Natural killer cell immune synapse formation and cytotoxicity are controlled by tension of the target interface

ABSTRACT Natural killer (NK) cells can kill infected or transformed cells via a lytic immune synapse. Diseased cells may exhibit altered mechanical properties but how this impacts NK cell responsiveness is unknown. We report that human NK cells were stimulated more effectively to secrete granzymes A and B, FasL (also known as FasLG), granulysin and IFNγ, by stiff (142 kPa) compared to soft (1 kPa) planar substrates. To create surrogate spherical targets of defined stiffness, sodium alginate was used to synthesise soft (9 kPa), medium (34 kPa) or stiff (254 kPa) cell-sized beads, coated with antibodies against activating receptor NKp30 (also known as NCR3) and the integrin LFA-1 (also known as ITGAL). Against stiff beads, NK cells showed increased degranulation. Polarisation of the microtubule-organising centre and lytic granules were impaired against soft targets, which instead resulted in the formation of unstable kinapses. Thus, by varying target stiffness to characterise the mechanosensitivity of immune synapses, we identify soft targets as a blind spot in NK cell recognition. This article has an associated First Person interview with the co-first authors of the paper.


Introduction
CD8 + T cells has been shown to rely on the local application of force adjacent to membrane adhesion sites (Andzelm et al., 2007;Basu et al., 2016). However, these data have relied on chemical manipulations of cells or used planar substrates with varying stiffness; there has not been a direct test of how immune cell responses vary against 3D spherical stimulants with defined stiffness.
During their lifetime, NK cells infiltrate a variety of tissues with differing mechanical properties, exposing them to large variations in matrix stiffness in the process. Atomic force microscopy (AFM) experiments have revealed that lung tissue is in the range of 2 kPa (Wells, 2013), muscle 10 kPa and cartilage 30 kPa, whilst cells developing in the bone marrow are exposed to substrate rigidities in the range of around 25-100 kPa (Jansen et al., 2015;Discher et al., 2009). Inflamed and diseased tissues also present with chronically altered tissue rigidities (Suresh, 2007;Previtera and Sengupta, 2015).
These changes depend on several factors including dysregulated vasculature and variation in the extracellular matrix composition (Padera et al., 2004;Muiznieks and Keeley, 2013). Fibrosis is a key component of multiple diseases affecting millions of people including liver cirrhosis, pulmonary fibrosis and arthritis (Friedman et al., 2013). Importantly, AFM measurements on fibrotic tissue reveal a characteristic tissue stiffening with elastic modulus values of 17 kPa, compared to 2 kPa for normal tissue (Lupher and Gallatin, 2006;Wells, 2013). Liver scarring and fibrosis are strongly coupled with hepatocyte death, and NK cell mediated killing of hepatocytes has been reported (Ochi et al., 2004;Ishiyama et al., 2006) Alongside changes to rigidity of the surrounding matrix, cells themselves exhibit a variety of stiffnesses ranging from 0.2 kPa to 10 kPa (Cross et al., 2007;Suresh, 2007).
The mechanical properties of a cell change during oncogenesis, with primary solid tumour cells typically stiffer than the healthy cells from which they arise (Paszek et al., 2005). However, a significant decrease in stiffness is correlated with metastatic Journal of Cell Science • Accepted manuscript potential, with one study having found a 73% reduction in stiffness following metastasis (Cross et al., 2007;Guck et al., 2005;Hou et al., 2009). Viral infection also induces cortical actin restructuring, with one strain of rubella virus accounting for an increase of 17.8% in Young's modulus, a measure of substrate stiffness (Kräter et al., 2018). Thus, stiffening or softening of cells within tissues may be indicative of transformation.
Whether there is a direct link between changes in tissue and/or cellular stiffness and NK cell responsiveness is currently untested.
Here, we set out to examine the impact of altered target rigidity on NK cell function.
Using polyacrylamide substrates coated with monoclonal antibodies (mAb) against NKp30 and lymphocyte function-associated antigen-1 (LFA-1) to mimic the activating target interface, we demonstrate that NK cell degranulation and secretions of IFNγ, granzyme A and B, granulysin and Fas ligand (FasL) were enhanced on stiffer substrates. Using gel microbeads of varying stiffness as surrogate targets we observed that interactions between NK cells and soft targets were typically asymmetrical and unstable. In contrast, stiff beads caused MTOC and perforin granule polarisation accompanied by strong degranulation. Thus, we demonstrate that target stiffness plays an important role in the formation of the NK cell synapse and downstream cellular activation.

