Phospholipid flipping facilitates annexin translocation across membranes

Annexins are phospholipid binding proteins that somehow translocate from the inner leaflet of the plasma membrane to the outer leaflet 1–4. For example, Annexin A2 is known to localise to the outer leaflet of the plasma membrane (cell surface) where it is involved in plasminogen activation leading to fibrinolysis and cell migration, among other functions 1,5–9. Despite having well described extracellular functions, the mechanism of annexin transport from the cytoplasmic inner leaflet to the extracellular outer leaflet of the plasma membrane remains unclear. As annexin A2 and A5 bind to negatively charged lipid head groups of the inner and outer leaflets of the plasma membrane in a calcium-dependent manner 4,10, we hypothesised that lipid remodelling may be involved in their translocation to the cell surface. Here, we show that phospholipid flipping activity is crucial for the transport of annexins A2 and A5 across membranes in cells and in liposomes. The lantibiotic, cinnamycin, which flips lipids 11,12, facilitates the transport of annexin A2 and A5 to the cell surface in many types of human cells. Furthermore, we identified TMEM16F (anoctamin-6) as a lipid scramblase required for transport of these annexins to the outer leaflet of the plasma membrane. This work reveals a mechanism for annexin translocation across membranes which depends on plasma membrane phospholipid flipping. This translocation process may be relevant to many other phospholipid binding proteins in diverse membrane compartments.

and A5 across membranes in cells and in liposomes. The lantibiotic, cinnamycin, which flips lipids 11,12 , facilitates the transport of annexin A2 and A5 to the cell surface in many types of human cells. Furthermore, we identified TMEM16F (anoctamin-6) as a lipid scramblase required for transport of these annexins to the outer leaflet of the plasma membrane. This work reveals a mechanism for annexin translocation across membranes which depends on plasma membrane phospholipid flipping. This translocation process may be relevant to many other phospholipid binding proteins in diverse membrane compartments.

