Biphasic mechanosensitivity of T cell receptor-mediated spreading of lymphocytes

Significance

The ability of a T cell to explore environmental mechanical cues, through bonds formed by its special receptors called T cell receptors (TCRs), is crucial for the first steps of immune recognition. Here we show that the response of T cells, quantified in terms of their spreading behavior, is biphasic with substrate stiffness when mediated through TCRs. However, when the ligands of the T cell integrins are additionally involved, the cellular response becomes monotonic. Based on a mesoscale model, this ligand-specific response can be attributed to differences in force sensitivity and effective stiffness of the link formed between the ligand/receptor pairs and the actin cytoskeleton. This may provide a general mechanism for immune cells to discriminate mechanosensitive bonds.

Abstract

Mechanosensing by T cells through the T cell receptor (TCR) is at the heart of immune recognition. While the mechanobiology of the TCR at the molecular level is increasingly well documented, its link to cell-scale response is poorly understood. Here we explore T cell spreading response as a function of substrate rigidity and show that remarkably, depending on the surface receptors stimulated, the cellular response may be either biphasic or monotonous. When adhering solely via the TCR complex, T cells respond to environmental stiffness in an unusual fashion, attaining maximal spreading on an optimal substrate stiffness comparable to that of professional antigen-presenting cells. However, in the presence of additional ligands for the integrin LFA-1, this biphasic response is abrogated and the cell spreading increases monotonously with stiffness up to a saturation value. This ligand-specific mechanosensing is effected through an actin-polymerization–dependent mechanism. We construct a mesoscale semianalytical model based on force-dependent bond rupture and show that cell-scale biphasic or monotonous behavior emerges from molecular parameters. As the substrate stiffness is increased, there is a competition between increasing effective stiffness of the bonds, which leads to increased cell spreading and increasing bond breakage, which leads to decreased spreading. We hypothesize that the link between actin and the receptors (TCR or LFA-1), rather than the ligand/receptor linkage, is the site of this mechanosensing.

Mechanosensitivity has emerged as a hallmark of many biological systems—from the molecular to the tissue level—and is implicated in health and disease (1–3). Sensitivity to environmental stiffness has now been demonstrated for almost all cell types. Readouts include extent of spreading, cell stiffness, motility, differentiation, and traction forces (4–6), and all typically increase monotonically with substrate stiffness, usually reaching a saturation value at a particular value of the stiffness, which may be related to the stiffness of the physiological cellular environment (5). From a theoretical point of view, adhesion and mechanosensing can be treated either within cell-scale macroscopic models (3) that usually predict a monotonous increase in cell spreading and traction with stiffness or using microscopic models that account for molecular mechanosensitivity (7) which could lead to biphasic behavior in traction forces or migration velocity (8). Experimental reports of such biphasic behavior for any readout are very rare for cells adhering to a surface; examples include motility of neutrophils (9) and smooth muscle cells (10) and, recently, traction forces in talin-silenced fibroblasts (11).

The ability of T cells to recognize pathogens and pathological cells depends on their mechanosensitivity (12, 13) which is thought to implicate the CD3 domain of the T cell receptor (TCR) complex (14). Since the natural ligands in this context are usually mobile in the plane of the antigen-presenting cell (APC) membrane, in vitro experiments often involve supported lipid bilayers (SLBs) carrying mobile ligands (15). The mobility of the ligands was shown to impact cell spreading (16, 17), via a mechanosensory mechanism (17). However, the mechanosensitivity of leukocytes in general and lymphocytes in particular has been much less studied than that of focal adhesion-forming cells, despite direct evidence that T cell response is sensitive to the substrate stiffness (14, 18, 19). Interestingly, for TCR-mediated T cell spreading, contradictory trends were reported in the past (18, 19). However, since one study used mouse cells on a hydrogel with elasticity of 10–200 kPa and the other used a human cell line on silicone elastomers at 100–2,000 kPa, a direct comparison of results was not possible. In the light of our results presented below, this contradiction in fact presages biphasic behavior.