Synthesis and characterisation of hydrogel substrates
NK cells play a crucial role in eliminating malignant and diseased cells but whether or not the rigidity of target cells affects NK cell cytotoxicity has not been widely studied. To test this, polyacrylamide substrates were synthesised to mimic the activating target interface covering a range of rigidities. By altering the final concentration of acrylamide/bis-acrylamide, substrates were produced with three different Young's Journal of Cell Science • Accepted manuscript moduli of 1 ±0.1, 22 ±1 and 142 ±35 kPa, as assessed by AFM ( Fig. 1 A). These values represent a range of physiologically relevant rigidities spanning soft, medium and stiff tissues respectively (Xu et al., 2012;Nikolaev et al., 2014;Buxboim et al., 2010). To activate NK cells and induce assembly of a cytotoxic synapse, hydrogel surfaces were coated with monoclonal IgG1κ antibodies equilibrated across the different surfaces ( Fig.   1 B).

NK cell degranulation efficiency is modified by the rigidity of the target substrate
To test whether substrate stiffness influences NK cell degranulation, NK cells were

NK cell cytokine secretions are enhanced on stiffer substrates
Next, we set out to determine whether NK cell cytokine secretions were modulated by the stiffness of the activating substrate. To test this, primary NK cells were placed on 1, 22 and 142 kPa surfaces coated with anti-LFA-1 or anti-LFA-1 plus anti-NKp30 mAb.
After 24 h, supernatants were collected and a range of proinflammatory cytokine Journal of Cell Science • Accepted manuscript secretions quantified using a cytokine bead array. Little, if any, IL-2 was secreted in response to ligation of LFA-1 and NKp30. However, IFNγ, FasL, granzyme A and B and granulysin were all secreted. Importantly, secretions increased with substrate stiffness, with high levels of IFNγ, FasL, granzyme A and B and granulysin all being secreted when NK cells were interacting with the stiffest 142 kPa substrates (Fig. 1 F). By plotting the fold change in each donor's secretion of IFNγ, there was a 1.6 and 4.5-fold change in secretion for 22 kPa and 142 kPa surfaces respectively, in comparison to 1 kPa surfaces ( Fig. S1 C).

Synthesis and characterisation of gel microbeads
Degranulation at the IS occurs following successful polarisation of the MTOC and lytic granules towards the NK cell-target interface. Here, granules fuse with the synaptic membrane and their contents are expelled into the synaptic cleft, as supramolecular attack particles (Ambrose et al., 2020;Bálint et al., 2020). It is currently unclear how altering substrate stiffness impacts these different stages of NK cell activation, which lead up to secretion. Whilst flat gel substrates are useful for studying effector functions, examination of organelle and protein polarisation to the synapse is challenging.
Capturing images in a defined Z-plane in the very initial moments of cell interaction with the 2D substrate excludes intracellular structures beyond a specific distance above the focal plane. We therefore sought to synthesise gel microbeads as a more physiological representation of a target cell than a flat 2D surface, where the stiffness could be altered in a controlled manner and the 3D aspects of NK cell activation accurately assessed. Sodium alginate has commonly been used to produce gel beads on a millimetre scale for cell encapsulation, due to their high degree of biocompatibility and mild gelling conditions in the presence of divalent cations (Lee and Mooney, 2012;Rowley and Mooney, 2002;Wang et al., 2005;Lemoine et al., 1998). Stiffness of alginate can be Journal of Cell Science • Accepted manuscript altered by changing either the viscosity of the alginate or the alginate/crosslinker concentration (Andersen et al., 2014;Chan et al., 2011). Using a water-in-oil emulsion method, aqueous sodium alginate was dripped into an oil bath and continuously stirred for 30 min to allow the formation of alginate beads. The addition of calcium chloride induced rapid gelation of the alginate beads and were collected by centrifugation after extensive washing with isopropanol ( Fig. 2 A).
Next, we exposed alginate beads synthesised from low and intermediate viscosity alginates to poly-L-lysine (PLL). Electrostatic interactions between the negatively charged alginate bead surface and the positively charged PLL residues induces the formation of a polyanionic/cationic coat at the bead surface (Thu et al., 1996) (Fig. 2 B).
The formation of a PLL coat has been shown to enhance the mechanical properties of alginate (Bhujbal et al., 2014;Kleinberger et al., 2013) and by exposing beads to differing concentrations of PLL, beads were produced across a wide range of rigidities.
To activate NK cells, PLL-coated alginate beads were coated with anti-LFA-1 and anti-NKp30 IgG1κ mAb. To ensure mAb levels across the bead batches were similar, beads were coated with anti-NKp30 mAb directly conjugated to an AF647 fluorophore and their fluorescent intensity compared across the bead batches. Using imaging flow cytometry, fluorescence was detected on the surface of all three bead batches (