Main
To assess the role of lipid remodelling in annexin transport across membranes, we studied the effect of the lipid flipping toxin, cinnamycin, in mammalian cells 11,12 . Cinnamycin is a 19-amino acid lantibiotic that flips phosphatidylethanolamine (PE) and phosphatidylserine (PS) from one leaflet of the lipid bilayer to the other, both in liposomes and cell membranes 11,12 . We confirmed that cinnamycin can flip lipids in HeLa cells, as measured by the amount of PE and PS on the cell surface determined by flow cytometry (Fig. 1a). We next treated HeLa cells with cinnamycin for 30 min in serum-free medium, washed the cells with serum-free medium and incubated the cells with EDTA to dissociate all available cell surface annexin A2 and A5 13 . Cinnamycin treatment dramatically increased the amount of annexin A2 and A5 in the EDTA eluate without impacting cell morphology or viability, as shown by microscopy and lactate dehydrogenase activity in the eluate fraction, respectively ( Fig. 1b and Extended Data Fig. 1a-b). As a negative control, no annexin A2 and A5 was detected in the eluate when cells were incubated in serum-free medium without EDTA (Fig. 1b). The increase in annexin A2 and A5 on the cell surface in cinnamycin-treated cells was specific, as no cytosolic proteins or transmembrane proteins such as actin, Arf1, Arf6 and transferrin receptor were detected in the eluate fractions ( Fig. 1b and Extended Data Fig. 1c). When cinnamycin was used at concentrations that compromised cell membrane integrity, we observed actin release in the eluate fraction (Extended Data Fig. 1d), suggesting that actin in the eluate correlated with cell lysis. Mass spectrometry confirmed that cinnamycin stimulated translocation of annexin A1, A2, A3, A4 and A5 to the cell surface (Fig. 1c). This phenomenon was not limited to HeLa cells and could be demonstrated in several other lines (Extended Data Fig. 1e). Furthermore, mastoparan X, another lipid-flipping toxin 14 , caused a similar increase in annexin A2 detectable on the cell surface (Extended Data Fig. 1f). 3 To assess the requirement for lipid flipping activity in annexin translocation to the cell surface, we conducted the assay at 4°C. At this temperature, cinnamycin binds to the membrane but lipid flipping is abrogated 11 . Cinnamycin-mediated annexin A2 translocation to the cell surface was inhibited at 4°C (Extended Data Fig. 1g). To evaluate the importance of annexin A2 lipid binding in the transport process, we analysed the translocation of an annexin A2 mutant (Y23A) which is defective in lipid binding and has a defect in cell surface localisation 1 . Cinnamycin was unable to facilitate the translocation of the annexin A2 lipid binding mutant ( Fig. 1d). Together, these data support a mechanism whereby annexin A2 and A5 are transported to the cell surface by first binding to lipids on the inner leaflet of the membrane before being translocated across to the cell surface during lipid remodelling.
To evaluate whether lipid flipping is the minimal requirement for annexin transport across membranes, we developed an in-vitro liposome system using recombinant annexin A5 and cinnamycin. We first confirmed that cinnamycin could flip lipids in liposomes using an assay based on the quenching of NBD-PE by dithionite (Extended Data Fig. 2). Next, liposome binding and sedimentation experiments showed that annexin A5 binds to phosphatidylcholine:phosphatidylethanolamine (PC:PE) liposomes in the presence of calcium and can be removed from the membrane by the calcium chelator EGTA (Fig. 2a, lane 1 vs 3).
Interestingly, pre-treatment with cinnamycin increased the EGTA-resistant annexin A5 fraction (Fig. 2a, lane 4). This suggests that annexin A5 was translocated from the outer leaflet of the liposome membrane (surface of the liposome) to the inner leaflet or lumen of the liposome where it was protected from removal by EGTA. To confirm this phenomenon, we used a proteinase K protection assay. Proteinase K is very efficient in cleaving exposed proteins from membrane surfaces 15 . PC:PE liposomes were pre-incubated with annexin A5 before incubation with proteinase K, which resulted in digestion of free and surface-bound annexin A5 detected by a dramatic decrease in amount of full-length annexin A5 seen by western blot (Fig. 2b, lane 1 vs 3). A small fraction of full-length annexin A5 and a near full-length cleavage product was also detected; this corresponded to fully-protected and partially membraneinserted annexin A5, respectively (Fig. 2b, lane 3). Protease protection was membranedependent, as all available annexin A5 was degraded in the presence of membrane-solubilising detergent (triton) (Fig. 2b, lane 4). Importantly, cinnamycin pre-treatment increased the levels of the full-length annexin A5 and the near full-length cleavage product detectable after proteinase K digestion (Fig. 2b, lane 5). Therefore, cinnamycin increased the proportion of annexin A5 protected from proteinase K degradation, in agreement with the EGTA assay. 4 A similar protection assay was used in combination with a fluorescence quenching assay, allowing a real-time readout for degradation. In this assay, annexin A5 was labelled with fluorescein isothiocyanate (FITC) molecules. At high density, the fluorescence of the FITC molecules is quenched due to their close proximity on annexin A5. When proteinase K is added, accessible FITC-annexin A5 is cleaved, allowing the FITC molecules to be spatially separated, triggering an increase in their intrinsic fluorescence (dequenching). Proteinase K-induced FITC-annexin A5 dequenching was unaffected when incubated with cinnamycin then triton without liposomes, or when incubated with PC-only liposomes (Fig. 2c (i-ii)), as annexin A5 does not bind PC alone 16 . By contrast, proteinase K-induced FITC-annexin A5 dequenching was reduced when PE-containing liposomes were used (Fig. 2c(iii)). This is consistent with results from the previous protection assay, indicating that there is a fraction of annexin A5 protected from protease cleavage ( Fig. 2c(iii)). Furthermore, in agreement with previous results, PE-containing liposomes pre-treated with cinnamycin exhibited further attenuated FITC dequenching, compared to control/DMSO ( Fig. 2c(iii)). Finally, the addition of triton completely recovered FITC fluorescence in both DMSO and cinnamycin conditions (Fig. 2c (iii)). Taken together, these three approaches suggest that lipid flipping induced by cinnamycin facilitates annexin A5 translocation across liposome membranes.