Here we explore mechanosensing in T cells via the TCR-CD3 complex, in presence or absence of ICAM-1—a ligand for the T cell integrin LFA-1. The ligand of choice is anti-CD3 which provides sufficient adhesion via the TCR-CD3 complex alone (17, 20)—something not possible with the weaker pMHC/TCR bond, at the same time eliciting the same signaling pathways as pMHC ligation. We follow the spreading of Jurkat T cells on functionalized surfaces, spreading being a marker of future proliferation for T lymphocytes (21). The spreading data are interpreted within a mechanistic model, and a possible mechanism underlying the unusual mechanical response of the T cells is suggested.

Results

Characterization of T Cell Spreading on Soft Substrates.

Soft substrates were glass-supported polydimethylsiloxane-based (PDMS silicone) elastomer layers with typical thickness of several micrometers, with stiffness, as quantified in terms of Young’s modulus, ranging from 500 Pa to several megapascals (SI Appendix, Table S1). To cover such a large stiffness range while retaining similar surface chemistry, different PDMS types with adapted polymer/cross-linker ratio were used. The PDMS surface was functionalized (Fig. 1A) with either anti-CD3 alone or, additionally, with ICAM-1 or an antibody targeting the coreceptor CD28. The surface density of anti-CD3 was estimated to be 400 ± 50 molecules/$\mathrm{\mu }{\mathrm{m}}^2$ (SI Appendix, SI Methods). To characterize the maximal spreading state, T cells were allowed to interact with the functionalized surfaces for 20 min and optionally imaged during this time to capture the dynamics, allowing them enough time to spread but not retract (17). They were then fixed, optionally labeled, and imaged in bright field (BF), reflection interference contrast microscopy (RICM), total internal reflection fluorescence (TIRF), and confocal microscopy. RICM images were analyzed to quantify the cell spread area and the actin TIRF microscopy (TIRFM) images were analyzed to quantify the extent of actin peripherality (see SI Appendix for details) (17).

Fig. 1.

T cell spreading on elastomers functionalized with anti-CD3. (A) Schematic representation of the experiment. T cells interact with elastomers with stiffness ranging from 500 Pa to 40 MPa and functionalized with anti-CD3. (B) T cells were allowed to spread for 20 min on either 20-kPa or 2,440-kPa substrates, were fixed, and were imaged in BF, RICM, TIRFM (actin), and confocal modes. (C) Cell spread area as quantified from RICM (0.5 kPa, n = 116, N = 4; 3 kPa, n = 92, N = 3; 4 kPa, n = 25, N = 1; 5 kPa, n = 128, N = 3; 20 kPa, n = 103, N = 3; 145 kPa, n = 91, N = 3; 300 kPa, n = 57, N = 1; 2,440 kPa, n = 119, N = 3; 7,000 kPa, n = 20, N = 1; 40 MPa, n = 41, N = 1; glass, n = 94, N = 4). The range of very soft (light green), soft (green), intermediate (blue), and hard (red) is color coded. (D) Time course of cell spreading imaged in RICM. (E) Percentage of nonspreading cells as time progresses and after fixation. (F) Quantification of cell spread area, excluding the nonspreading cells. The moment of addition of cells to the experimental chamber is taken as time zero. At least 100 cells for each time point and each stiffness are shown. Note that weakly adherent cells are washed away during fixation, thus driving the average area toward higher values. Data are average over all cells for a given condition, and error bars are SEM. **P < 0.01; *P < 0.05; ns, P > 0.1 indicates no significant difference. (Scale bars, 4 $\mathrm{\mu }{\mathrm{m}}^2$.)

TCR-Mediated Biphasic Spreading Response.