Target stiffness directly impacts NK cell degranulation efficiency
Next we set out to determine whether engaging targets of varying rigidities impacts NK cell degranulation. Primary NK cells were mixed with 9, 34 and 254 kPa (soft, medium and stiff targets respectively) beads coated with either BSA, anti-LFA-1 mAb alone or anti-LFA-1 with anti-NKp30 mAb. In response to beads coated with BSA or anti-LFA-1 alone, there was little, if any, staining of CD107a above that of unstimulated NK cells (incubated in media alone) confirming minimal NK cell degranulation with BSA or anti-LFA-1 coated beads (Fig. 3, A and B). However, upon co-incubation with beads coated with anti-LFA-1 and anti-NKp30 mAbs, the proportion of NK cells which stained with CD107a, and therefore degranulated, increased as the stiffness of the target increased; 5.0 ±3.8%, 8.7 ±4.2% and 19.8 ±14.0% for soft, medium and stiff targets respectively ( Fig. 3 B).
Whilst these results suggest that stiffness increases degranulation it could not be ruled out that these observations were due to the thicker PLL membrane on the medium and stiff beads, or some other non-specific mechanism of activation. To confirm that the differences in degranulation responses were due to the altered substrate rigidity, we blocked the NKp30 receptor on NK cells using anti-NKp30 mAb prior to the start of the degranulation assay. This should abrogate degranulation in the absence of any other stimulation. Indeed, blocking the NKp30 receptor completely abolished degranulation when incubated with soft, medium and stiff beads demonstrating degranulation was Journal of Cell Science • Accepted manuscript occurring via NKp30 (Fig. 3, C and D). Altogether, these data establish that increasing the tension of the target substrate results in enhanced degranulation triggered via NKp30.

The NK cell spreading response is sensitive to target stiffness
The NK cell cytolytic response is triggered in multiple stages (Orange, 2008) and thus, we next set out to assess at which stage the NK cell response was impaired against soft targets. Maturation of the synapse is dependent on an initial spreading of the NK cell against the target interface, mediated and sustained by actin polymerisation at the cell periphery. Thus, we tested whether or not actin polymerisation at the synapse was impacted by changes in target stiffness.
Primary NK cells were mixed with soft, medium and stiff bead targets for 20 min, fixed and f-actin stained with phalloidin-AF647 ( Fig. 4 A). To assess the amount of f-actin at the synapse, the fluorescence intensity of phalloidin staining at a cell/bead interface was compared to a region of the same size at the back of the NK cell. Low target stiffness did not entirely stop accumulation of f-actin at the synapse (Fig. 4, A and B). However, for medium and stiff targets, the mean increase in f-actin at the synapse was much more pronounced, doubling when anti-NKp30 mAb was present (with a fold-change intensity increase from 0.7 ±0.4 to 1.5 ±1.0 and from 0.8 ±0.4 to 1.6 ±0.9 for 34 kPa and 254 kPa targets respectively, Fig. 4 B). Additionally, the proportion of NK cells in conjugates was higher against 254 kPa targets (Fig. S3 A).
To assess NK cell spreading against the different targets, the length of each synapse was measured (example shown in Fig. 4 C). In conjugates with NKp30 ligated, there was a strong increase in NK cell spreading against stiff targets, more than for soft targets (9.0 ±0.5 µm vs 7.8 ±0.6 µm, Fig. 4 D).

Polarisation of the MTOC towards the immune synapse is dependent on the mechanical properties of the target interface
Triggering of an activating IS results in rapid polarisation of perforin-rich granules towards the cell/target interface which is dependent on the polarisation of the MTOC.
Thus, we next assessed whether or not the MTOC polarised similarly against targets of different stiffness. Primary NK cells were incubated with soft, medium and stiff targets for 20 min, fixed, permeabilised and stained for f-actin, MTOC and perforin granules. To test whether or not activation by soft or stiff beads could be affected by ligand density, we also assessed the degree of MTOC polarisation where (1)

