How target cell stiffness controls immune cell nibbling behavior

To study trogocytosis by macrophages, we developed a fluorescence-based assay that uses the key features of trogocytosis (surface engulfment) and phagocytosis (surface and volume engulfment) to separate the two processes. For a volume marker, we labelled the cytosol of target cells with the pH-sensitive rhodamine derivative pHrodo, which selectively fluoresces in low pH environments (for example, the phagolysosome). For a surface marker, we used a fluorescent antibody, AlexaFluor647 anti-CD47 IgG, that also activates macrophages through engagement with their Fcγ receptor (FcγR). CD47 is a marker of self and a potent ‘don’t eat me’ inhibitory signal against macrophage phagocytosis, and antibody labelling attenuates its inhibitory effects. Macrophages that only trogocytose would only be marked by an AlexaFluor647 signal, while those that phagocytose (as well as those that phagocytose and trogocytose) would be marked by both an internalized pHrodo signal and AlexaFluor647 signal.

When we mix target Jurkat T cells with RAW 264.7 macrophage-like cells, we observe two clear phenotypes by fluorescence microscopy. Some macrophages fully engulf the labelled Jurkat T cells, with surface antibody signals and volume signals observed to be colocalized in the phagolysosome of the macrophage, which are labelled with CellTracker Green CMFDA (Fig. 1a, bottom). Other macrophages internalize small ‘bites’ from the surface of Jurkat T cells that are positive for the anti-CD47 antibody surface signal but negative for the volume marker, pHrodo (Fig. 1a, top). From fluorescence microscopy images of macrophages incubated with opsonized Jurkat T cells for 1 h, we found that a large proportion of macrophages nibbled target cells, while full phagocytosis events were quite rare (Extended Data Fig. 1).

To test this observation with larger populations, we developed a flow cytometry assay to quantify trogocytic and phagocytic efficiency. We mixed cytosol-labelled macrophages with surface- and volume-labelled target cells and analysed the populations of macrophages that contained both target surface and volume signals (indicating phagocytosis), only the surface signal (indicating trogocytosis only) or neither signal. We define phagocytic and trogocytic efficiency as the percentage of the total number of macrophages that have undergone phagocytosis or trogocytosis, respectively.

Consistent with the microscopy experiments, we find that trogocytic events are notably more frequent than phagocytic events when target cells are opsonized with anti-CD47 (Fig. 1b). We estimate that trogocytosis is 3–4× more frequent than phagocytosis in our 1-h experiments with a 1:1 ratio of macrophages to target cells.

To test whether trogocytosis was dependent on CD47 labelling or could occur with labelling of other cell surface proteins, we non-specifically biotinylated the primary amines of proteins on the surface of target cells using an NHS-ester biotin compound. We then opsonized target cells with anti-biotin, labelled them and mixed them with macrophages. We found that trogocytic efficiency was nearly identical to that for the anti-CD47 labelling at the same antibody concentration, indicating that trogocytosis is not dependent on the surface molecule labelled (Fig. 1b).

To ensure that the trogocytosis we measure is dependent on FcγR binding to the antibody, we measured the trogocytic and phagocytic efficiency of macrophages first treated with an FcγR-blocking antibody (mouse anti-CD16/32). Both trogocytosis and phagocytosis are inhibited when the macrophage FcγR is blocked by 0.04 μM antibody (Extended Data Fig. 2a,b). Similarly, we observed that biotinylated, but non-opsonized, target cells with AlexaFluor647 streptavidin-labelled membranes are not trogocytosed or phagocytosed, confirming that trogocytosis is dependent on FcγR engagement (Extended Data Fig. 2c).

We next explored whether the trogocytosis we observe is unique to macrophage-like RAW 264.7 cells, which are convenient to work with but have known differences from primary mouse macrophages and other macrophage cell lines^18,19. We quantified the trogocytic and phagocytic efficiency of opsonized Jurkat T cells incubated with primary bone marrow-derived macrophages (BMDMs), lipopolysaccharide (LPS)-stimulated RAW 264.7 cells and another immortalized mouse macrophage cell line, J774a.1 cells. We observed a similar trogocytic efficiency across all four macrophage cell lines (Fig. 1c). This shows that both the ability to trogocytose and the trogocytic efficiency are not strongly affected by macrophage cell type.

