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

communica ions biology A icle h ps://doi.o g/10.1038/s42003-025-07551-9 Hexokinase-I di ec ly binds o a cha ged memb ane-bu ied glu ama e o mi ochond ial VDAC1 and VDAC2 Check o upda es Sebas ian Bieke 1,2,6, Michael Timme1,2,6, Nils Woge1,2,DinaG.Hassan 1,2,3, Chelsea M. B own 4, Siewe J. Ma ink 4,ManuelN.Melo 5& Joos C. M. Hol huis 1,2 Binding o hexokinase HKI o mi ochond ial ol age-dependen anion channels (VDACs) has a - eaching physiological implica ions. Howe e , he s uc u al basis o his in e ac ion is unclea . Combining compu e simula ions wi h expe imen s in cells, we he e show ha complex assembly elies on in ima e con ac s be ween he N- e minal α-helix o HKI and a cha ged memb ane-bu ied glu ama e on he ou e wall o VDAC1 and VDAC2. P o ona ion o his esidue blocks complex o ma ion in silico while acidifica ion o he cy osol causes a e e sable elease o HKI om mi ochond ia. Memb ane inse ion o HKI occu s adjacen o he bilaye - acing glu ama e whe e a pai o pola channel esidues media es a ma ked hinning o he cy osolic leafle . Dis up ing he memb ane hinning capaci y o VDAC1 d ama ically impai s i s abili y o bind HKI in silico and in cells. Ou da a e eal key opological and mechanis ic insigh s in o HKI-VDAC complex assembly ha may benefi he de elopmen o he apeu ics o coun e pa hogenic imbalances in his p ocess. Vol age-dependen anion channels (VDACs) a e abundan β-ba el p o- einsin heou e memb aneo mi ochond ia(OMM) ha se eas hemain condui s o he la ge flux o ions, ATP/ADP, NAD +/NADH and K ebs’ cycle in e media es om and in o mi ochond ia1,2.Inmammals, h ee iso o ms exis (VDAC1-3) wi h non- edundan unc ions3,4.VDAC1and VDAC2 a e he mos abundan ly exp essed iso o ms in mos issues. Besides hei cen al ole in con olling he flow o me aboli es ac oss he OMM, bo h iso o ms ac as sc amblases ha media e phospholipid impo in o mi ochond ia5. Addi ionally, VDAC1 and VDAC2 unc ion as dynamic ansloca ion pla o ms o a a ie y o p o eins ha con ol he pe meabili y o he OMM o cy och ome c o ei he p omo e o p e en mi ochond ial apop osis. VDAC binding pa ne s include he p o- apop o ic Bcl-2 p o eins BAX and BAK6–8, which media e he decisi e s ep in OMM pe meabiliza ion by which cy och ome cand o he apop o- genic ac o s a e eleased in o he cy osol o igge he apop o ic cascade9. Mo eo e , ce amides, cen al in e media es o sphingolipid me abolism, exe hei p o-apop o ic ac i i y, a leas in pa , by in e ac ing di ec ly wi h VDAC210. VDAC1 and VDAC2 also unc ion as he physiological ecep o s o hexokinases (HKs). These enzymes phospho yla e glucose o gene a e glucose-6-phospha e (G-6-P), an ATP-dependen eac ion ha se es as en y poin o glucose in o he glycoly ic pa hway o ene gy p oduc ion o , al e na i ely, in o he pen ose phospha e pa hway o gene a e anabolic in e media es11. Ele a ed le els o mi ochond ially bound HK iso o ms HKI and HKII lead o ahigh a e o glycolysis and lac a e p oduc ion,a me abolic signa u e e e ed o as he Wa bu g e ec 12. This me abolic swi ch om oxida i e o glycoly ic me abolism is a cen al hallma k o umo p og es- sion, allowing p e-malignan lesions o main ain a high me abolic a e in oxygen-dep i ed a ascula en i onmen s13–15. Mo eo e , mi ochond ially bound HKs p o ec cance cells om d ug-induced mi ochond ial apop- osis by diminishing he p opensi y o VDACs o in e ac wi h p o- apop o ic Bcl-2 p o eins BAX and BAK16–18. Con e sely, a educ ion in HKI concen a ionin hespinalco dis hough oenhance binding o VDAC1 o specific amyo h ophic la e al scle osis ype I-associa ed a ian s o supe - oxide dismu ase 1 (SOD1), he eby p omo ing o ma ion o oxic SOD1 agg ega es, mi ochond ial dys unc ion and cell dea h in mo o neu ons19,20. The impo ance o HKI-VDAC in e ac ions in ca cinogenesis and neu odegene a i e disease has p omp ed a sea ch o small molecules and pep ides capable o dis up ing o s abilizing his p o ein-p o ein complex21–23. Howe e , hese e o s a e hampe ed by a lack o s uc u al 1Molecula Cell Biology Di ision, Depa men o Biology/Chemis y, Uni e si y o Osnab ück, 49076 Osnab ück, Ge many. 2Cen e o Cellula Nanoanaly ics, Osnab ück Uni e si y, A ille ies aße 77, 49076 Osnab ück, Ge many. 3Depa men o En i onmen al Medical Sciences, Facul y o G adua e S udies and En i onmen al Resea ch, Ain Shams Uni e si y, Cai o, Egyp . 4G oningen Biomolecula Sciences and Bio echnology Ins i u e, Uni e si y o G oningen, Nijenbo gh 7, 9747 AG G oningen, The Ne he lands. 5Ins i u o de Tecnologia Química e Biológica An ónio Xa ie , Uni e sidade No a de Lisboa, A . da República, 2780-157 Oei as, Po ugal. 6 These au ho s con ibu ed equally: Sebas ian Bieke , Michael Timme. e-mail: m.n.melo@i qb.unl.p ;[email p o ec ed] Communica ions Biology | (2025) 8:212 1 1234567890():,; 1234567890():,; insigh s in o how HKI and VDAC assemble in o a complex. Like HKII, HKI con ains a sho N- e minal, 20-amino acid hyd ophobic α-helix ha enables OMM binding, p esumably h ough i s in e ac ion wi h VDACs24–26. Two p o ein-p o ein docking s udies epo ed models o complex o ma ion based on a di ec plugging o he N- e minal helices o HKI/HKII in o he po e o VDAC127,28.Asignifican sho coming o hese models is ha hey ail o add ess a c i ical ole o a memb ane-bu ied glu ama e a posi ion 73 (E73) loca ed on he ou side wall o VDAC1 in HKI binding29,30. Mo eo e , a molecula docking simula ion s udy e ealed a high-a fini y binding si e o a pep ide mimicking he N- e minuso HKI on he ou side wall o VDAC1 in close p oximi y o he bilaye - acing E73 esidue31. Ano he modeling s udy pos ula ed ha HKII ini ially binds he OMM h ough inse ion o i s hyd ophobic N- e minus in o he cy osolic leafle and hen in e ac s wi h he ou e wall o VDAC1 o o m a bina y complex32. Whe he he in e ac ion o HKI wi h VDACs ollows a simila scena io emains o be es ablished. A p esen , he memb ane opology o sidedness o VDAC channels has no been defini i ely assigned, wi h complemen a y