Given that cinnamycin lipid flipping activity is sufficient to translocate annexin across membranes in cells and liposomes, we looked for a mammalian protein that could drive this process in cells. Plasma membrane lipid asymmetry is maintained by transmembrane proteins that flip lipids from the inner leaflet to the outer leaflet and vice versa 17,18,19 . One family of proteins with lipid flipping activity are the scramblases 20 . Scramblase activity is calciumdependent and energy-independent [19][20][21] . Scramblase dysfunction results in Scott's syndrome, a mild bleeding disorder resulting from a lack of phosphatidylserine externalisation 22 . Scott's syndrome has been attributed to mutations in the phospholipid scramblase known as TMEM16F (also called anotamin-6) 21 . Due to the importance of TMEM16F lipid flipping activity in vivo, we investigated a role for TMEM16F in the translocation of annexin A2 and A5 to the cell surface. Clonal TMEM16F HeLa knockout cell lines were generated using CRISPR/Cas9 and matched wild-type controls with no targeting were also isolated (Extended Data Fig. 3a, b). We confirmed gene targeting using sequencing and qPCR (Extended Data   Fig. 3c). We were unable to measure the protein level of TMEM16F as the few antibodies that we tried did not show specific signals on western blot (data not shown). However, we confirmed a functional defect in lipid flipping in the TMEM16F knockout cell lines by 5 measuring the level of PS on cell surface in cells challenged with ionomycin ( Fig. 3a(i)). We treated TMEM16F wild-type or -deficient cells with ionomycin for 10 min at 37°C in the presence of annexin A5-Cy5 and propidium iodide. Cell surface PS (and PE) were analysed by measuring the amount of annexin A5-Cy5 bound to live cells by flow cytometry. TMEM16Fdeficient cells were unable to externalise PS in response to an increase in intracellular calcium, whereas wild-type cells and positive matched non-targeted controls showed an increase in the amount of PS on the cell surface ( Fig. 3a(i) and Extended Data Fig. 4a). This confirms that TMEM16F activity is abolished in TMEM16F knockout cells. Interestingly, the level of PS (and PE) on the cell surface of TMEM16F deficient cells under unstimulated conditions was also slightly reduced compared to the wild-type and untargeted controls ( Fig. 3a(ii) and Extended Data Fig. 4b), showing for the first time that TMEM16F is constitutively active.
To investigate the role of TMEM16F in translocation of annexins, we assessed the level of annexin A2 and A5 on the cell surface of TMEM16F deficient cells. Cells were treated with EDTA to release annexin A2 and A5 and the eluate was assessed by western blot, as described earlier. Strikingly, in both TMEM16F-deficient clones, annexin A2 and A5 were severely reduced in the EDTA eluate (Fig. 3b). Annexin A2 and A5 were not detected in the serum-free medium (SFM) eluate, and actin and LAMP-2 were absent from all eluates (SFM or EDTA) indicating this was a specific process (Fig. 3b). To ensure that the lack of annexin on the cell surface was due to reduced translocation rather than reduced retention at the cell surface, we assessed the level of annexin released into the supernatant over 24 hours. No annexin A2 and A5 were detected in the supernatant from wild-type or TMEM16F-deficient cells (Extended Data Fig. 4c). This demonstrated that annexin A2 and A5 were not translocated and thus present in the medium in TMEM16F-deficient cells (Extended Data Fig. 4c), whereas annexin A2 and A5 are translocated across the membrane in wild-type cells and detectable on the cell surface ( Fig. 3b). The lack of annexin A2 and A5 on the cell surface in TMEM16F-deficient cells is not due to off-target effects as the phenotype is consistent across both clones targeted with different sgRNAs. Furthermore, the phenotype was normalised when the TMEM16F-deficient cells were reconstituted with mouse, mCherry-tagged TMEM16F (mCherry-mTMEM16F) via lentiviral transduction that restored lipid flipping activity (Fig. 3c, d and Extended Data 5a, b and c). This established that TMEM16F is required for transport of annexin to the cell surface.
Our data indicate that cinnamycin stimulates lipid flipping in a manner comparable to TMEM16F, therefore, we set out to determine if its activity could substitute for TMEM16F.
Cells were treated with cinnamycin and the level of PS and PE externalisation was assessed by 6 annexin A5-Cy5 binding to live cells by flow cytometry. Cinnamycin stimulated lipid flipping in both wild-type and TMEM16F-deficient cells ( Fig. 4a and Extended Data Fig. 6a, b). Therefore, cinnamycin can be used as a surrogate for TMEM16F lipid flipping activity.
TMEM16F-deficient cells treated with DMSO showed reduced annexin A2 and A5 in the EDTA eluate (Fig. 4b), however when treated with cinnamycin there was a significant increase in the amounts of annexin A2 and A5 detectable in the EDTA eluate in these cells (Fig. 4b).
Cinnamycin also increased the level of annexin A2 and A5 on the cell surface in wild-type HeLa, as expected (Fig. 4b). This demonstrates that lipid flipping activity is sufficient for translocation of annexin A2 and A5 from the cytosol to the cell surface and was not due to a lack of annexin retention at the cell surface as annexin A2 was not present in the medium during cinnamycin treatment as described earlier (Extended Data Fig. 6c).
In summary, we have identified a mechanism for membrane translocation of annexins which is mediated by lipid flipping/remodelling. This process provides an explanation for the longstanding mystery of how some cytoplasmic proteins reach the cell surface independent of conventional secretion or described modes of unconventional secretion 23-26 . In cells, this lipid flipping-dependent translocation is primarily mediated by TMEM16F and enables transport of a protein from the inner intracytoplasmic domain of the plasma membrane to the exterior outer leaflet. This transport is ATP-independent and calcium-dependent and does not require the complex translocon-type machinery as seen for conventional trafficking of proteins into the ER or mitochondria. 10004974). Oligonucleotides for TMEM16F CRISPR targeting and sequencing were synthesized from Sigma-Aldrich (Extended Table 1).