In the first set of experiments, the substrates were functionalized with anti-CD3 alone (Fig. 1A). Cells spread more on 20-kPa than on 2,440-kPa elastomers but the actin is peripherally distributed in both cases (Fig. 1B and SI Appendix, Fig. S1). Fig. 1C and SI Appendix, Fig. S2 summarize the final cell area measured for different elastomer stiffness. Note that for the same stiffness, changing the silicone type does not affect cell area, showing that the data are independent of the details of the chemical structure of the PDMS matrix. Cells on the softest elastomer (0.5 kPa) spread moderately (area about 200 $\mathrm{\mu }{\mathrm{m}}^2$). The area increases as the stiffness is increased up to 5 kPa, reaching a maximum of about 300 $\mathrm{\mu }{\mathrm{m}}^2$. Thereafter the cell spread area decreases with increasing stiffness, falling back to roughly 200 $\mathrm{\mu }{\mathrm{m}}^2$ for 2 MPa and to less than 150 $\mathrm{\mu }{\mathrm{m}}^2$ at 7 MPa. On equivalently functionalized glass, with nonspecific interactions fully blocked, the cells spread to a mere 120 $\mathrm{\mu }{\mathrm{m}}^2$. [To achieve full blocking, we used immobilized ligands grafted on SLBs (17). Without full blocking, area on anti–CD3-coated glass is high at 300 $\mathrm{\mu }{\mathrm{m}}^2$ due to nonspecific effects (17).] We verified that on PDMS of all types, cells fail to spread if anti-CD3 is not present and that the ligand density under the cells is identical to the background, thus making sure that the ligands are not ripped off during adhesion and spreading (SI Appendix, Fig. S3). If the PDMS is coated with poly-l-lysine instead of anti-CD3, cells spread marginally more on hard than on soft PDMS (SI Appendix, Fig. S4, i), and no hint of biphasic behavior is seen. However, on hydrogel (polyacrylamide) coated with anti-CD3, the biphasic spreading behavior is reproduced (SI Appendix, Fig. S4, ii).

This remarkable biphasic spreading behavior is also reproduced in human primary CD4⁺ T cells, which spread more on soft than on hard PDMS substrates (SI Appendix, Fig. S5). Taken together, the above experiments indicate that this unusual dependence of the spreading on elasticity is a specific TCR-complex–mediated response and that both human T cell lines and primary human cells show this behavior.

T cells tend to spread isotropically on functionalized glass or SLBs functionalized with anti-CD3 and ICAM-1, resulting in a roughly circular shape (17), whereas on the elastomers studied here, they may exhibit irregular shapes (SI Appendix, Fig. S6). A rough analysis of the membrane-to-substrate distance based on RICM images shows that although the average distance is higher for cells that are spread less, the membrane accesses lowest distances of 25–30 nm, compatible with TCR to anti-CD3 molecular contact, on both the softest and hardest substrates studied (SI Appendix, Fig. S7). The distribution of actin at the adhesive interface, a major determinant of cell shape, tends to be peripheral for a fully spread T cell, the peripherality becoming less pronounced for weakly spread cells (17, 22). Here, a peripheral ring-like distribution is seen for all stiffness values (Fig. 1B and SI Appendix, Fig. S1, quantified in SI Appendix, Fig. S8), even when the cell area is relatively small. The recruitment of TCR and ZAP-70 (zeta-chain–associated protein kinase 70—one of the first molecules to be recruited upon TCR engagement) imaged in TIRFM does not show appreciable variations with stiffness (SI Appendix, Fig. S1), hinting at a mechanical rather than a signaling-based mechanosensing mechanism.

Spreading Kinetics.

We next explored the time evolution of T cell spreading on soft (5 kPa), intermediate (130 kPa), and hard (2,440 kPa) substrates. Fig. 1D shows an example of single-cell time-lapse RICM demonstrating that the cells on hard substrates lag behind in spreading already in the time window 0–5 min, a period shown previously to be critical for antigen recognition (23). Fig. 1 E and F quantify this effect at the scale of the population. It is seen that on hard substrates there is a population of cells that never spread (Fig. 1E; cells with area less than 50 $\mathrm{\mu }{\mathrm{m}}^2$ are considered nonspreading). Furthermore, the cells that do spread do so to a lesser extent on the hard substrate (Fig. 1F). The time evolution of the actin organization is similar in the two cases (SI Appendix, Fig. S9).

Ligand Specificity and Role of Myosin.

To test the specificity and robustness of this effect, we first costimulated the cells with surfaces grafted with anti-CD3 and anti-CD28. Such costimulation elicited no further spreading beyond that already observed for anti-CD3 alone, and the dependence on stiffness also remained unaffected (Fig. 2A), confirming the absence of a role for CD28 in TCR-mediated mechanosensing (14). Further, on surfaces coated with anti-HLA, there was negligible spreading on both soft and hard substrates (Fig. 2B). T cells are known to spread moderately on anti–HLA-coated glass without being specifically activated (23), hinting that the effect seen here is specific to TCR. Finally, treatment of cells on anti–CD3-coated substrates with blebbistatin has no effect on spreading and its stiffness dependence (Fig. 2C). Suppression of actin polymerization (using cytochalasin D) or of ARP-2/3–mediated branching (using CK666), however, completely prevents spreading on both hard and soft elastomers (Fig. 2C and SI Appendix, Fig. S10). We therefore conclude that the biphasic stiffness dependence of spreading requires engagement of the TCR complex and is independent of myosin-generated contractions but depends on actin polymerization and branching.