expe imen al app oaches yielding di e gen and con a- dic ing esul s33–35. Knowledge o he ac ual opog aphy o VDACs is a p e equisi e o any comp ehensi e analysis o hei ole as mi ochond ial sca olds o a b oad a ie y o p o eins. He e, we combined molecula dynamics simula ions wi h expe i- men al s udies in cells o define he s uc u al and opological de e minan s ha go e n HKI binding o VDAC1 and VDAC2. We find ha complex assembly c i ically elies on di ec in e ac ions be ween he N- e minal α- helix o HKI and a memb ane-bu ied, dep o ona ed glu ama e on he ou e wall o bo h channel isome s. P o ona ion o his esidue abolished complex assembly in simula ions. Consis en wi h his esul , we show ha VDAC- dependen mi ochond ial ansloca ion o a epo e ca ying he N- e m- inal α-helix o HKI is exquisi ely sensi i e o fluc ua ions in cy osolic pH. Mo eo e , we find ha a pai o pola channel esidues flanking he memb ane-bu ied glu ama e causes a ma ked hinning o he cy osolic leafle , p o iding a low-ene gy passageway o HKI o acili a e complex assembly. Taken oge he , ou da a o e undamen al mechanis ic insigh s in o HKI-VDAC complex o ma ion and indica e ha he C- e mini o VDAC channels mus ace he in e memb ane space o p o ide unc ional binding pla o ms o HKI. Resul s A memb ane-bu ied Glu in VDACs is c i ical o s abilizing he mi ochond ial pool o HKI The bulk o HKI no mally esides on mi ochond ia, wi h VDACs se ing as essen ial binding pla o ms. While VDAC1 is widely iewed as p incipal HKI docking si e, he ole o VDAC2 is less well defined. As expec ed, GFP- agged HKI exp essed in HeLa cells ex ensi ely co-localized wi h he OMM ma ke Tom20 (Fig. 1a; Supplemen a y Fig. 1a). Remo al o ei he VDAC1 o VDAC2didno significan ly a ec mi ochond ial localiza ion o HKI- GFP. Howe e , loss o bo h channels abolished mi ochond ial localiza ion o he enzyme and caused i s accumula ion in he cy osol, e en hough a po ion o he enzyme was ound associa ed wi h he ER and plasma memb ane (Fig. 1a, b; Supplemen a y Fig. 1a, b; Supplemen a y Fig. 2a). Mo eo e , endogenous HKI p o ein le els we e significan ly educed in VDAC1/2 double KO cells while subcellula ac iona ion expe imen s showed ha in hese cells, endogenous HKI p ima ily esides in he cy osol (Supplemen a y Fig. 2c, d). Rein oducing VDAC1 o VDAC2 in o VDAC1/2 double KO cells es o ed bo h mi ochond ial localiza ion and exp ession o HKI (Fig. 1a, b; Supplemen a y Fig. 2e). These da a indica e ha VDAC1 and VDAC2 each con ibu e o s abilizing he mi ochond ial pool o HKI. Bo h VDAC1 and VDAC2 ha bo a uniquely posi ioned glu ama e (Glu)in he ansmemb ane egiono β-s and 4 –Glu73 in VDAC1 and Glu84 in VDAC2 – ha aces he bilaye ’s hyd ophobic co e. P io wo k e ealed ha Glu73 in VDAC1 is equi ed o HKI binding30. Consis en wi h his, subs i u ion o Gln o Glu73 in VDAC1 abolished i s abili y o es o e mi ochond ial localiza ion and exp ession o HKI in VDAC1/2 double KO cells (Fig. 1a, b; Supplemen a y Fig. 2e). Likewise, a VDAC2 mu an in which Gln was subs i u ed o Glu84 ailed o s abilize he mi ochond ial HKI pool. In con as , subs i u ion o Asp o Glu73 in VDAC1 o Glu84 in VDAC2 yielded a channel ha suppo ed mi ochon- d ial ec ui men o HKI o a le el beyond ha obse ed o i s wild ype coun e pa (Fig. 1a, b). Taken oge he , hese esul s sugges ha HKI binding o VDAC1 and VDAC2 c i ically elies on a nega i ely-cha ged, memb ane-bu ied Glu esidue on he ou e channel wall. Mi ochond ial ec ui men o HKI is media ed by i s N- e minal α-helix HKIcon ainsanN- e minalα-helixo 20-aminoacids(HKI-N) ha enables binding o he OMM24, p esumably by in e ac ing di ec ly wi h VDACs. As expec ed, a unca ed HKI a ian lacking his egion (HKIΔ2-14) ailed o localize o mi ochond ia and displayed a cy osolic dis ibu ion (Fig. 1c). To confi m ha HKI-N alone is su ficien o mi ochond ial localiza ion, we used he 17 N- e minal amino acids o HKI o a HaLo-Tag and exp essed he cons uc in HeLa cells. In wild- ype cells, HKI-N ex ensi ely co- localized wi h OMM ma ke Tom20. In con as , when exp essed in VDAC1/2 double KO cells, HKI-N ailed o a ge mi ochond ia and localized o he cy osol (Fig. 1d). This indica es ha HKI binds VDACs p ima ily ia i s N- e minal helix, possibly in ol ing di ec con ac wi h he bilaye - acing Glu (Fig. 1e). The N- e minal helix o HKI has been shown o bind memb anes e en in he absence o VDACs, p esumably owing o i s pa ially hyd ophobic na u e36. This implies ha mi ochond ial ec ui men o HKI in ol es wo consecu i e s eps, namely inse ion o i s N- e minal helix in he cy osolic leafle o he OMM ollowed by VDAC binding o o m a bina y complex. HeliQues analysis37 e ealed ha HKI-N o ms an α-helix wi h an apola ace composed mos ly o non-pola and hyd ophobic esidues and a pola ace p ima ily con aining hyd ophilic and cha ged esidues (Fig. 2a, b). The amphipa hic na u e o HKI-N p edic s a memb ane binding mode whe eby i s apola ace engages wi h he hyd ophobic memb ane co e and he pola ace wi h he lipid head g oups (Fig. 2c). To in es iga e he memb ane binding a fini y o he HKI-N, we pe - o med coa se-g ain molecula dynamics (CG-MD) simula ions using he Ma ini3 o cefield38,39. A bilaye mimicking he OMM was cons uc ed40 and an α-helical pep ide comp ising HKI-N wi h an addi ional Gln a i s C- e minus (co esponding o Gln18 in HKI) was es ained on o he cy osolic memb ane su ace. A e li ing he es ain s, he deso p ion o he pep ide in o he aqueous phase was moni o ed o e ime41.TheHKI-Npep ide emained memb ane-bound, wi h i s apola ace bu ied in o he hyd o- phobic memb ane co e and wi h esidence imes o >5000 ns. Leu7 is a key componen o he memb ane-o ien ed HKI-N apola ace, si ing a i s e y cen e (Fig. 2b), and hus likely in cons an con ac wi h he hyd ophobic memb ane co e. Subs i u ion o Gln o Leu7 sho ened he HKI-N mem- b ane esidence ime o ~350 ns (Fig. 2d), suppo ing a c i ical ole o he apola ace in memb ane binding. Mo eo e , subs i u ion o Gln o Leu7 in GFP-HKI abolished i s mi