Lentiviral transfection
HEK293FT packaging cells growing in 10-cm dishes were transfected with a mix of 11.68 µg packaging vector (psPAX2), 5.84 µg envelope vector (pMD2.G) and 18.25 µg ANO6-Plvx-mCherry-c1 vector. PEI (polyethylenimine) was used as transfection reagent. 48h after transfection, cell culture medium was collected and replaced by fresh medium; this step was repeated 2 times at intervals of 24 h. Virus preparations were then combined. Viral particles were added to cells, spin at 1,000g for 30 min and incubated overnight. After 24 h, medium was replaced by full medium and cells were incubated for 5 more days. Transduced cells were selected with puromycin and sorted to enrich mCherry-expressing cells.

Mass spectrometry
Samples were submitted to the Cambridge Institute for Medical Research-Institute of Metabolic Science proteomics facility where they were analyzed using Thermo Orbitrap Q Exactive with EASY-spray source and Dionex RSLC 3000 UPLC.  (table S1). PCR products were sequenced by Sanger sequencing and insertions and deletions analysed by using the Tracking of Indels by DEcomposition (TIDE) web tool 29 . Additionally, to analyse insertions and deletions larger than 50 base pairs, the R code was kindly provided by Prof van Steensel. TIDE analysis showed that the expected region had been targeted and each knockout clone was devoid of wild type TMEM16F DNA ( Fig. S3A and B). Clonal wild type control lines that had been through transfection, selection and single cell cloning steps, but had not efficiently targeted TMEM16F, were used as matched positive controls for each exon ( Fig. S3A and B). Due to the lack of antibodies specific for TMEM16F we were unable to analyse expression at the protein level, therefore, we assessed the mRNA levels which were reduced in TMEM16F knockout cells (Fig. S3C).

Ionomycin and cinnamycin PS flow cytometry assay
Approximately 1x10

Statistical analysis
Significance levels for comparisons between groups were determined with t-tests.          only NBD in the outer membrane leaflet is accessible to react with dithionite, a partial decrease in fluoresce is observed. An additional decrease in fluorescence was detected when membranesolubilising detergent (triton 0.5%) was added, which exposes the NBD in the inner leaflet.
Cinnamycin-treated LUV were more susceptible to dithionite-mediated NBD reduction, since some of the NBD-PE in the inner leaflet (which was previously inaccessible to dithionite) was