Journal of Cell Science • Accepted manuscript
Prior to polarisation, perforin-rich granules translocate along microtubules and cluster around the MTOC (Mentlik et al., 2010;Orange, 2008). By staining CD107a and the MTOC simultaneously we tested whether or not granule coalescence and polarisation were affected by target stiffness (Fig. 5 B). By dividing the distance from the IS to the furthest granule by the length of the cell, we calculated the percentage of the cell occupied by granules (Fig. 5 D). The granules were typically spread throughout the cell in conjugates with 9 kPa targets, occupying 70 ±27% and 67 ±27% of the cell, for beads coated with anti-LFA-1 alone and with NKp30 mAb, respectively. There was a marked decrease in granule spread when NK cells were conjugated to 254 kPa targets, with the proportion of the cell containing granules being 38 ±24% and 25 ±10%. Granule coalescence and MTOC polarisation together are critical steps in cytolysis.
Analysing the two parameters together reduces uncertainty that the MTOC is classed as polarised or granules clustered by chance. We scored all images by eye as polarised/unpolarised (MTOC) and clustered/dispersed (granules) and plotted the four frequencies for beads coated in anti-LFA-1 alone and with anti-NKp30 (Fig. 5

F and G).
As granule coalescence occurs before MTOC polarisation , cells in which granules were dispersed (white bars with red lines) essentially indicate a lack of reactivity (whether or not the MTOC shows an apparent polarisation). These constituted the majority of conjugates with 9 kPa beads coated in NKp30 mAb (7 ±12% and 50 ±17% for conjugates with polarised and unpolarised MTOC, respectively). Conjugates

Journal of Cell Science • Accepted manuscript
with an unpolarised MTOC but clustered granules (amber bars) represented 23 ±21% of conjugates overall. Only 20 ±10% of conjugates were cytolytic, with clustered granules and polarised MTOC (green bars). In contrast, 254 kPa targets coated in NKp30 mAb far more effectively triggered these steps in cytolytic conjugate formation, with 77 ±12% of conjugates having clustered granules and a polarised MTOC. Blocking NKp30 abrogated MTOC polarisation against beads coated in LFA-1 and NKp30 mAb, and reduced MTOC polarisation against beads coated in LFA-1 mAb alone (Fig. S4). It is not entirely clear as to why anti-NKp30 mAb blocking would affect polarisation triggered via LFA-1 alone, but perhaps this relates to the nanoscale proximities of activating receptors and integrins on the cell surface such that blocking one impacts the accessibility of the other.

Synapse stability is sensitive to target stiffness
Target stiffness influences several steps in NK cell cytotoxicity, with stiff targets inducing slightly more NK cell spreading, and increased polarisation of the MTOC and perforin granules. T-cell kinapses denote asymmetrical, unstable interactions between T-cells and targets in which the lymphocyte is not sufficiently activated to halt motility (Dustin, 2008). Using beads coated in anti-LFA-1 with or without anti-NKp30, we set out to characterise the proportion of conjugates in which the NK cell phenotype indicated motility, with the focus of f-actin at the leading edge lamellipodium separate from the IS.
Talin polarisation to the contact between lymphocyte and targets is dependent on LFA-1 ligation, and is a precursor to both synapse and kinapse formation (Smith et al., 2005).
Importantly, against all bead conditions we observed a similar accumulation of talin in focal zones (Fig. 6 A, Fig. S5), similar to those previously described against ICAM-1 coated polystyrene beads (Mace et al., 2009). This indicated that the NK cells were interacting with, rather than bypassing, the beads, irrespective of their stiffness.

Journal of Cell Science • Accepted manuscript
We categorised each conjugate, according to the positions of the MTOC and organisation of f-actin within the NK cell, into four groups: unpolarised kinapse, polarised kinapse, unpolarised synapse, and polarised synapse (Fig. 6 B). Conjugates were denoted kinapses in which the distribution of f-actin was clearly asymmetrical relative to the line perpendicular to the bead/cell interface. All conjugates in which the degree of symmetry was unclear were discarded. Cells with an MTOC ratio <0.3 were considered polarised.
The proportions of each type of interaction (unpolarised/polarised, synapse/kinapse) were determined for interaction between NK cells and 9 kPa, 34 kPa and 254 kPa targets (Fig. 6 C). Against 254 kPa targets, synapses outnumbered kinapses in NK cell interactions that were both polarised and unpolarised, indicating that these targets provided a threshold stiffness for the establishment of symmetrical synapses. Against 34 kPa targets coated in anti-NKp30, the percentage of unstable kinapses that resulted in either polarised or unpolarised NK cells constituted 10 ±10% and 12 ±6% of interactions, respectively. Against soft 9 kPa targets, interactions were predominantly either unpolarised and/or kinapses. Analogous data were obtained when NK cells were stimulated via the antibody-dependent cellular cytotoxicity pathway, through CD16 ligation (Fig. S6). Thus, a soft target surface does still allow interaction with an NK cell, evidenced by talin recruitment, but does not meet a threshold required for symmetrical synapse formation. Taken together these results demonstrate that target mechanical properties and receptor ligation both contribute to the formation of a stable symmetrical synapse.