Fig. 2.

(A–C) Comparison of final cell area, after 20 min of spreading followed by fixation, on anti-CD3 (A) with costimulatory dually functionalized substrates exhibiting both anti-CD3 and anti-CD28; (B) with anti-HLA, where cells spread weakly or not at all; and (C) after treatment with blebbistatin (+BB), cytochalasin D (+CytoD), or CK666 (+CK666). Note that for a given stiffness, neither costimulation nor blebbistatin has any effect on final area. Anti-CD3 data are reproduced here from Fig. 1 for direct comparison. ***P < 0.001. Data are averages, and error bars are SEM.

Additional Ligands for Integrins Abrogate Biphasic Response.

In the next set of experiments we explored the role of LFA-1 by dual functionalization of the substrates with anti-CD3 and ICAM-1. Consistent with past reports on glass with only ICAM-1 on the surface (17), with or without simultaneous stimulation with soluble anti-CD3, there is no spreading on PDMS (SI Appendix, Fig. S11). For the case of cofunctionalization with both anti-CD3 and ICAM-1 (Fig. 3A), the cells spread the least on the softest substrates used and they spread more as the substrate stiffness is increased, already reaching a saturation value of about 400 $\mathrm{\mu }{\mathrm{m}}^2$ in the kilopascal range (Fig. 3 B and C). Thus, the maximal area reached here is larger than that with anti-CD3 alone. The spreading dynamics do not differ between the soft (5 kPa) and hard (2,440 kPa) cases (Fig. 3D). In terms of actin organization, no difference is detected between soft and hard on one hand and with or without ICAM-1 on the other hand; the actin is mainly peripheral in all cases (SI Appendix, Fig. S8). Again, the spreading behavior is strikingly reproduced in human CD4⁺ naive T cells isolated from peripheral blood, which fail to spread on anti–CD3-functionalized hard PDMS without ICAM-1 but do spread on hard PDMS dually functionalized with ICAM-1 and anti-CD3, as well as on anti–CD3-coated soft PDMS with or without ICAM-1 (SI Appendix, Fig. S5).

Fig. 3.

T cell spreading on elastomers, with stiffness ranging from 500 Pa to 2.5 MPa, cofunctionalized with anti-CD3 and ICAM-1. (A) Schematic representation of the experiment. (B) T cells on such substrates imaged in BF, RICM, and TIRFM (actin) modes after 20 min of spreading and fixation. (C) Corresponding cell spread area (0.5 kPa, n = 124, N = 5; 20 kPa, n = 51, N = 2; 5 kPa, n = 59, N = 2; and 2,440 kPa, n = 89, N = 3). (D) Cell spread area as a function of time, with at least 100 cells for each time point and each stiffness (average area is slightly overestimated due to 30–40 nonspreading cells ignored for the analysis). Data are average over all cells for a given condition, and error bars are SEM. ***P < 0.001; ns, P > 0.1 indicates no significant difference. (Scale bars, 4 $\mathrm{\mu }$m.)

Model of Biphasic Cellular Response Based on Force-Dependent Bond Rupture.

Myosin-generated forces are a major player in traditional clutch-based models (7). However, since here we show that for T cells the treatment with a myosin-inhibiting drug has little impact on cell spreading on immobilized ligands (Fig. 2C), we identify actin polymerization-generated forces as the basis for a theoretical description of mechanosensitive spreading of T cells.

To mathematically describe cell spreading, we take a closer look at the cell edge, consisting of lamellipodia-like protrusions of the cell membrane. Spreading is effectuated as a result of actin polymerization and branching just behind the membrane at the edge of the cell (Fig. 4 A–C). Within a one-dimensional model of such a cell edge, the actin is modeled as a strip of unit length and width which polymerizes with velocity $v_p$ and pushes the edge of the cell forward, at the same time generating a retrograde flow of the actin away from the edge (Fig. 4C) (17). For a fully spread cell, the edge velocity is on average zero, and therefore the actin retrograde velocity is $v_p$. Here we make the assumption that $v_p$ depends on the nature and number of ligands alone. The value of $v_p$ can therefore be taken from independent experiments using immobile ligands on which LifeAct-labeled Jurkat cells were allowed to adhere and spread. In the presence of anti-CD3 alone (henceforth called the TCR case) this was measured to be about 25 nm/s, and in the presence of additional ICAM-1 (henceforth called the TCR+LFA-1 case), about 100 nm/s.