ochond ial localiza ion in HeLa cells (Fig. 2e, ). Toge he , hese da a sugges ha memb ane inse ion o i s N- e minal α- helixisap e equisi e o HKIbinding oVDACin heOMM. HKI-N binding o VDACs is di ec ly con olled by p o ona ion o he memb ane-bu ied Glu To elucida e he s uc u al basis o HKI-VDAC complex o ma ion, we nex pe o med CG-MD simula ionso HKI-N binding o VDAC1 and VDAC2. As he o egoing expe imen ssugges ed ha HKI-VDACcomplexassembly equi es a nega i ely-cha ged, memb ane-bu ied Glu esidue on he ou e channelwall(Fig.1a, b), we fi s se ou o es ima e he pKa alues o he co espondingGlu esiduesin VDAC1and VDAC2using i a ableMa ini simula ions42. This e ealed ha he pKa alueo Glu73inVDAC1isshi ed compa ed o a ee glu ama e in solu ion43 bu byless hanoneuni ,i.e. om 4.3 o ~4.8 (Fig. 3b, c). Fo Glu84 o VDAC2, he es ima ed pKa alue is ~5.1 (Supplemen a y Fig 3a, b). This indica es ha a neu al pH, bo h Glu73 in VDAC1and Glu84in VDAC2a e in hei dep o ona ed,nega i elycha ged h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 2 Fig. 1 | Mi ochond ial localiza ion o HKI elies on i s N- e minal α-helix and a memb ane-bu ied Glu in VDACs. a Fluo escence images o wild- ype (WT) and VDAC1/2-DKO HeLa cells exp essing EGFP- agged HKI (g een) alone o in combina ion wi h HA- agged VDAC1, VDAC1E73Q, VDAC1E73D, VDAC2, VDAC2E84Q o VDAC2E84D,fixedand hens ainedwi hDAPI (blue) andan an ibody agains Tom20 (magen a). Line scans showing deg ee o o e lap be ween HKI and Tom20 signals along he pa h o he a ow shown in he zoom-in. Scale ba , 10 μm. bPea son’s co ela ion co-e ficien analysis be ween HKI and Tom20 signals in cells as in (a). Fo each iolin plo , he middle line deno es he median, and he op and bo om lines indica e he 75 h and 25 h pe cen ile. F om le o igh , n= 20 (WT), 20 (VDAC1/2-DKO), 20 (VDAC1/2-DKO +VDAC1), 20 (VDAC1/2- DKO +VDAC1E73Q), 20 (VDAC1/2-DKO +VDAC2), 20 (VDAC1/2- DKO +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) o e a leas 2 independen expe i- men s. p alues we e calcula ed by unpai ed wo- ailed es . cFluo escence images o WT HeLa cells exp essing EGFP- agged HKI o N- e minal unca ion mu an HKI Δ2-14 ,fixed and hen s ained wi h DAPI (blue) and an an ibody agains Tom20 (magen a). Line scans showing deg ee o o e lap be ween HKI and Tom20 signals along he pa h o he a ow shown in he zoom-in. Scale ba , 10 μm. dFluo escence images o li e WT and VDAC1/2-DKO HeLa cells co-exp essing EGFP- agged Tom20 (magen a) and Halo- agged HKI-N (N- e minal HKI esidues 1-17, g een). Line scans showing deg ee o o e lap be ween HKI-N and Tom20 signals along he pa h o he a ow shown in he zoom-in. Scale ba , 10 μm. eModels o complexes o med be ween HKI, Halo- agged HKI-N and VDAC1/2. The memb ane-bu ied Glu is ma ked in ed. h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 3 s a e e en when esiding in he hyd ophobic memb ane in e io . Conse- quen ly, we pe o med CG-MD simula ions o HKI-N binding o VDAC1 andVDAC2 wi h hebilaye - acingGluin hedep o ona ed(cha ged)s a e. CG-MD-simula ions o channels wi h p o ona ed (neu al) Glu esidues se ed as con ol o e i y he impo ance o ha ing Glu in i s nega i ely cha ged o m o HKI-N binding. As he memb ane opology o VDACs is no known, each channel was also simula ed in wo o ien a ions, namely wi h i s C- e minus acing he cy osol—whe e HKI-N was p esen –o he in e -memb ane space (IMS). Main simula ions we e pe o med in an OMM-mimicking bilaye wi h an agg ega e ime o 1.41 ms (Supplemen- a y Table 1)—only a ainable using CG-MD. S ikingly, HKI-N o med s able con ac s wi h bo h VDAC1 and VDAC2 p o ided ha he channel’sC- e minus aced he IMS and he memb ane-bu ied Glu was dep o ona ed (Fig. 3d– ; Supplemen a y Fig. 3c, d). When hese condi ions we e me , he N- e minal hal o HKI-N was obse ed o inse e ically in o he cy osolic memb ane leafle along one side o he channel wall and bind di ec ly o Glu73–in VDAC1 and Glu84–in VDAC2 (Supplemen a y Mo ies 1 and 2). HKI-N esidues mos equen ly in di ec con ac wi h he dep o ona ed Glu we e Me 1, Ala4 and Gln5, all si ua ed on he same side along he axis o he α-helix (Fig. 3g; Supplemen a y Fig. 3 ), wi h Me 1-Glu73/Glu84 con ac s occu ing o 17.20 ± 3.11% and 15.68 ± 5.83% o he agg ega e simula ion ime o VDAC1 and VDAC2, espec i ely (e o s indica e SEMs o e 3 eplica es). Thesebindinge en swe eo enobse edmul iple imespe simula ionand eachedμs du a ions(Fig.3e;Supplemen a yFig.3d; Supplemen a yFig. 4). P o ona ion o he memb ane-bu ied Glu se e ely educed he con ac p e alence o 0.31 ± 0.25% and 2.14 ± 0.74% o VDAC1 and VDAC2, espec i ely. Flipping he memb ane o ien a ion o he channel in each case comple ely abolished complex o ma ion. Unde hese condi ions, HKI-N ailed o inse in o he cy osolic leafle and no in e ac ion wi h he bilaye - acing Glu occu ed. Ins ead, con ac s wi h VDAC1 and VDAC2 became andom and sho -li ed (<10 ns), in ol ing channel esidues acing he cy osol (Fig. 3 , g; Supplemen a y Fig. 3e, ). Consis en wi h he localiza ion s udies o GFP- agged HKI in HeLa cells (Fig. 1a, b), VDAC channels wi h a Glu- o-Gln subs i u ion lacked a fini y o HKI-Nin simula ions, ega dless o hei ansbilaye o ien a ion (0.06 ± 0.05% and 1.67 ± 0.31% con ac p e alence o VDAC1 and VDAC2, espec i ely). On he o he hand, VDAC channels wi h a Glu- o- Asp subs i u ion e ained he abili y o bind HKI-N, p o ided ha he Asp was dep o ona ed and he channel’sC- e minus aced he IMS (17.39 ± 4.50% and 36.30 ± 13.26% con ac p e alence o VDAC1 and VDAC2, espec i ely; Supplemen a y Fig. 5). Collec i ely, hese esul s indica e ha HKI-VDAC binding c i ically elies on bo h he memb ane opology o VDACs and he p o ona ion s a e o he bilaye - acing Glu. Acidifica ion o cy osolic pH igge s dissocia ion o HKI-N om mi ochond ia To challenge he idea ha HKI-VDAC complex o ma ion is con olled by he p o ona ion s a e o he bilaye - acing Glu, we nex in es iga ed he impac o cy osolic acidifica ion on he subcellula dis ibu ion o Halo- agged