While different macrophages trogocytose the same target cell with similar efficiency, does the same macrophage trogocytose different target cells with similar efficiency? To test this, we used a panel of cultured cells derived from blood cancers, namely, Jurkat T cells, HL60 cells and Raji B cells. We opsonized the target cells with anti-CD47, which is highly expressed on all three cell types (Extended Data Fig. 3), and incubated them with RAW 264.7 macrophages in separate experiments. Interestingly, we see large differences in trogocytic and phagocytic efficiency for the different target cell lines, with the trogocytic efficiency of HL60 and Raji B cells at only ~20%, compared with ~70% for Jurkat cells (Fig. 2a). Notably, this is also the case for LPS-stimulated macrophages, especially for the filamin-deficient melanoma cell line M2. Strikingly, while LPS-stimulated macrophages phagocytose HL60 and Jurkat cells at a statistically similar amount, the ratio of trogocytosis to phagocytosis is 2× higher for Jurkat cells than for HL60 cells.

As Jurkat T cells have approximately 2.2× more CD47 on their cell surface than HL60 and Raji B cells (Extended Data Fig. 3), differences in trogocytic efficiency could be due to differences in antibody surface density. We altered antibody solution concentrations to achieve the same surface density of anti-CD47 across cell types (Fig. 2b), and we tested non-specific labelling of the target cells by biotinylating and opsonizing them with anti-biotin (Fig. 2c). When antibody surface density is matched, we observe that trogocytic efficiency is the highest in Jurkat cells, while phagocytic efficiency is the highest in HL60 cells. To ensure that our results are not simply from blocking the ‘don’t eat me’ interaction of target CD47 with macrophage SIRPα (due to anti-CD47 or biotinylation of CD47), we co-incubated macrophages with Raji B cells opsonized with anti-CD19. We observed that Raji B cells opsonized with anti-CD19 show ~4× more trogocytosis than phagocytosis, which is comparable to the results shown in Fig. 2b, albeit with an overall 2× lesser magnitude (Extended Data Fig. 4).

As trogocytosis involves pinching off small (<1 μm) bits of membrane, one potentially relevant difference between the target cells could be the deformability of their cell surface. This deformability is dependent not only on membrane tension but also on the strength and abundance of linkages between the membrane and underlying cytoskeleton. Together, the resistance to deformation can be captured by ‘cortical tension’, often measured with micropipette aspiration^20–22. We quantified the cortical tension of antibody-labelled target cells using a home-built micropipette aspiration system with an in-line pressure sensor. Using glass pipettes with an inner diameter of ~5 μm, we applied suction to individual cells until a ‘tongue’ was pulled into the syringe to a consistent length of ~1.5 μm (Fig. 2d). We then quantified cortical tension of the cell using equation (1), the Young–Laplace law

where ΔP_suction, R_pipette and R_cell are the suction pressure, the pipette radius and the cell radius, respectively.

We find that the cortical tension of HL60 cells, which are phagocytosed more than the other cells, is notably higher than that of the Jurkat T cells and Raji B cells (Fig. 2e). If we take the average trogocytic efficiency of target cells with matched anti-CD47 surface densities, the trogocytic efficiency for target cells decreases (Pearson correlation coefficient of −0.71) with increasing cell tension (Fig. 2f).

The cortical tension we measured for cells is a combination of membrane tension and interactions with the cortical cytoskeleton^23,24. We wondered whether membrane tension alone could be used to alter macrophage trogocytosis, independent of any cytoskeletal interactions. To test this, we created target particles with only membrane using GUVs composed of palmitoyl oleoyl phosphatidylcholine (POPC), 1 mol% biotin dioleoyl phosphatidylethanolamine (biotin-DOPE) and 0.5 mol% lissamine rhodamine phosphatidylethanolamine (liss rho PE). The GUVs were formed in a sucrose medium that was osmotically tuned using an osmometer to cause the GUVs to sink but remain isotonic in the cell culture media. To obtain GUVs of a consistent size and avoid the confounding effects of small vesicles formed at the same time as the GUVs, we developed a home-built differential density column (SeparatorMax 3000; Extended Data Fig. 5) to separate the GUVs on the basis of size. We collected fractions of GUVs in the size range of 5–20 μm in diameter and opsonized them with fluorescent anti-biotin at a surface density of ~400 antibodies per μm (ref. ²) for a 10-μm GUV.