Fig. 4.

The model and fit to data. (A) Schematic representation of the model of T cell spreading as seen from the top. A portion of lamellipod is considered, and the direction of the force exerted by the actin flow is shown. (B) Mesoscopic side view of the cell edge showing ligand–receptor bonds that couple actin retrograde flow to the elastic substrate. (C) Molecular connections are replaced by a single effective linker, represented here at various states of binding and extension. The barbed end of the polymerizing actin filament pushes on the membrane and generates the retrograde flow. (D) Fit of model to normalized area data. Note that the data are reproduced from Fig. 1C (TCR, excluding two data points on viscoelastic PDMS; last black-encircled point on right is on glass) and Fig. 3C (TCR+LFA-1). (E) Force density in pascals. (F) Force on a single bond and the fraction of receptors that are bound.

The balance of forces acting on the actin in a lamellipod of unit size gives $F=T$, where $F$ is the friction force density and $T$ is a tensile force that is probably dominated by actomyosin tension at very low spreading and the membrane tension in a moderate to well-spread cell. Making the reasonable assumption that $T$ increases with cell area $A$, and opting for a simple linear dependence, $F=\gamma A$, where $\gamma$ is an unknown tension (17). We next compute the dependence of $F$ on the substrate elasticity $E$ and compare it with the experimentally measured $A$, both $F$ and $A$ being suitably normalized [each with respect to its values at a given elasticity (chosen as 5 kPa here)] (Fig. 4D).

$F$ is related to the molecular parameters of the ligand/receptor pairs through the force $f$ exerted on the individual linkers by the retrograde flow of actin, such that $F=Nnf$ where $N$ is the molecular density (here 400/$\mathrm{\mu }{\mathrm{m}}^2$ for TCR and 800/$\mathrm{\mu }{\mathrm{m}}^2$ for the TCR+LFA-1 case), and $n$ is the fraction of bound receptors that is determined self-consistently with $F$. Here, ligands are immobilized and, contrary to previous work using mobile ligands on SLBs (15, 17), are not displaced by cell-generated forces. They form specific bonds with cellular transmembrane receptors, whose intracellular moieties interact with actin via adapter proteins that are mostly known for integrins (24), but are still a matter of debate for TCRs. For simplicity, the entire complex, and its link to actin, is modeled here as a single spring-like bond (henceforth called a linker, Fig. 4C). The main frictional dissipation occurs at the receptor–actin link, such that $f=\eta v_p$, where $\eta$ is a frictional coefficient, which can be interpreted as arising from a bond kinetics defined by a constant on-rate $k_{on}$, a force-dependent off-rate $k_{o\hspace{0.17em}f\hspace{0.17em}f}$, and an effective bond stiffness $k_b$, and given by $\eta =\frac{k_b}{k_{o\hspace{0.17em}f\hspace{0.17em}f}}$ (25, 26). Importantly, $k_{o\hspace{0.17em}f\hspace{0.17em}f}=k_{{o\hspace{0.17em}f\hspace{0.17em}f}_0}{e}^{\frac{f}{f^B}}$, where $k_{{o\hspace{0.17em}f\hspace{0.17em}f}_0}$ is the off rate at zero force and $f^B$ is the characteristic force at which a bond becomes force sensitive (27). $k_b$ here is an effective stiffness that should account for the substrate as well as the linker with an intrinsic bond elasticity $k_b^0$. Following ref. 25, we write $k_b=\frac{k_b^0\cdot Ea}{k_b^0+Ea}$, where $a$ is a molecular length scale. $k_b$ increases with $E$, reaching a saturation at $k_b^0$.