HKI-N in HeLa cells. Cy osolic pH was adjus ed by incuba ing cells inabu e wi h hedesi edpHin hep esenceo H+/K+ionopho e nige icin p < 0.0001pp < 0.0001 Fig. 2 | HKI-N binding o memb anes. a A omic model o HKI (PDB: 1BG3, cyan) wi h he N- e minal α-helix (HKI-N) highligh ed in esidue- ype colo ing. bHeliQues analysis and VDW/Dynamic Bonds ep esen a ion o he coa se- g ained HKI-N backbone e eals an α-helix wi h a pola and apola ace. cModel p edic ing ha he apola ace o HKI-N media es memb ane binding, wi h he fi s hal o he α-helix p o uding deepe in o he memb ane bilaye . dMemb ane esidence ime analysis o HKI-N and HKI-NL7Q using CG-MD simula ions. Helices we e bound o OMM-mimicking memb anes ollowing he es aining p o ocol desc ibed in Me hods. A e li ing he es ain s, he dis ance o he helix esidue closes o he memb ane’s op leafle was measu ed, un il i su passed 1.4 nm. S ills show ep esen a i e configu a ions om each condi ion. Plo s ep esen he ime p og ession o he helix-memb ane dis ances o six independen eplicas pe condi ion, as o e laid semi anspa en aces; e ical ises co espond o each ace’s memb ane-lea ing e en , om which poin ha ace is no longe d awn. eFluo escence images o WT HeLa cells exp essing EGFP- agged HKI o HKIL7Q (g een), fixed and hen s ained wi h DAPI (blue) and an an ibody agains Tom20 (magen a). Line scans showing deg ee o o e lap be ween HKI and Tom20 signals along he pa h o he a ow shown in he zoom-in. Scale ba , 10 μm. Pea son’s co ela ion co-e ficien analysis be ween HKI and Tom20 signals in cells as in (e). n= 20 (HKI) and 28 cells (HKIL7Q) o e h ee independen expe imen s. p alues we e calcula ed by unpai ed wo- ailed es . h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 4 (Fig. 4a; Supplemen a y Fig. 6a). Equilib a ion o cy osolic pH wi h he pH o he ex e nal bu e was quan i a i ely assessed wi h he in acellula pH indica o pH odoTM RedAM(Supplemen a yFig.6b).Tomoni o ad op in cy osolic pH in eal ime, we ook ad an age o he ac ha he fluo opho e o EGFP is mo e sensi i e o acidic pH when compa ed o mChe y44 and HaloTag Ligand JF646. Thus, in cells exp essing Tom20-EGFP, JF646- Fig. 3 | HKI-N binding o VDAC1 is di ec ly con olled by he p o ona ion s a e o he memb ane-bu ied Glu. a A omic model o HKI (cyan, wi h esidue- ype colo ed HKI-N) bound o VDAC1 (yellow) wi h he memb ane-bu ied Glu (E73) ma ked in ed.bS ill om a i a able MD simula ion o VDAC1 (yellow) o e alua e he p o ona ion s a e o E73 ( ed) a pH 5.0. PO 4 g oups in he POPC-based bilaye a e ma ked in o ange and p o ons a e ma ked in blue.cTi a ion cu e showing he deg ee o dep o ona ion o E73 in VDAC1, simula ed a a pH ange o 3–8. dS ills om an MD simula ion showing HKI-N bound o VDAC1 wi h a dep o ona ed E73 ( ed) and IMS- acing C- e minus. Glyce ol g oups in he OMM-mimicking bilaye a e ma ked in cyan.eS ills om an MD simula ion, showing he app oach and binding o HKI-N o VDAC1 wi h a dep o ona ed E73 ( ed) and IMS- acing C- e minus. Rela i e du a ion o con ac s be ween HKI-Me 1 and specific esidues o VDAC1 wi h a p o ona ed o dep o ona ed E73 and cy osol- o IMS- acing C- e minus simula ed in an OMM-mimicking bilaye . Shown a e he combined da a o h ee indi idual eplicas wi h a o al simula ion ime be ween 169 μs and 211 μs pe condi ion.gRela i edu a ion o con ac sbe weenVDAC1-E73 andspecific esidues o HKI-N unde he same condi ions as in ( ). h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 5 labeled Tom20-Halo and mChe y used o he OMM ancho o AKAP1 (OMM-mChe y), a shi in cy osolic pH om 7.4 o 6.0 s ongly educed EGFP fluo escence wi hou a ec ing he o he wo fluo opho es (Fig. 4b, c; Supplemen a y Fig. 6c, d). S ikingly, acidifica ion o he cy osol eadily igge ed he ansloca ion o JF646-labeled HKI-N-Halo om mi ochond ia in o he cy osol. Dissocia ion o HKI-N-Halo om mi o- chond ia was al eady measu able when lowe ing he cy osolic pH o 6.8 and g adually p og essed wi h inc eased acidifica ion so ha a pH 6.0 he bulk o HKI-N-Halo esided in he cy osol (Fig. 4b, d; Supplemen a y Fig. 6c, d). Raising he cy osolic pH om 6.0 back o 7.4 es o ed he mi ochond ial h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 6 localiza ion o HKI-N-Halo (Fig. 4b). Consis en wi h he CG-MD simu- la ions, hese esul s suppo he no ion ha HKI-VDAC binding is con- olled by he p o ona ion s a e o he bilaye - acing Glu e en hough we canno exclude ha p o ona ion o addi ional acidic esidues also play a ole. HKI-VDAC binding c i ically elies on an asymme ic posi ioning o he memb ane-bu ied Glu The o egoing CG-MD simula ions e ealed ha he ansbilaye o ien a- ion o VDACs is a c i ical de e minan o HKI binding (Fig. 3 , g; Sup- plemen a y Fig. 3e, ). In e es ingly, we no iced ha he memb ane-bu ied Glu in VDACs is asymme ically posi ioned a ew Å away om he bilaye cen e and esides in he cy osolic leafle when he channel’sC- e minus aces he IMS, he o ien a ion compa ible wi h HKI binding (Fig. 5a). We he e o e hypo hesized ha channels wi h he opposi e opology may ail o bind HKI because he memb ane-bu ied Glu in ha o ien a ion lies oo deep in he lipid bilaye o he enzyme’sN- e minal α-helix o make s able con ac s. To e i y his idea, we subs i u ed Phe o Glu73 and Glu o Phe71 in VDAC1, e ec i ely c ea ing a channel in which he asymme ic posi ion o he memb ane-bu ied Glu is flipped ac oss he bilaye cen e (Fig. 5a). Nex , we pe o med CG-MD simula ions o p obe HKI-N binding o he VDAC1E73F/F71E mu an channel in bo h memb ane o ien a ions and wi h a dep o ona edGlu.UnlikeVDAC1, heVDAC1E73F/F71E a ian wasunable o o m s able con ac s wi h HKI-N i espec i e o i s ansbilaye o ien a ion (Fig. 5b). Mo eo e , unlike VDAC1, he VDAC1E73F/F71E a ian comple ely ailed o es o e mi ochond ial localiza ion o GFP-HKI in VDAC1/2- double KO cells (Fig. 5c). These esul s indica e ha bilaye dep h o he cha ged Glu on he ou e channel wall, al hough c i ical, is no he sole Fig. 4 | Cy osolic pH con ols mi ochond ial associa ion o HKI-N. a