We incubated opsonized GUVs with macrophages and observed by confocal microscopy that macrophages can both fully engulf GUVs and partially nibble them (Fig. 3a). When a GUV is trogocytosed, we observe colocalization of the lipid dye and the fluorescent antibody in trogocytic bites, indicating that the macrophages extract small patches of membrane from GUVs rather than detach the anti-biotin AF647 antibody. These initial experiments showed that antibody-opsonized GUVs can be both trogocytosed and phagocytosed, similar to target cells.

To quantify the trogocytic and phagocytic efficiency of macrophages incubated with GUVs, we loaded dark GUVs with a soluble rhodamine fluorophore and opsonized them with AlexaFluor647 anti-biotin. After 30 min of co-incubation with macrophages, we analysed the macrophages via flow cytometry and found that trogocytosis is a much more common event than phagocytosis, consistent with the target cells tested (Fig. 3b). We might have expected that all GUVs would either by trogocytosed or phagocytosed because they all have the same lipid composition and vary only minimally in size. However, there must be a difference—even in this minimal system—that it is sufficient to drive trogocytosis rather than phagocytosis.

Interestingly, GUVs created via electroformation are formed with a wide range of membrane tensions, even when they are ‘isotonic’ (Extended Data Fig. 6) due to the random formation of membrane pores²⁵. To directly investigate the role of membrane tension in trogocytosis, we first used micropipette aspiration to measure the membrane tension of a population of GUVs. We incubated GUVs with macrophages in an imaging dish, allowed GUVs to settle and interact with the macrophages for 10 min; thereafter, we individually measured the membrane tension of GUVs that were being either phagocytosed or trogocytosed. We identified phagocytic events as GUVs whose surface was wrapped >50% by macrophage phagocytic extensions and trogocytic events as GUVs connected to macrophages in which small ‘bites’ were internalized in the macrophage. We chose >50% wrapping as the threshold for phagocytosis because the micropipette occupies one side of the GUV and because we and others²⁶ have found that >50% coverage leads to full engulfment >75% of the time. We found two distinct tension regimes: At tensions of 0.26 ± 0.33 mN m⁻¹, GUVs are exclusively trogocytosed, while at tensions of 11.7 ± 11.2 mN m⁻¹, GUVs are phagocytosed (Fig. 3c).

These measurements show that membrane tension is a sufficient signal to drive macrophage trogocytosis or phagocytosis of GUVs. However, is that decision made at initial contact and activation, or does the macrophage dynamically sense membrane tension and alter its decision? To address this, we tested the response of macrophages to dynamic changes in membrane tension with the micropipette aspiration system and simultaneous confocal microscopy. We added opsonized GUVs to one side of an imaging chamber and selected a GUV with a low membrane tension (0.27 mN m⁻¹, in the trogocytosis regime). Upon presenting the GUV to a macrophage, we observed the macrophage forming filopodial extensions but then quickly retracting and beginning to internalize small ‘bites’ of a GUV membrane. Using the micropipette, we then applied suction to the GUV, increasing membrane tension above the threshold for phagocytosis (3.68 mN m⁻¹, in the phagocytosis regime). Within 3 min, the macrophage rapidly extended a phagocytic cup around the GUV and engulfed it completely within a matter of minutes (Fig. 3d,e and Supplementary Video 1; see Supplementary Videos 2–4 for more examples). These experiments show that membrane tension alone is a significant driver of trogocytosis in macrophages, and changes in membrane tension can act as a switch between phagocytosis and trogocytosis.