Discussion
Diseased tissue often presents with altered extracellular matrix protein compositions giving rise to changed stiffness of both the matrix and cells present (Engler et al., 2006;Cross et al., 2007;Liu et al., 2015;Paszek et al., 2005). There have been several studies of how differences in force impact on TCR signalling (Simonson et al., 2006;O'Connor et Journal of Cell Science • Accepted manuscript al., 2012;Judokusumo et al., 2012;Springer and Dustin, 2012) and T cell function (Saitakis et al., 2017;Ricker et al., 2016;O'Connor et al., 2012). Despite their role in tumour surveillance, and the established link between mechanical changes and oncogenesis, the effect of target stiffness change on NK cell effector functions has been far less studied.
Here, we set out to investigate how altering the stiffness of the target interface influenced NK cell cytotoxic functions. Polyacrylamide hydrogels coated with mAb against the activating receptor NKp30 and the integrin LFA-1 were used to mimic an activating target interface. We establish that the pro-inflammatory cytokine IFNγ, as well as granulysin, FasL and granzymes A and B are secreted by human NK cells in response to LFA-1 and NKp30 ligation with increasing secretion observed on stiffer substrates. NK cells on soft substrates activated by the addition of PMA and ionomycin degranulated to the same extent as NK cells on the stiffest substrates implying that NK cells on soft substrates are not incapacitated, but rather receptor-mediated activation of NK cells is mechanically sensitive.
To our knowledge, all previous evidence of varied cellular cytotoxic responses to mechanical stiffness has relied on 2D substrates. We present a novel methodology for using PLL-coated alginate beads as surrogate targets to study single cell responses in 3D. This approach establishes that the mechanism of stable IS formation is optimised for stiff targets.
A small number of ill-formed 9 kPa beads were present in our production process, which exhibited some level of internal staining. But these cannot account for differences in stimulation for a number of reasons. Aberrant 9 kPa beads with internal staining were excluded from experiments based on image analysis. Moreover, beads of 34 kPa and 254 kPa did not exhibit any level of internal stain, yet there were significant differences between the ability of these beads to stimulate NK cells. The density of Journal of Cell Science • Accepted manuscript protein staining was consistent across different bead types. And perhaps most importantly, if the level of antibody coating was changed dramatically, soft beads with high staining were impaired in their ability to activate NK cells, while hard beads with low staining still activated NK cells effectively.
Mechanical force and ligand binding can be functionally intertwined (Comrie et al., 2015) and a lack of opposing force with soft targets perhaps fails to engage the molecular clutch upon LFA-1 engagement. Additionally, soft targets are prone to ruffling and may provide insufficient traction forces for the outward growth of the immune synapse. A similar phenomenon was seen when studying fibroblast movement over rubber (Harris et al., 1980). Greater forces produced by stiffer targets likely allow faster and more sustained actin retrograde flow, with symmetrical spreading of the NK cell around the target. Against soft targets random slippage events occur during synapse formation (Matalon et al., 2018) and the contractile forces of molecular motors dissipate, which can cause actin filaments to slide (Gupta et al., 2016).
The tension of the target interface is also likely to be important in aiding myosingenerated forces, which serve to remodel the f-actin meshwork at the IS, allowing granules to translocate to the synaptic membrane (Carisey et al., 2018). Inhibiting myosin motor protein activity, either by RNA-interference mediated knockdown or by chemical inhibitors, blocks granule fusion to the NK cell membrane (Sanborn et al., 2009;Andzelm et al., 2007). Total internal reflection fluorescence imaging of lytic granules at the IS in YTS cells spread on CD28-coated slides showed that myosin inhibitors diminished penetration of granules into f-actin at the IS (Sanborn et al., 2009).
Moreover, inhibition of myosin activity led to an increased density of fibres resulting in fewer granule-permissive channels at the synapse, despite retention of the overall actinmesh architecture (Carisey et al., 2018). Microtubules play a crucial role in mediating Journal of Cell Science • Accepted manuscript MTOC and perforin translocation to the synapse (Huse, 2012). Of all the effects of varied target stiffness which we tested here, the strongest correlate was MTOC polarisation.
The extent to which signalling through NKp30 is directly affected by target stiffness remains to be established. NKp30 signals via an associated CD3ζ chain which induces recruitment of the tyrosine kinases Syk and ZAP70. These kinases in turn phosphorylate transmembrane adaptor molecules such as LAT leading to activation of PLC-γ1 and the Vav family of proteins (Watzl and Long, 2010;Watzl et al., 2014). T cells plated on activating soft polyacrylamide substrates were devoid of phosphorylated Syk and ZAP70 kinases within the cell centre even by 30 min post-seeding. In contrast, stiffer substrates showed rapid accumulation of phosphorylated ZAP70 and Syk already within 2 minutes post-seeding (Judokusumo et al., 2012). Thus, the low percentage of NK cells degranulating with soft targets may result from poor recruitment of phosphorylated ZAP70 to the NK cell synapse (Matalon et al., 2018).
Several studies have documented a reduction in stiffness during early apoptosis as a result of cytoskeletal degradation (Pelling et al., 2009;Su et al., 2019;Schulze et al., 2009). A lack of reactivity to such soft targets may have evolved to conserve energy and/or cellular components that would otherwise be wasted in an attempt to kill targets which were already apoptotic. Metastatic cancer cells are one critical exception in which a lower stiffness signals danger. Invasive cells such as SKOV3 and HEY display softer mechanical characteristics allowing the shape changes required to undergo metastasis (Swaminathan et al., 2011). On account of their reduction in stiffness, NK cells may fail to respond to these softer targets. Glioma cells are another example in which disease progression is correlated with a decrease in cellular stiffness (Pepin et al., 2018).
Intriguingly, glioblastoma cells evade T cell killing by locally preventing IS formation, with interactions characterised by kinapse dynamics (Díaz et al., 2018). There are settings in which it may be preferential for an NK cell not to form a full IS, with a Journal of Cell Science • Accepted manuscript kinapse-like mode of interaction allowing deeper penetration into solid tumours (Deguine et al., 2010).
Overall, our data establishes the importance of target stiffness in regulating NK cell cytotoxicity. We have revealed that soft targets represent a blind spot in NK cell killing, which suggests that acquisition of mechanical softness by diseased cells may be a novel immune evasion mechanism. F12 Ham, 10% human serum, 1% Non-essential amino acids, 1 mM sodium pyruvate (Sigma), 2 mM L-glutamine, 50 U/mL penicillin streptomycin, 50 µM 2-mercaptoethanol (Gibco). NK cells were stimulated with rhIL-2 (200 U/ml; Roche) and rested for 6 days prior to experiments. Cells were kept at 37°C in a 5% CO2 atmosphere.