The parameters $f^B$, $k_b^0$, $k_{{o\hspace{0.17em}f\hspace{0.17em}f}_0}$, and $k_{on}$ correspond to physical quantities for the TCR case and may, in principle, be measurable from single-molecule experiments (28, 29). For TCR+LFA-1, however, a single effective linker represents both TCR-mediated and LFA-1–mediated bonds—the attributed molecular parameters being effective quantities that embody the real parameters corresponding to each linker type as well as possible cooperative effects. In both cases, $F$ can be written as

$$F=\frac{Nn{k}_b^0\hspace{0.17em}Ea\hspace{0.17em}{v}_p}{\left(k_b^0+Ea\right)\hspace{0.17em}{k}_{{o\hspace{0.17em}f\hspace{0.17em}f}_0}\hspace{0.17em}{e}^{\frac{f}{f^B}}}.$$ [1]

Using the linear relation of $F$ and $A$, Eq. 1 is fitted to area data (Fig. 4C), to obtain the four molecular parameters. [$F$ is defined by three independent variables which, however, contain four physical parameters (SI Appendix). Here $k_{on}$ was fixed to 1 s⁻¹. The fit is robust to the precise choice of $k_{on}$.] For the TCR case, $k_{o\hspace{0.17em}f\hspace{0.17em}f}^0$ = 0.001 ${\mathrm{s}}^{-1}$, $f^B$ = 0.3 pN, and $k_b$ = 0.3 pN/nm; and for the TCR+LFA-1 case, $k_{o\hspace{0.17em}f\hspace{0.17em}f}^0$ = 0.03 ${\mathrm{s}}^{-1}$, $f^B$ = 0.4 pN, and $k_b$ = 0.002 pN/nm. Calculating the real force density (in units of pascals) shows that, as expected, it reflects the behavior of the normalized area and is within the range of expected cell forces (Fig. 4E).

The model strikingly captures both monotonic and biphasic behavior (Fig. 4D). To understand the physics behind this, we need to consider the dependence of $f$ and $n$ on $E$. For a given actin polymerization velocity $v_p$, the force per bond $f$ increases with $E$ due to the renormalization of $k_b$ (Fig. 4F), leading to an increase in the force density $F$ and therefore $A$. However, when $f$ becomes appreciably larger than $f^B$, bonds start breaking faster than they form and so $n$ drops, leading to a drop in $F$ and $A$, even though $f$ continues to increase with $E$. To display biphasic behavior therefore, the bonds need to be stiff and very sensitive to force (high $k_b^0$ and low $f^B$), ensuring that $k_{o\hspace{0.17em}f\hspace{0.17em}f}$ overtakes $k_{on}$ before $E$ reaches $k_b^0/a$. This should be the case for the TCR. In the case of TCR+LFA-1, however, $k_b$ saturates before $E$ reaches $k_b^0/a$. SI Appendix, Fig. S15 shows that independent of $k_{o\hspace{0.17em}f\hspace{0.17em}f}^0$, high $k_b$ and low $f^B$ are essential for biphasic behavior.

Discussion

Ligand/receptor-mediated cell mechanosensing is an ubiquitous process initiated by binding of cell surface receptors to their ligands on another surface. The mechanical cues presented via the ligand and transmitted through the bound receptor are ultimately converted to a cellular action, here read out as cell spread area. Strikingly, here the transmission does not appear to pass via a biochemical signal, and the stiffness dependence of the readout depends on the specific receptors that are solicited—TCR alone or TCR as well as LFA-1. It is well known that the cell surface TCRs need to be engaged to activate LFA-1, and conversely, ligation of LFA-1 to its ligand feeds back on TCR activation (30). The synergy between anti-CD3– and ICAM-1–induced adhesion/activation is strikingly demonstrated in the case of mobile ligands where the cells adhere and initiate actin polymerization but fail to effectively spread in absence of LFA-1 stimulation (17).

Importantly, a large body of work reports that tissue cells, such as fibroblasts that adhere principally via integrins, spread more on hard than on soft substrates (4–6, 31) and actomyosin-dependent clutch models were extensively used to explain this behavior (8, 32, 33). Incidentally, whereas a recent work found that cellular behavior may depend not only on substrate stiffness but also on porosity (34), the latter being different for elastomers and hydrogels (34, 35), we could recapitulate the biphasic trend on both types of soft supports.