Our experiments with cultured cells and GUVs indicate that the deformability of the target surface is sufficient to control whether the target is phagocytosed or trogocytosed. This raises the possibility that intentionally stiffening the cortex of cells could suppress trogocytosis and promote phagocytosis. Interestingly, cells that have undergone apoptosis are significantly stiffer (up to 3×) than their healthy counterparts^27–29. Macrophages are responsible for efferocytosis, or the phagocytic clearance of apoptotic cells. We hypothesized that the increase in stiffness upon apoptosis would bias macrophages towards phagocytosis over trogocytosis.

To induce apoptosis in Jurkat T cells, we used raptinal, a potent activator of caspase-3, which induces apoptosis within 30 min of incubation, resulting in >80% cell death (as measured by propidium iodide) within 2 h (refs. ^30,31). In the presence of 10 μM raptinal for 2 h, we measured 85% of Jurkat T cells dead as marked by propidium iodide (in contrast to 1% of untreated Jurkat T cells). After co-incubation with LPS-stimulated macrophages, we saw a 3× increase in phagocytosis (and a 15× decrease in trogocytosis) over healthy cells when Jurkat cells were opsonized with IgG (Fig. 4a, left). Because apoptotic cells have additional ‘eat me’ signals (for example, phosphatidylserine lipids flipped to the outer leaflet of the plasma membrane), we repeated the experiment without IgG opsonization (Fig. 4a, right). To measure trogocytosis without a fluorescent IgG, we biotinylated the target cells and then labelled the membrane with AlexaFluor647 streptavidin (and dark anti-biotin for opsonized cells). In Fig. 4a (right), phagocytosis and trogocytosis in both raptinal-treated and -untreated Jurkat cells are attenuated; however, phagocytosis is still higher for raptinal-treated Jurkat cells.

To further investigate the role of cortical tension in trogocytosis of living cells, we co-incubated LPS-stimulated macrophages with a melanoma-derived cell type, M2. M2 cells are deficient in filamin A, a crosslinker of F-actin at the cortex. The cortex of M2 cells is weakened and, as a result, they are measurably softer than similar cells expressing filamin A^32,33. In comparison with Jurkat T cells and HL60 cells, M2 cells opsonized at the same surface density of anti-CD47 are trogocytosed significantly more (Fig. 4c), consistent with our findings that lower cortical tension results in greater trogocytosis.

We next sought to test whether controlled stiffening of the cortex would allow us to titrate the amount of phagocytosis we observed. Previous work has shown that gentle glutaraldehyde fixation leads to measurable differences in cortical tension³⁴, with concentrations of glutaraldehyde below 0.01% leaving cells viable. We treated opsonized target cells with either 0%, 0.002% or 0.0025% glutaraldehyde for 30 min. At these fixation concentrations, anti-CD47 density on treated target cells is approximately the same as untreated cells (300–500 ± 100 antibodies per μm²).

To confirm the effects of gentle fixation on cortical tension, we measured treated target cells using micropipette aspiration. We chose HL60 and Jurkat cells because they engage in the least and most trogocytosis, respectively. Cortical tensions for treated cells increased above the tension regime we measured for trogocytosis of GUVs, especially for Jurkat and HL60 cells treated with 0.0025% glutaraldehyde (Fig. 4d). To quantify the trogocytic efficiency of macrophages with the opsonized and stiffened target cells, we washed and incubated them with macrophages for 1 h and then analysed them with flow cytometry. We found that trogocytosis was almost eliminated for cells treated with 0.0025% glutaraldehyde. By contrast, phagocytosis increased significantly for cells treated with 0.0025% glutaraldehyde (Fig. 4e).

As we previously observed that Jurkat cells, before antibody surface density matching with HL60 and Raji B cells, had greater trogocytic efficiency with greater antibody coverage (2.2× higher than HL60 and Raji B cells), we wondered whether we could rescue trogocytosis of the stiffened cells by increasing antibody surface density. When we titrated antibody surface density for gently fixed Jurkat and HL60 cells, we found that trogocytic efficiency was indeed increased at the highest antibody densities (Fig. 4f). Interestingly, trogocytic efficiency is rescued at much higher concentrations of antibody than for wild type target cells. Not only does it take more antibody to induce trogocytosis of gently fixed cells, but the trogocytic efficiency is never greater than what was achieved for wild type target cells at lower antibody densities (Extended Data Fig. 7). This suggests that we are unable to coat the surface sufficiently to rescue the full trogocytic efficiency of a soft wild type target.