Preparation and synthesis of polyacrylamide hydrogels
To prepare glass surfaces for gel binding, 3 µL bind-silane (GE healthcare), 47 µL glacial acetic acid and 950 µL ethanol (Fisher Scientific) were mixed together and 200 µL pipetted onto glass-bottomed dishes (No. 0, 14 mm diameter; MatTek). After 3 min, dishes were rinsed with ethanol and left to dry at room temperature (RT). Separately, to clean glass coverslips (No. 0, 13 mm diameter), they were initially sonicated in acetone Journal of Cell Science • Accepted manuscript for 15 min, rinsed twice with water and rocked for 1 h in NaOH (1 M). Coverslips were washed extensively with water and stored in 70% ethanol until use.
Hydrogels were synthesised using Protogel © (National Diagnostics) adapted from a method described by Pelham and Wang (Wang and Pelham, 1998)

Quantification of hydrogel stiffness
The hydrogel stiffness was characterised by the reduced Young's Modulus (E). Stiffness values were obtained from generating force curves gathered using a Bruker Bioscope Catalyst AFM with a Nanoscope V controller (Bruker UK Ltd.) mounted on a Nikon Eclipse Ti-I optical microscope (Nikon Instruments Europe B.V.) operating under the NanoScope controller software (v9.15) (Bruker Corp.). The gels were probed using a silicon nitride spherically-tipped cantilever (Windsor Scientific limited; nominal spring constant (k) of 0.3 N/m; 2.5 µm nominal radius). The spring constant was calibrated using the thermal oscillation tuning method (Lübbe et al., 2013). For the AFM setup, the cantilever was aligned over the sample using a camera (Axiocam MRc, Zeiss) attached to the microscope and the relationship between the photodiode signal and cantilever deflection (deflection sensitivity) obtained from force curves generated on glass.