Schema ic ou line o expe imen al s a egy o de e mine he impac o cy osolic acidifica ion on mi ochond ial associa ion o HKI-N. bFluo escence images o li e HeLa cells co- exp essing EGFP- agged Tom20 (g een), OMM-mChe y (cyan)andHalo- aggedHKI- N(magen a) g own in Op imem ( op), ea ed wi h 10 μM nige icin in pH 6.0 bu e o 5 min (middle) and hen wi h 10 μM nige icin in pH 7.4 bu e o 5 min (bo om). Line scans showing deg ee o o e lap be ween OMM and HKI-N signals along he pa h o he a ow shown in he zoom-in. Scale ba , 10 μm. cFluo escence images o li e HeLa cells co-exp essing EGFP- agged Tom20 (g een), OMM-ancho ed mChe y (cyan)andHalo- agged Tom20 (magen a) ea ed as in (b). Line scans showing deg ee o o e lap be ween OMM and Tom20-Halo signals along he pa h o he a ow shown in he zoom-in. Scale ba , 10 μm. dQuan i a i e assessmen o mi ochond ia-associa ed le els o OMM- mChe y, Tom20-EGFP, Tom20-Halo and HKI-N-Halo in li e HeLa cells a e ea - men wi h nige icin in bu e a indica ed pH o 5 min. Fluo escence alues in co e- sponding pH bu e we e se ela i e o alues o same cell in Op i-MEM. Da a a e means ± SD, n= 6 cells pe condi ion o e ou independen expe imen s. p alues we e calcula ed by unpai ed wo- ailed es . Fig. 5 | HKI-VDAC binding c i ically elies on an asymme ic posi ioning o he memb ane-bu ied Glu. a S ills om MD simula ions o VDAC1 and VDAC1E73F/F71E wi h he memb ane- acing Glu and Phe esidues a posi ions 71 and 73 ep esen ed as ed and whi e balls, espec i ely. The g aphs show he posi ion o Glu73 in VDAC1 and Glu71 in VDAC1E73F/F71E ela i e o he memb ane cen e (dashed line) o e he cou se o a simula ion. bRela i e du a iono con ac sbe weenHKI-Me 1andspecific esidues o VDAC1o VDAC1E73F/F71E wi hcy osol-o IMS- acing C- e mini. Da a o VDAC1 a e aken om Fig. 3 and shown o compa ison. Fo VDAC1E73F/F71E da a o h ee indi idual simula ionswe e combined wi h a o al simula ion ime be ween 148 μs and 162 μs pe condi ion. cFluo escence images o VDAC1/2-DKO HeLa cells co- exp essing EGFP- agged HKI (g een) and HA- agged VDAC1 o VDAC1E73F/F71E, fixed and hen s ained wi h DAPI (blue) and an ibodies agains he HA-epi ope ( ed) and Tom20 (magen a). Line scans showing deg ee o o e lap be ween HKI and Tom20 signals along he pa h o he a ow shown in he zoom-in. Scale ba , 10 μm. h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 7 de e minan o HKI binding and ha o he unique ea u es on he memb ane- acing su ace o VDACs also play a ole. VDAC channels cause hinning o he lipid monolaye p oximal o he memb ane-bu ied Glu VDACs expe ience a global hyd ophobic misma ch wi h he lipid bilaye in which heya einse edin—ha inga hyd ophobicin e aceo only~2.4 nm, which is significan ly less han ha o biological memb anes (~4 nm)45. Cumula i ely wi h his o e all misma ch, p e ious MD simula ions o VDAC1 e ealed addi ional memb ane hinning and wa e de ec s nea he ou wa d- acing Glu5,46. By ex ending hese s udies o VDAC1 in i s HKI binding-compe en o ien a ion (wi h he channel’sC- e minus acing he IMS), we ound ha his localized memb ane hinning is mainly confined o he cy osolic leafle , adjacen o he nega i ely cha ged Glu (E73–;Fig.6a). He e, he cy osolic leafle eached a minimal hickness o 0.71 ± 0.04 nm, which was conside ably hinne han he a e age hicknessnea he channel wall ou side o his egion (1.49 ± 0.01 nm) o in he absence o p o ein (1.94 ± 0.001 nm; Supplemen a y Fig. 7a, b). In he de ec egion, we also Fig. 6 | VDAC channels cause lipid packing de ec s and memb ane leafle hin- ning p oximal o he bilaye - acing Glu. (a) Leafle -specific memb ane hinning g aphs o VDAC1 and VDAC2 simula ed in a POPC bilaye wi h C- e mini acing he IMS leafle . G ay sphe es indica e he VDAC backbone and he posi ion o he bilaye acing Glu is ma ked by an a ow. Memb ane hinning was calcula ed as he a e age dis ance o he lipid backbone phospha es o he global memb ane cen e . (b) Leafle -specific wa e de ec g aphs o VDAC1 and VDAC2 simula ed as in (a). Wa e de ec s we e calcula ed as he amoun o wa e molecules de ec ed wi hin a z-dis ance o 1.5 nm o he global memb ane cen e . (c) Occupancies o lipid PO 4 g oups in simula ions o VDAC1 and VDAC2 as in (a). Occupancy su aces enclose olumes wi h a e age occupancy o 0.5% o g ea e . The posi ion o he bilaye - acing Glu is ma ked. (d) Cy osolic leafle hinning and wa e de ec g aphs o VDAC1, VDAC1E73Q and VDAC1E73D simula ed in a POPC bilaye wi h C- e mini acing he IMS (bo om) leafle . The bilaye - acing acidic esidues we e p o ona ed o dep o ona ed, as indica ed. Analysis was done as in (a) and (b). h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 8 obse ed a la ge deg ee o wa e pene a ion (Fig. 6b). Simula ions o VDAC2 e ealed a simila hinning o he cy osolic leafle along wi h wa e de ec snea hecha gedGlu(E73 –;Fig.6a, b), wi h a minimal hickness o 0.52 ± 0.05 nm and an a e age hickness o 1.46 ± 0.01 nm nea he channel wall ou side o he de ec egion (Supplemen a y Fig. 7a, b). When mapping he occupancy o he lipid phospha es o plo ing he hickness o VDAC- su ounding lipids, we obse ed ha he egion o memb ane hinning did no pe ec ly o e lap wi h he posi ion o he cha ged Glu (Fig. 6c; Sup- plemen a y Fig. 7b). This sugges ed ha memb ane hinning may no ely on a cha ged Glu bu a he on ou wa d- acing pola esidues in i s icini y. Indeed, p o ona ion o he bilaye - acing Glu o i s subs i u ion by Gln in VDAC1 o VDAC2 g ea ly diminished he wa e de ec s in ei he case, bu had li le impac on he memb ane hinning capaci y o he channels (Fig. 6d; Supplemen a y Fig. 8). When he bilaye - acing Glu was eplaced by a dep o ona ed Asp (D73–), memb ane hinning and wa e de ec s we e e ained (Fig. 6d; Supplemen a y Fig. 8). Hence, while he nega i ely cha ged Glu c ea es condi ions ha acili a e he pene a ion o wa e , i appea s ha he memb ane hinning capaci y o VDACs is media ed by o he esidues on he ou e channel wall. Pola esidues p oximal o he memb ane-bu ied Glu p o ide a ga eway o