Our experiments indicate that both cortical tension and antibody density play important roles in driving trogocytosis of cellular targets. How might the two interact to control interface mechanics between the macrophage and target cell? To gain intuition, we develop a simple scaling relationship for the interfacial mechanics. We assume the target antibody binds to macrophage FcγRs at a high-enough concentration to trigger downstream signalling machinery that leads to cytoskeleton-mediated active stresses at the interface (Fig. 5a). On the basis of prior experimental measurements, these stresses take the form of both extensile and compressive normal stresses on the target membrane³⁵. We further assume that these stresses scale with the antibody density on the target (Supplementary Note). These normal stresses will tend to deform the interface, and this deformation is resisted by the target cortical tension. Balancing the forces due to active stresses and cortical tension at the interface, one gets equation (2)

This gives the length scale equation () 3

where γ_t is the target cortical tension and σ_normal(ρ_AB) is the antibody-dependent normal stress at the interface. This scaling corresponds to the minimum length scale of local membrane deformation or bites because below this scale, tension dominates the effects of active stresses and damps out active fluctuations. A similar length scale can be derived by considering the balance between active stresses and membrane bending (Supplementary Note), which is typically smaller than R_min for physiologically relevant values of tension and membrane bending stiffness.

Importantly, this scaling implies that increasing antibody density or decreasing target membrane tension has an equivalent effect on setting the minimum length scale of target membrane deformations. For small antibody density or large tensions, the active stresses are weak compared with those due to tension, and the minimum scale of deformation due to active stress becomes comparable or larger than the interface size R (R_min » R). In this case, the macrophage is incapable of pinching off bites smaller than the interface and would completely engulf the target (that is, phagocytosis). By contrast, for small tensions or large antibody densities, the active stresses can locally overwhelm membrane tension and pinch off bites smaller than the interface size (R_min « R), leading to trogocytosis. Below a second, lower critical antibody density, one expects no engagement, and neither behaviour is observed. Thus, one can also interpret this scaling as a critical antibody density or tension that separates the distinct behaviours of phagocytosis and trogocytosis.

If we take our experimentally measured values for γ_t ≈ 10⁻² – 10⁻¹ mN m⁻¹ and the previously measured traction forces that give normal stresses of order 100 Pa (ref. ³⁵), our scaling predicts a deformation size R_min of 0.1–1 μm. Interestingly, this is consistent with the scale of trogocytic ‘bites’ measured through confocal microscopy, which are on the scale of 1 μm for macrophages incubated with Jurkat T cells and HL60 cells (Fig. 5b). This suggests that the macrophage can pinch off bites of a size close to the minimum scale of interfacial deformation. Our scaling analysis assumes smooth deformations in the membrane, which is controlled by the resistance of the membrane to stretching or bending. For sharp deformations, we expect the contribution from bending to dominate. However, to see sharp indentations below a radius of 0.1 μm would require local active stresses beyond what are expected physiologically (see Supplementary Note, equation (5)).

Our scaling also predicts that the critical antibody density that separates phagocytic and trogocytic behaviours in experiments must be proportional to the target cell cortical tension. To test this, we fit the sigmoidal curve below to the data in Fig. 4c and extracted the inflection point of the curve as the critical antibody density in equation (4), ρ_crit

For both HL60 and Jurkat cells, we find that the density of antibody necessary to initiate trogocytosis indeed increases with cortical tension (Fig. 5c). We assume that, while this scaling relationship is general, the difference in slope between HL60 and Jurkat cells is probably due to important differences in how the active stresses developed at the interface are related to the local antibody density. In this simple scaling, we do not consider the possible contributions from dynamic remodelling of the cytoskeleton, spatial heterogeneities due to lipid rafts, or receptor clustering and cooperative effects, which will probably have cell-type-specific differences owing to distinct surface proteomes and properties. For the parameters we measured, antibody density and cortical tension, our modelling is consistent with the positive correlation between the critical antibody density and cortical tension that we observe (Fig. 5c).