Journal of Cell Science • Accepted manuscript
The local reduced modulus was determined for 5 different locations on the gel surface in a 1 x 1 µm 2 region, indented (100 nm depth) at a frequency of 1 Hz with lateral spacing of 0.1 µm. Analysis was carried out using force curve analysis software (NanoScope analysis v1.40; Bruker) whereby a baseline correction was applied to each curve before a force fit was applied using the Hertzian (spherical) model (Lin et al., 2007) and a maximum force fit of 70%. Any force values falling more than 2 SD away from the mean value for a given area were discarded to account for failed indents.
For 34/254 kPa beads, post-gelation, 10 mL of isopropanol (Fisher) was added to the stirring mixture, incubated for 10 min whereupon the solution was transferred to a 50 mL falcon tube. For 9 kPa beads, after 1 h the bead mixture was immediately transferred to a 50 mL falcon tube. Beads were spun down at 1500 g for 10 min and the supernatant removed. The beads were spun down a further four times, using fresh isopropanol each time (40 mL). Beads were then washed twice and resuspended in 100 mL sterile water, and filtered through a 40 µm cell-strainer (Fisher) to remove beads or debris >40 µm.
The flow-through was spun down at 1500 g for 10 min and beads resuspended at 2 x 10 6 /mL in sterile water.

Poly-L-lysine and protein coating
Poly-L-lysine (PLL; MW 15-30 kDa; P7890; Sigma) was dissolved in sterile water (0.1% w/v) and stored in sterile conditions at 4°C for up to 6 months. 2x10 6 /mL beads were incubated in differing dilutions of PLL (0.0005, 0.002 and 0.005% for 9, 34 and 254 kPa Journal of Cell Science • Accepted manuscript beads respectively) in a total volume of 10 mL. 9 kPa beads were incubated with PLL for 2 min whilst continuously rocked (35 RPM) and 34/254 kPa beads were vortexed continuously for 10 min for optimal coating. Post-incubation with PLL, beads were spun down (1500 g) for 5 min, the supernatant removed and washed twice with PBS.
PLL-coated beads were counted using a haemocytometer and 0.4x10 6 /mL beads were transferred to a 1.5 mL Eppendorf tubes in 200 µL PBS. Anti-NKp30 mAb (P30-15, BioLegend), anti-CD16 mAb (3G8, BioLegend) and/or anti-LFA-1 mAb (TS2/4 or HI111 where specified, BioLegend) were added as indicated (each at a total concentration of 50/10/10 µg/mL for 9, 34 and 254 kPa beads respectively) and left for 1 h whilst being continuously shaken at RT. For beads coated with anti-LFA-1 mAb only, a matched nonstimulatory murine isotype control was used in place of anti-NKp30 to ensure protein densities were similar. After 1 h, beads were spun down (2000 g) and washed twice with 1 mL PBS and blocked with 3% BSA/PBS for 30 min whilst continuously shaking.
Beads were washed with 1 mL PBS before adding to cells. To test the amount of anti-NKp30 mAb bound to the bead surface, beads were incubated for 1 h with anti-NKp30-AF647 (labelled with AF647 NHS ester, Thermo Fisher) at RT. The gMFI for each bead batch was determined using flow cytometry (Imagestream, Amnis).

Characterising bead stiffness
Bead stiffness was characterised using the same system as per hydrogel stiffness quantification. Beads were immobilised to prevent bead movement during force curve acquisition. Briefly, 1x10 5 /mL PLL-coated beads were extruded through a 10 µm membrane filter (13 mm diameter; Millipore) using a 1 mL syringe. The membrane filter was subsequently fixed to a glass slide and submerged in PBS. Individually-trapped beads within the membrane pores were found using a camera (Axiocam MRc, Zeiss) attached to the microscope.

Journal of Cell Science • Accepted manuscript
Force curves were obtained, for beads submerged in PBS, via an indentation made with a borosilicate glass colloidal probe mounted on a cantilever (CP-CONT-BSG-A, sQUBE; nominal spring constant (k) of 0.3 N/m; 2.5 µm nominal radius). The spring constant was calibrated using the thermal oscillation tuning method (Lübbe et al., 2013). The local reduced modulus for each bead was determined by indenting 10 different points on each bead.

Cytokine bead array
To stimulate cytokine secretions, 2x10 5 NK cells were incubated on flat polyacrylamide substrates coated with anti-LFA-1 or anti-LFA-1 plus anti-NKp30 mAb for 24 h. The supernatants were subsequently collected and stored at -20°C for further analysis.
Cytokines in the supernatant were detected using a cytokine bead array kit (CD8/NK cell LEGENDplex kit; BioLegend), carried out according to the manufacturer's instructions. The samples were analysed by flow cytometry (FACSVerse; BD) and data were analysed using flow cytometry analysis software (LEGENDplex software v8.0; BioLegend).