HKI-VDAC binding We conside ed ha hinning o he cy osolic leafle nea he memb ane- bu ied Glu o VDACs may p o ide a low-ene gy passageway o he N- e minalα-helix o HKI o acili a e HKI-VDAC binding. A close inspec ion o he ou e wall o VDAC1 in a eas exhibi ing he highes deg ee o memb ane hinning e ealed wo pola esidues, Th 77 and Se 101, which a e posi ioned wi hin close ange o he memb ane-bu ied Glu (Fig. 7a). In CG-MD simula ions, subs i u ion o Leu o Th 77 o Se 101 in each case led o a localized bu ma ked educ ion in he memb ane hinning capaci y o VDAC1 ca ying a cha ged Glu (Fig. 7a, b; Supplemen a y Fig. 7c). The a eas occupied by Th 77 and Se 101 each had i s own local minimum o leafle hickness ha wasselec i ely abolishedby mu a ion.The hicknesson he S101 side displayed he lowes local minimum and co esponds o he global minimum (0.71 ± 0.04 nm). Fo ha eason, he T77L mu a ion did no a ec he global minimum (0.65 ± 0.01 nm). Upon in oducing he S101L mu a ion, he minimum on he T77 side becomes he new global minimum wi h a highe hickness alue (0.90 ± 0.01 nm). When he wo mu a ions we e combined, leafle hinning in he egion p oximal o he cha ged Glu was u he educed (1.11 ± 0.01 nm), essen iallyabolishing he leafle hinning specific o his egion(Fig.7a, b; Supplemen a y Fig. 7c). This was accompanied by a subs an ial educ ion in wa e de ec s (Fig. 7b). These esul s indica e ha Th 77 and Se 101 each con ibu e o a local dis o ion o he cy osolic memb ane leafle , possibly acili a ing access o HKI o he cha ged, memb ane-bu ied Glu. Consis en wi h his idea, CG- MD simula ions e ealed ha subs i u ion o Leu o Th 77 o Se 101 in VDAC1 diminished con ac s be ween HKI-N and he cha ged Glu ( om 17.20 ± 3.11% o 10.77 ± 4.75% and 11.11 ± 4.08% con ac p e alence o T77L and S101L, espec i ely; Fig. 7c, d). Combining hese subs i u ions u he educed HKI-N binding o a 4.57 ± 2.20% con ac p e alence. Looking a con ac li e ime dis ibu ions, i can be seen ha hese mu a ions a ec binding by educing he on- a e a he han o - a e o he binding p ocess (Supplemen a y Fig. 4), indica ing ha he pola ace su ounding he Glu ac s indeed as an access pa hway. Impo an ly, he diminished capaci y o he mu an channels o bind HKI-N in silico s ongly co ela ed wi h an impai ed abili y o hese channels o es o e mi ochond ial ec ui men o HKI in VDAC1/2-double KO cells (Fig. 7e, ). Collec i ely, hese esul s indica e ha Th 77 and Se 101 a e co e componen s o a memb ane hinning pa hway by which he N- e minal α-helix o HKI gains access o he memb ane-bu iedGluo VDACs, he ebyp o idinga ga eway o HKI-VDAC binding. Discussion While binding o HKI o mi ochond ial VDACs is c ucial o cell g ow h and su i al, he s uc u al basis o HKI-VDAC complex assembly is no known. Using a CG-MD simula ions app oach complemen ed wi h unc ional s u- dies in cells, we iden ified co e s uc u al and physicochemical ea u es ha go e n binding o HKI o VDAC1 and VDAC2. As schema ically ou lined in Fig. 8, ou esul s indica e ha a bilaye - acing nega i ely cha ged Glu on he ou e channel wall plays a c ucial ole in HKI binding by p omo ing s able con ac s be ween he channel and he enzyme’samphipa hicN- e minal α- helix (HKI-N). P o ona ion o he Glu esidue abolishes HKI-N binding in simula ions while ansien acidifica ion o he cy osol causes a e e sable elease o HKI-N om mi ochond ia. Memb ane inse ion o HKI occu s adjacen o he cha ged Glu whe e a pai o pola channel esidues causes a ma ked hinning o he cy osolic memb ane leafle , c ea ing a unnel ha likely se es as low-ene gy passageway o he enzyme’sN- e minal α-helix o acili a e complex assembly. Consis en wi h his model, we ound ha dis- up ing he memb ane hinningcapaci y o VDAC1 significan lyimpai ed i s abili y o bind HKI bo h in silico and in cells. In line wi h p e ious wo k47, we demons a e ha HKI-N is essen ial and su ficien o VDAC binding. Howe e , HKI-N can also bind mem- b anes independen ly o VDACs36. B eaking he apola ace o HKI-N by a single poin mu a ion significan ly weakened memb ane binding in silico and abolished mi ochond ial localiza ion o HKI in VDAC1/2-exp essing cells. F om his we in e ha memb ane pa i ioning o HKI-N is a p e- equisi e o VDAC binding. Ou findings a e ha d o econcile wi h a p e ious model o HKI-VDAC complex o ma ion ha is based on di ec plugging o HKI-N in o he channel’scen alpo e 27. Ins ead, ou da a indica e ha HKI-VDAC complex assembly is a mul is ep p ocess whe eby HKI ini ially binds he OMM h ough memb ane adso p ion in ol ing he apola in e ace o HKI-N. We en ision ha hinning o he cy osolic memb ane leafle by a pai o pola channel esidues, Th 77 and Se 101 in VDAC1,c ea esa unnel ha se esas he modynamic ap o HKI binding by enabling he enzyme’sN- e minal α-helix o il and inse a he VDAC/memb ane in e ace o become aligned o s able in e ac ions wi h he cha ged Glu on he ou e channel wall. Addi ionally, ou da a p o ide impo an clues ega ding he ansbilaye o ien a iono VDACchannelsin heOMM.Thesidednesso heseβ-ba el p o eins has been p obed wi h a ious app oaches wi hou eaching gene al consensus. Fo ins ance, s udies onhuman VDAC1 ca ying a clea age si e o cy osolic caspases indica e ha he channel’sC- e minus aces he IMS34.In con as , a spli -NeonG een complemen a ion s udy sugges s ha he C- e - minus o human VDAC2 aces he cy osol35. Based on packing analysis o mu ine VDAC1 c ys als in a lipidic en i onmen , Ujwal e al48.p oposed ha VDACs a e dual opology memb ane p o eins ha may achie e an i-pa allel a angemen s in he OMM. Howe e , ou MD simula ions clea ly indica e ha HKI-VDAC complex o ma ion is only possible wi h channels in one o ien a ion, namely whe eby hei C- e mini ace he IMS. I is only in his o ien a ion ha he pola channel esidues c i ical o memb ane- hinning a e posi ioned accu a ely o es ablish a passageway o cy osolic HKI o each he bilaye - acing Glu and o m a s able complex. While ou findings do no ule ou he possibili y o a dual opology o VDAC channels, hey