From the scaling relationship described above and assuming an interface size R of ~1 μm, we predict a tension-dependent switch in macrophage behaviour at a tension around 0.1–1 mN M⁻¹ for the range of normal stresses measured in experiments (50–150 Pa)³⁵. On the basis of this prediction, we can construct an approximate phase diagram that predicts the macrophage behaviour as a function of antibody density and cortical tension (Fig. 5d). Our experimental data for Jurkat and HL60 cell lines are quantitatively consistent with this phase diagram, with cortical tension predicted to have a more significant effect than antibody density.

Phosphocholine lipids (Avanti Polar Lipids) were used as purchased without further purification. Lipid stock solutions in chloroform contained a quaternary mixture of 97.5 mol% POPC, 1 mol% biotin-PE, 1 mol% PEG2K DSPE and 0.5 mol% lissamine rhodamine PE. GUVs are diluted in an ionic solution of phosphate-buffered saline (PBS), and all lipids in our mixtures are zwitterionic. We added PEG2K DSPE to block GUVs from aggregating in the charge-screened PBS solution.

Antibodies used to opsonize GUVs and cultured cells were purchased from Santa Cruz Biotechnology and used without further labelling or purification. Biotin and CD47 were bound by, respectively, AlexaFluor647-labelled anti-biotin mouse IgG (clone BK-1/39, Santa Cruz Biotechnologies) and AlexaFluor647-labelled anti-CD47 mouse IgG (B6H12, Santa Cruz Biotechnologies). The FcγRs on RAW cells were bound by anti-CD16/32 mouse IgG (Biolegend).

RAW 264.7 murine male macrophage-like cell line, J774A.1 murine female macrophage-like cell line, Jurkat T human cell line, Raji B human B cell line, M2 human melanoma cell line and HL60 human promyeloblast cell lines were obtained from the UC Berkeley Cell Culture Facility. RAW 264.7, J774A.1, Jurkat T, M2 and HL60 cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 media (Corning) supplemented with 10% heat-inactivated foetal bovine serum (HI-FBS, Thermo Fisher Scientific) and 1% penicillin–streptomycin (Thermo Fisher Scientific). Raji B cells were cultured in RPMI 1640 media supplemented with 10% HI-FBS, 1% sodium pyruvate (Thermo Fisher Scientific) and 1% penicillin–streptomycin. RAW 264.7 and J774A.1 cells were cultured in non-tissue culture-treated 10-cm dishes (VWR); all other cell lines were cultured in tissue culture-treated 10-cm dishes (Corning) at 37 °C with 5% CO₂. Pre-stimulated RAW 264.7 cells were incubated with 100 ng ml⁻¹ LPS (Sigma) for 24 h before coculture with target cells.

BMDMs from male C57BL/6 (B6) mice were a kind gift from the Portnoy Lab (UC Berkeley). BMDMs were grown in RPMI 1640 media supplemented with 10% HI-FBS and 1% penicillin–streptomycin at 37 °C with 5% CO₂. BMDMs were used in experiments within 24 h of thawing.

Solutions containing 0.25 mg total lipids were spread evenly on slides coated with indium tin oxide (70–100 Ω sq⁻¹; Sigma Aldrich). The slides were placed under vacuum for >30 min to permit complete evaporation of chloroform. A capacitor was created by sandwiching a 0.3-mm rubber septum between two lipid-coated slides. The gap was filled with ~200 μl of 285 mM sucrose (osmotically matched to PBS using an osmometer (Precision Systems)). GUVs 10–100 μm in diameter were electroformed⁵¹ by application of an a.c. voltage of 1.5 V at 10 Hz across the capacitor for 1 h at 55 °C.