Degranulation assays
For measuring NK cell degranulation, 5 x10 5 NK cells were mixed with 5 x10 5 beads in a total volume of 100 µl media ( Journal of Cell Science • Accepted manuscript µg/mL; Sigma) was used to stimulate NK cell degranulation via soluble stimulation and was added to a separate well of NK cells 1 h before the end of the assay. Where indicated, blocking NKp30 was carried out by incubating NK cells (1x10 6 /mL) with anti-NKp30 mAb (P30-15; 20 µg/mL), or an isotype-matched control mAb for comparison, for 1 h at 37°C prior to the start. These cells were washed in fresh medium prior to the start of the degranulation assay.

Imaging NK cell -bead interactions
To prepare NK cell-bead conjugates, protein-coated beads were spun down at 2000 g for 5 min and resuspended in human serum-free medium. Beads were plated out into glassbottomed wells (Labteks no.1.5; Nunc; 4x10 5 soft beads per well and 2x10 5 medium/stiff beads per well) pre-coated with 10 µg /mL fibronectin (F0985, Sigma).
Soft beads were plated at a higher concentration as they were more difficult to locate when imaging. Beads were allowed to settle for 1 h at 37°C. NK cells were spun down at 300 g for 5 min and the supernatant removed. The cell pellet was resuspended in media with 10% FBS and 2x10 5 cells added into each well. Conjugates were left to form for 20 min, then fixed by the addition of 4% PFA/PBS for 20 min and permeabilised for 10 min with 0.1% Triton-X100/PBS. Cells were subsequently blocked overnight with 3% BSA/PBS. For MTOC imaging, cells were stained with 1 µg/mL anti-pericentrin antibody (ab4448, Abcam) for 2 h at 4°C followed by 5 µg/mL AF568 labelled anti-rabbit IgG H&L secondary antibody (A11035, Invitrogen). To image f-actin at the synapse fixed conjugates were stained with 33 nM phalloidin-AF647 or phalloidin-AF488 (A22287 and A12379, Thermo Fisher) for 1 h at RT. To image granule polarisation, cells were stained with anti-LAMP-1-AF647, 5 μg/mL for 1 h (sc-20011, Santa Cruz Biotechnology).

Journal of Cell Science • Accepted manuscript
Conjugates were imaged by confocal microscopy (Leica TCS SP8) using a 100X/1.40 NA oil-immersion objective and white light laser source. Images were acquired using sequential imaging to avoid spectral overlap and analysed using ImageJ (Schneider et al., 2012) (National Institutes of Health). Accumulation of f-actin at the synapse was determined by the fold increase in MFI staining at the cell bead interface divided by the MFI from a region at the back of the cell of the same size. Spreading of NK cells against the beads was assessed by measuring the length of f-actin at the bead/cell interface.
Polarisation of the MTOC was assessed by measuring the ratio of the distance from the MTOC to the cell-bead interface to the distance from the synapse to the back of the cell. The experiments across this manuscript were based on image analysis of conjugates between a single cell and a single bead. Any bead which displayed aberrant properties was not scored in these assays; beads which appeared abnormal were entirely excluded.
The brightness and contrast of images has been adjusted in representative images to make the MTOC more clearly visible, but all analyses were carried out on unadjusted images.

Imaging flow cytometry
Beads coated in mAb were washed in PBS and analysed by imaging flow cytometry (ImagestreamX MK-II; Amnis). Both brightfield and fluorescent images were captured using 40X zoom and data analysed using the IDEAS® software package (Merck). Bead debris and doublets were gated out prior to bead analysis and a minimum of 5,000 beads were acquired per experiment.

Statistical analysis
All statistical analysis was carried out using Graphpad Prism (Graphpad software, version 7). For each data set, the Shapiro-Wilk normality test was used to evaluate the distribution of values. For groups of data displaying normal Gaussian distributions the statistical significance was determined by the use of parametric tests. Two-tailed t-tests were used to examine statistical significance between two groups of data and one-way ANOVA used for three or more data groups. To compare statistically significant differences between groups, one-way ANOVA with Tukey's multiple comparisons was carried out. For data not showing normal distribution, non-parametric equivalent tests were used. Wilcoxon-signed rank tests were used as the non-parametric t-test equivalent whilst the Kruskal-Wallis test or matched values Friedman test with Dunn's post-testing was used to replace the one-way ANOVA. A calculated P value <0.05 was considered statistically significant (indicated by a single asterisk*). P<0.01(**), P<0.001(***) and P<0.0001(****) whilst P≥0.05 was considered not significant (ns).    Journal of Cell Science • Accepted manuscript