clea ly indica e ha onlyoneo he wopossible ansbilaye o ien a ionsp o idesa unc ional binding pla o m o HKI. Ti a able MD simula ions o VDAC1 and VDAC2 e ealed ha a neu al pH, he bilaye - acing Glu is p edominan ly in i s dep o ona ed, ully nega i ely-cha ged s a e. Al hough i is ene ge ically un a o able o a cha ged esidue o be exposed o he hyd ophobic memb ane in e io , memb ane hinning imposed by pola esidues in close p oximi y o he bilaye - acing Glu may explain why i s pKa alue is shi ed by less han one uni in compa ison o a ee Glu. Con e ging lines o e idence indica e ha he p o ona ion s a us o he bilaye - acing Glu is a key de e minan o HKI binding. To begin wi h, p o ona ion o his Glu in VDAC1 and VDAC2 in each case p o ed su ficien o ab oga e HKI-N binding in simula ions. Replacing Glu wi h he non- i a able Gln abolished HKI-N binding o VDAC channels in simula ions and dis up ed VDAC-dependen mi o- chond ial localiza ion o HKI in cells. Con e sely, eplacing Glu o i a- able Asp p omo ed complex o ma ion bo h in silico and in cells. Mild acidifica ion o he cy osol om pH 7.4 o 6.0 ins an ly dissocia ed HKI-N h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 9 37. Gau ie , R., Dougue , D., An onny, B. & D in, G. HELIQUEST: A web se e o sc een sequences wi h specificα-helical p ope ies. Bioin o ma ics 24, 2101–2102 (2008). 38. Souza, P. C. T. e al. Ma ini 3: a gene al pu pose o ce field o coa se-g ained molecula dynamics. Na . Me hods 18, 382–388 (2021). 39. Ma ink, S. J. e al. Two decades o Ma ini: Be e beads, b oade scope. Wiley In e discip. Re .: Compu . Mol. Sci. 13, e1620 (2023). 40. Ho a h, S. E. & Daum, G. Lipids o mi ochond ia. P og. Lipid Res. 52, 590–614 (2013). 41. Melo, M. N. Coa se-g ain simula ions o memb ane-adso bed helical pep ides. Me hods Mol. Biol. 2405, 137–150 (2022). 42. Chia iello, M. G., G ünewald, F., Za mien o-Ga cia, R. & Ma ink, S. J. pH-dependen con o ma ional swi ch impac s s abili y o he PsbS dime . J. Phys. Chem. Le . 14, 905–911 (2023). 43. Isom, D. G., Cas añeda, C. A., Cannon, B. R., Velu, P. D. & Ga cía- Mo eno, E. B. Cha ges in he hyd ophobic in e io o p o eins. P oc. Na l Acad. Sci. Usa. 107, 16096–16100 (2010). 44. Dohe y, G. P., Bailey, K. & Lewis, P. J. S age-specificfluo escence in ensi y o GFP and mChe y du ing spo ula ion in Bacillus Sub ilis. BMC Res. No es 3, 303 (2010). 45. La a gue, E. e al. Lipid composi ion o he memb ane go e ns he oligome ic o ganiza ion o VDAC1. h ps://doi.o g/10.1101/2024.06. 26.597124 (2024). 46. Villinge , S. e al. Func ional dynamics in he ol age-dependen anion channel. P oc. Na l Acad. Sci. Usa. 107, 22546–22551 (2010). 47. Rossi, A. e al. De ec i e mi ochond ial py u a e flux a ec s cell bioene ge ics in Alzheime ’s disease- ela ed models. Cell Rep. 30, 2332–2348 (2020). 48. Ujwal, R., Cascio, D., Chap al, V., Ping, P. & Ab amson, J. C ys al packing analysis o mu ine VDAC1 c ys als in a lipidic en i onmen e eals no el insigh s on oligome iza ion and o ien a ion. Channels (Aus in). 3, 167–170 (2009). 49. Quach, C. H. T. e al. Mild alkaliza ion acu ely igge s he Wa bu g e ec by enhancing hexokinase ac i i y ia ol age-dependen anion channel binding. PLoS ONE 11, e0159529 (2016). 50. Be gdoll, L. A. e al. P o ona ion s a e o glu ama e 73 egula es he o ma ion o a specific dime ic associa ion o mVDAC1. P oc. Na l Acad. Sci. Usa. 115, E172–E179 (2017). 51. Sun, L., Shukai , S., Naik, T. J., Moazed, F. & A dehali, H. Glucose phospho yla ion and mi ochond ial binding a e equi ed o he p o ec i e e ec s o hexokinases I and II. Mol. Cell. Biol. 28, 1007–1017 (2008). 52. Lau e wasse , J. e al. Hexokinases inhibi dea h ecep o -dependen apop osis on he mi ochond ia. P oc. Na l Acad. Sci. USA 118, e2021175118 (2021). 53. Co be , C. & Fe on, O. Tumou acidosis: om he passenge o he d i e ’s sea . Na . Re . Cance 17, 577–593 (2017). 54. Reshkin, S. J. e al. Na+/H+exchange ‐dependen in acellula alkaliniza ion is an ea ly e en in malignan ans o ma ion and plays an essen ial ole in he de elopmen o subsequen ans o ma ion‐ associa ed pheno ypes. FASEB J. 14, 2185–2197 (2000). 55. G illo-Hill, B. K., Choi, C., Jimenez-Vidal, M. & Ba be , D. L. Inc eased H+e flux is su ficien o induce dysplasia and necessa y o iabili y wi h oncogene exp ession. Eli e 4, e03270 (2015). 56. Dadsena, S., Hassan, D. G. & Hol huis, J. C. M. Un a eling he molecula p inciples by which ce amides commi cells o dea h. Cell S ess 3, 280–283 (2019). 57. Pa wa dhan, G. A., Be e ly, L. J. & Siskind, L. J. Sphingolipids and mi ochond ial apop osis. J. Bioene g. Biomemb . 48,153–168 (2016). 58. Og e men, B. Sphingolipid me abolism in cance signalling and he apy. Na . Re . Cance 18,33–50 (2018). 59. Galluzzi, L., Kepp, O., Tajeddine, N. & K oeme , G. Dis up ion o he hexokinase-VDAC complex o umo he apy. Oncogene 27, 4633–4635 (2008). 60. Smilansky, A. e al. The ol age-dependen anion channel 1 media es amyloid β oxici y and ep esen s a po en ial a ge o Alzheime disease he apy. J. Biol. Chem. 290, 30670–30683 (2015). 61. Mag i, A. & Messina, A. In e ac ions o VDAC wi h p o eins in ol ed in neu odegene a i e agg ega ion: an oppo uni y o ad ancemen on he apeu ic molecules. Cu . Med. Chem. 24, 4470–4487 (2017). 62. Shoshan-Ba ma z, V., Zaka , M., Rosen hal, K. & Abu-Hamad, S. Key egions o VDAC1 unc ioning in apop osis induc ion and egula ion by hexokinase. Biochim. Biophys. Ac a—Bioene g. 1787, 421–430 (2009). 63. Jain, A., Beu el, O., Ebell, K., Ko nee , S. & Hol huis, J. C. M. Di e ing CERT-media ed ce amide anspo o mi ochond ia igge s Bax- dependen apop osis. J. Cell Sci. 130, 360–371 (2017). 64. Jain, A., Dadsena, S. & Hol huis, J. C. M. A swi chable ce amide ans e p o ein o dissec ing he mechanism o ce amide-induced mi ochond ial apop osis. FEBS Le . 594, 3739–3750 (2020). 65. Gha a i Zaki, A. e al. De elopmen o a chemogene ic app oach o manipula e in acellula pH. J. Am. Chem. Soc. 145, 11899–11902 (2023). 