Sucrose solutions containing electroformed GUVs were diluted in 1 ml of 285 mM glucose and injected into a home-built differential density column, lovingly named the SeparatorMax 3000 (Extended Data Fig. 5). Specifically, ten ports with an inner diameter of 2 cm, spaced 2.5 cm apart, were drilled into a cast acrylic tube of 33.6 cm. This tube was connected to two different baths containing a low-density glucose solution (285 mM) and a high-density glucose solution (300 mM) in series. Solutions were loaded into the column via a peristaltic pump with vigorous hand mixing. The resulting solution column is a linear gradient of glucose density from 300 mM glucose (bottom) to 285 mM glucose (top). Vesicles were allowed to sediment overnight (~12 h) and collected from each of the ten ports of the column. We kept the GUVs from ports 5–8 (labelling from the top to bottom), and they had diameters from 5 to 20 μm. We further purified GUVs by spinning the solution at 300g for 10 min and taking the bottom 100 μl of solution. This solution was then diluted in PBS containing 4 μM AlexaFluor647-labelled anti-biotin mouse IgG (clone BK-1/39, Santa Cruz Biotechnologies). GUVs used in flow cytometer-based phagocytosis and trogocytosis assays were loaded with 1 μM soluble rhodamine dye dissolved in 285 mM sucrose. GUVs were purified from excess dye in solution by running them through the SeparatorMax 3000.

All live cells were maintained at 37 °C and 5% CO₂ with a stage top incubator (Okolab) during imaging. For confocal microscopy, cells were imaged with a confocal microscope (Eclipse Ti, Nikon) with a spinning disk (Yokogawa CSU-X, Andor), sCMOS camera (Prime 95B, Photometrics) and 60× objective (Apo TIRF, 1.49NA, oil, Nikon). The spinning disk confocal microscope was controlled with Nikon Elements (Nikon). Images were analysed and prepared using Fiji (http://www.imagej.net/software/fiji↗).

Micropipettes were made from capillaries (1.0 mm outer diameter, 0.58 mm inner diameter, 100 mm length, borosilicate glass; G100-4, Harvard Apparatus) drawn out with a filament pipette puller (Sutter Instruments). Pipette tips were forged with an adapted microforge (MicroData Instruments) to obtain a smooth opening of ~5 μm in diameter. Pipettes were subsequently filled with 0.45 μm filtered PBS using a home-built syringe-pulling device, named The Siphonator 3000, for >30 min. Pipettes were treated for 15 min in a 0.45 μm filtered 10% bovine serum albumin solution before each experiment to passivate the pipette surface and prevent unwanted adhesion of GUVs and target cells to the pipette glass.

The treated pipettes were inserted into a TransferMan 4r micromanipulator (Eppendorf) mounted on a confocal microscope (Eclipse Ti2) to facilitate the manipulation of target cells and/or GUVs for surface tension measurements. To provide controlled suction, the pipettes were connected to a syringe pump (CellTram, Eppendorf) and a pressure sensor (DP103, Validyne Engineering), which measured the suction pressure applied to the pipette.

To measure surface tension, the pipette was aligned to the target interface and increasing suction pressure was applied until a membrane tube with a length equal to the pipette radius was pulled into the pipette. Using Laplace’s law, the measured suction pressure (ΔP_suction) was used to measure the interface tension (γ)

where ΔP_suction, R_pipette and R_cell are the suction pressure, the pipette radius and the cell radius, respectively. The interface tensions of GUVs and target cells were modulated by adjusting the suction pressure applied through the same pipettes, thereby varying the interfacial tension.

All experiments were independently repeated at least three times with similar results. Sample sizes and statistical tests are included in the figure legends wherever appropriate. No statistical methods were used to pre-determine sample sizes. No randomization was used to collect the data. Data distribution was assumed to be normal, but this was not formally tested. Data collection and analysis were not performed in a blinded manner, and no data were excluded from analyses.

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

Nature Cell Biology thanks Zheng Liu and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

is available for this paper at 10.1038/s41556-025-01807-6.

The online version contains supplementary material available at 10.1038/s41556-025-01807-6.

Datasets used to generate figures in the paper are available via the Open Science Framework at 10.17605/OSF.IO/6FP4V. All other data supporting the findings of this study are available from the corresponding author on reasonable request.are provided with this paper. Source data

Original code to generate each of the subfigures (and uploaded as individual scripts) in the paper is available via GitHub at https://github.com/fletchlab-git/Trogocytosis↗.