis, J. C. M. Diverting CERT-mediated ceramide transport to mitochondria triggers Baxdependent apoptosis. J. Cell Sci. 130, 360–371 (2017). 64. Jain, A., Dadsena, S. & Holthuis, J. C. M. A switchable ceramide transfer protein for dissecting the mechanism of ceramide-induced mitochondrial apoptosis. FEBS Lett. 594, 3739–3750 (2020). 65. Ghaffari Zaki, A. et al. Development of a chemogenetic approach to manipulate intracellular pH. J. Am. Chem. Soc. 145, 11899–11902 (2023). 66. Arganda-Carreras, I. et al. Trainable Weka segmentation: a machine learning tool for microscopy pixel classification. Bioinformatics 33, 2424–2426 (2017). 67. Mulichak, A. M., Wilson, J. E., Padmanabhan, K. & Michael Garavito, R. The structure of mammalian hexokinase-1. Nat. Struct. Biol. 5, 555–560 (1998). 68. Böhm, R. et al. The structural basis for low conductance in the membrane protein VDAC upon β-NADH binding and voltage gating. Structure 28, 206–214 (2020). 69. Schredelseker, J. et al. High resolution structure and double electronelectron resonance of the zebrafish voltage-dependent anion channel 2 reveal an oligomeric population. J. Biol. Chem. 289, 12566–12577 (2014). 70. Aleshin, A. E. et al. The mechanism of regulation of hexokinase: New insights from the crystal structure of recombinant human brain hexokinase complexed with glucose and glucose-6-phosphate. Structure 6,39–50 (1998). 71. Claveras Cabezudo, A., Athanasiou, C., Tsengenes, A. & Wade, R. C. Scaling protein–water interactions in the martini 3 coarse-grained force field to simulate transmembrane helix dimers in different lipid environments. J. Chem. Theory Comput. 19, 2109–2119 (2023). 72. Spinti, J. K., Neiva Nunes, F. & Melo, M. N. Room for improvement in the initial martini 3 parameterization of peptide interactions. Chem. Phys. Lett. 819, 140436 (2023). 73. Kroon, P. C. et al. Martinize2 and vermouth: unified framework for topology generation. Elife 12, RP90627 (2023). 74. Wassenaar, T. A., Ingólfsson, H. I., Böckmann, R. A., Tieleman, D. P. & Marrink, S. J. Computational lipidomics with insane: a versatile tool for generating custom membranes for molecular simulations. J. Chem. Theory Comput. 11, 2144–2155 (2015). 75. Pedersen, K. B. et al. The Martini 3 lipidome: expanded and refined parameters improve lipid phase behavior. https://doi.org/10.26434/ chemrxiv-2024-8bjrr (2024). 76. Periole, X., Cavalli, M., Marrink, S. J. & Ceruso, M. A. Combining an elastic network with a coarse-grained molecular force field: Structure, dynamics, and intermolecular recognition. J. Chem. Theory Comput. 5, 2531–2543 (2009). 77. Hess, B., Kutzner, C., Van Der Spoel, D. & Lindahl, E. GRGMACS 4: Algorithms for highly efficient, load-balanced, and scalable molecular simulation. J. Chem. Theory Comput. 4, 435–447 (2008). https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 16 78. Su, J., Marrink, S. J. & Melo, M. N. Localization preference of antimicrobial peptides on liquid-disordered membrane domains. Front. Cell Dev. Biol. 8, 350 (2020). 79. Grünewald, F. et al. Titratable Martini model for constant pH simulations. J. Chem. Phys. 153, 024118 (2020). 80. Van Gunsteren, W. F. & Berendsen, H. J. C. A leap-frog algorithm for stochastic dynamics. Mol. Simul. 1, 173–185 (1988). 81. Timme, M. & Melo, M. N. Analysis scripts and computational models used in ‘Hexokinase-I directly binds to a charged membrane-buried glutamate of mitochondrial VDAC1 and VDAC2’[data set]. (2024). https://doi.org/10.5281/zenodo.14144611. Acknowledgements The authors gratefully acknowledge Ladislav Bartos and Robert Vácha (National Centre for Biomolecular Research, Masaryk University, Brno, Czech Republic) for providing the scripts for membrane thinning and water defects analysis, and Varda Shoshan-Barmatz (Ben-Gurion University of the Negev, Israel) for the pEGFP-HKI construct.This work was supported by the Deutsche Forschungsgemeinschaft (378148610 and 448344643 to J.C.M.H.), the German Egyptian Research Long-term Scholarship Program (GERLI project 57222240 to D.G.H.), the European Research Council (ERC Advanced grant 101053661 „COMP-O-CELL“to S.J.M.) and the FCT— Fundação para a Ciência e a Tecnologia I.P. (through MOSTMICRO-ITQB R&D Unit with projects UIDB/04612/2020 and UIDP/04612/2020, and LS4FUTURE Associated Laboratory with projects LA/P/0087/2020 and CEECIND/04124/2017/CP1428/CT0008 to M.N.M.). Author contributions M.N.M. and J.C.M.H. designed the research with critical input from S.B. and M.T.; S.B. performed experiments in cells with critical input from D.H.; M.T. carried out the CG-MD simulations with critical input from N.W.; C.M.B. carried out all titratable MD simulations; J.C.M.H. provided expertise for experiments in cells and helped interpret the data; M.N.M. and S.J.M. provided expertise for CG-MD simulations and helped interpret the data; J.C.M.H. wrote the manuscript; all authors discussed results and commented on the manuscript. Funding Open Access funding enabled and organized by Projekt DEAL. Competing interests The authors declare no competing interests. Additional information Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s42003-025-07551-9. Correspondence and requests for materials should be addressed to Manuel N. Melo or Joost C. M. Holthuis. Peer review information Communications Biology thanks Subhendu Ghosh and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Laura Rodríguez Pérez and Christina Karlsson-Rosenthal. A peer review file is available. Reprints and permissions information is available at http://www.nature.com/reprints Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. © The Author(s) 2025 https://doi.org/10.1038/s42003-025-07551-9 Article Communications Biology | (2025) 8:212 17