66. A ganda-Ca e as, I. e al. T ainable Weka segmen a ion: a machine lea ning ool o mic oscopy pixel classifica ion. Bioin o ma ics 33, 2424–2426 (2017). 67. Mulichak, A. M., Wilson, J. E., Padmanabhan, K. & Michael Ga a i o, R. The s uc u e o mammalian hexokinase-1. Na . S uc . Biol. 5, 555–560 (1998). 68. Böhm, R. e al. The s uc u al basis o low conduc ance in he memb ane p o ein VDAC upon β-NADH binding and ol age ga ing. S uc u e 28, 206–214 (2020). 69. Sch edelseke , J. e al. High esolu ion s uc u e and double elec on- elec on esonance o he zeb afish ol age-dependen anion channel 2 e eal an oligome ic popula ion. J. Biol. Chem. 289, 12566–12577 (2014). 70. Aleshin, A. E. e al. The mechanism o egula ion o hexokinase: New insigh s om he c ys al s uc u e o ecombinan human b ain hexokinase complexed wi h glucose and glucose-6-phospha e. S uc u e 6,39–50 (1998). 71. Cla e as Cabezudo, A., A hanasiou, C., Tsengenes, A. & Wade, R. C. Scaling p o ein–wa e in e ac ions in he ma ini 3 coa se-g ained o ce field o simula e ansmemb ane helix dime s in di e en lipid en i onmen s. J. Chem. Theo y Compu . 19, 2109–2119 (2023). 72. Spin i, J. K., Nei a Nunes, F. & Melo, M. N. Room o imp o emen in he ini ial ma ini 3 pa ame e iza ion o pep ide in e ac ions. Chem. Phys. Le . 819, 140436 (2023). 73. K oon, P. C. e al. Ma inize2 and e mou h: unified amewo k o opology gene a ion. Eli e 12, RP90627 (2023). 74. Wassenaa , T. A., Ingól sson, H. I., Böckmann, R. A., Tieleman, D. P. & Ma ink, S. J. Compu a ional lipidomics wi h insane: a e sa ile ool o gene a ing cus om memb anes o molecula simula ions. J. Chem. Theo y Compu . 11, 2144–2155 (2015). 75. Pede sen, K. B. e al. The Ma ini 3 lipidome: expanded and efined pa ame e s imp o e lipid phase beha io . h ps://doi.o g/10.26434/ chem xi -2024-8bj (2024). 76. Pe iole, X., Ca alli, M., Ma ink, S. J. & Ce uso, M. A. Combining an elas ic ne wo k wi h a coa se-g ained molecula o ce field: S uc u e, dynamics, and in e molecula ecogni ion. J. Chem. Theo y Compu . 5, 2531–2543 (2009). 77. Hess, B., Ku zne , C., Van De Spoel, D. & Lindahl, E. GRGMACS 4: Algo i hms o highly e ficien , load-balanced, and scalable molecula simula ion. J. Chem. Theo y Compu . 4, 435–447 (2008). h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 16 78. Su, J., Ma ink, S. J. & Melo, M. N. Localiza ion p e e ence o an imic obial pep ides on liquid-diso de ed memb ane domains. F on . Cell De . Biol. 8, 350 (2020). 79. G ünewald, F. e al. Ti a able Ma ini model o cons an pH simula ions. J. Chem. Phys. 153, 024118 (2020). 80. Van Guns e en, W. F. & Be endsen, H. J. C. A leap- og algo i hm o s ochas ic dynamics. Mol. Simul. 1, 173–185 (1988). 81. Timme, M. & Melo, M. N. Analysis sc ip s and compu a ional models used in ‘Hexokinase-I di ec ly binds o a cha ged memb ane-bu ied glu ama e o mi ochond ial VDAC1 and VDAC2’[da a se ]. (2024). h ps://doi.o g/10.5281/zenodo.14144611. Acknowledgemen s The au ho s g a e ully acknowledge Ladisla Ba os and Robe Vácha (Na ional Cen e o Biomolecula Resea ch, Masa yk Uni e si y, B no, Czech Republic) o p o iding he sc ip s o memb ane hinning and wa e de ec s analysis, and Va da Shoshan-Ba ma z (Ben-Gu ion Uni e si y o he Nege , Is ael) o he pEGFP-HKI cons uc .This wo k was suppo ed by he Deu sche Fo schungsgemeinscha (378148610 and 448344643 o J.C.M.H.), he Ge man Egyp ian Resea ch Long- e m Schola ship P og am (GERLI p ojec 57222240 o D.G.H.), he Eu opean Resea ch Council (ERC Ad anced g an 101053661 „COMP-O-CELL“ o S.J.M.) and he FCT— Fundação pa a a Ciência e a Tecnologia I.P. ( h ough MOSTMICRO-ITQB R&D Uni wi h p ojec s UIDB/04612/2020 and UIDP/04612/2020, and LS4FUTURE Associa ed Labo a o y wi h p ojec s LA/P/0087/2020 and CEECIND/04124/2017/CP1428/CT0008 o M.N.M.). Au ho con ibu ions M.N.M. and J.C.M.H. designed he esea ch wi h c i ical inpu om S.B. and M.T.; S.B. pe o med expe imen s in cells wi h c i ical inpu om D.H.; M.T. ca ied ou he CG-MD simula ions wi h c i ical inpu om N.W.; C.M.B. ca ied ou all i a able MD simula ions; J.C.M.H. p o ided expe ise o expe imen s in cells and helped in e p e he da a; M.N.M. and S.J.M. p o- ided expe ise o CG-MD simula ions and helped in e p e he da a; J.C.M.H. w o e he manusc ip ; all au ho s discussed esul s and com- men ed on he manusc ip . Funding Open Access unding enabled and o ganized by P ojek DEAL. Compe ing in e es s The au ho s decla e no compe ing in e es s. Addi ional in o ma ion Supplemen a y in o ma ion The online e sion con ains supplemen a y ma e ial a ailable a h ps://doi.o g/10.1038/s42003-025-07551-9. Co espondence and eques s o ma e ials should be add essed o Manuel N. Melo o Joos C. M. Hol huis. Pee e iew in o ma ion Communica ions Biology hanks Subhendu Ghosh and he o he , anonymous, e iewe (s) o hei con ibu ion o he pee e iew o his wo k. P ima y Handling Edi o s: Lau a Rod íguez Pé ez and Ch is ina Ka lsson-Rosen hal. A pee e iew file is a ailable. Rep in s and pe missions in o ma ion is a ailable a h p://www.na u e.com/ ep in s Publishe ’s no e Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in published maps and ins i u ional a filia ions. Open Access This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License, which pe mi s use, sha ing, adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as long as you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons licence, and indica e i changes we e made. The images o o he hi d pa y ma e ial in his a icle a e included in he a icle’s C ea i e Commons licence, unless indica ed o he wise in a c edi line o he ma e ial. I ma e ial is no included in he a icle’s C ea i e Commons licence and you in ended use is no pe mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will need o ob ain pe mission di ec ly om he copy igh holde . To iew a copy o his licence, isi h p://c ea i ecommons.o g/licenses/by/4.0/. © The Au ho (s) 2025 h ps://doi.o g/10.1038/s42003-025-07551-9 A icle Communica ions Biology | (2025) 8:212 17