Recently, to explain nonmonotonous mechanosensitivity in early spreading of fibroblasts, a myosin-independent mechanism was suggested that depends instead on integrin catch-bond kinetics (36). In the present case too, it may seem natural to attribute the biphasic switch to a catch-bond mechanism operative for the TCR in absence of stimulation of the integrin. However, the theoretical model here shows that a biphasic behavior in fact may also arise in a slip-bond system through the interplay of bond kinetics and elastic coupling to the substrate. We could speculate that as seen both by Oakes et al. (36) and by us, mechanosensing associated with early spreading may be actin polymerization driven and perhaps nonmonotonic with substrate stiffness. Theoretically, while early work linked friction force generated by polymerization with kinetics of force-insensitive bonds (26), more recently mesoscopic models have linked actin polymerization-driven cellular-scale spreading (17, 22) to force-dependent molecular parameters; separately, traction force generation at externally pulled filopodia (25) has also been linked to force-dependent bond kinetics.

An important point here is that addition of ligands of integrins abrogates the biphasic response. Earlier work already pointed to the separate but complementary role of the TCR and LFA-1 in force generation and actin organization (37–39). Coengagement of the TCR complex and LFA-1 is therefore not merely additive but synergistic in the sense that ligation of each species with its ligand has an impact on the activation and functioning of the other.

Now the question arises about the origin of the mechanosensitivity: Is the bond under question the ligand/receptor bond or is it the linkage to the actin? Interestingly, previous single-molecule probing of TCRs on T cells (40) yielded different force response compared with cell-free experiments with purified TCRs (41). Furthermore, the characteristic force $f^B$ for a typical ligand–receptor bond is expected to be in the range of several piconewtons. Antibody–antigen bonds as well as integrin-mediated bonds are expected to be particularly stable and have a high $f^B$. Here we find that $f^B$, at about 0.3 pN for the TCR and about 0.4 for TCR+LFA-1, is an order of magnitude smaller than that expected for most ligand–receptor bonds. This leads us to speculate that the highly force-sensitive bond that is responsible for the peculiar mechanosensing seen here is in fact the previously mooted CD3 to actin link (17, 42), rather than the ligand–receptor (anti-CD3/TCR) bond. On integrin engagement, this putative link is probably reinforced and the effective link is less force sensitive. Interestingly, measurements in the range of elasticity spanning up to a few kilopascals do report higher traction forces and activation on stiffer substrates (12, 18, 43). This range, roughly coinciding with the first rising phase of our biphasic curve, is also the physiological range of elasticity for professional antigen-presenting cells, which may be modified under pathological conditions (44).

Whether T cells exploit the potentially biphasic response of the TCR-CD3 complex for their physiological function remains to be explored, but the molecular link proposed here as highly force sensitive, which we hypothesize might be the TCR-complex to actin link, should inform future studies on T cell mechanoresponse.

Materials and Methods

Clean glass coverslips were spin coated with liquid PDMS (CY52-276, Sylgard 184, Dow Corning; or QGel 920, Quantum Silicones) mixed in the desired base/cross-linker ratio and were baked. Stiffness was measured using an atomic force microscope (45). The PDMS surface was functionalized successively with biotin-conjugated BSA (Sigma), Fluorescent Neutravidin (ThermoFisher), and finally with anti-CD3 (clone UCHT1, eBioscience). ICAM-1 conjugated with a 6-his tag (ThermoFisher) was linked via BSA-nitrilotriacetic acid (in-house coupling). Jurkat E6 T lymphocytes (clone E6-1, ATCC) were cultivated under standard conditions and were used at a concentration of 0.6 million cells/mL. Live cells were imaged every 5 min during 20 min in 100 adjacent fields (tiles). Cells were fixed after 20 min in 2% paraformaldehyde (Merck). Actin, TCR, and ZAP70 were labeled with either rhodamine–phalloidin or an antibody-conjugated fluorophore. Primary cells were purified and were used as above. TIRFM/RICM imaging was done with an inverted microscope (AxioObserver, Zeiss), equipped with a 100$\times$, 1.45-N.A. oil or a custom 100$\times$, 1.46-N.A. oil antiflex objective (Zeiss). Analysis was done with Fiji-ImageJ v1.49d or IGOR Pro (WaveMatrics) software. Averages are calculated from the number of cells (n) in pooled experiments (N). Student’s t test (bilateral distribution) is not significant (ns) when P $>$ 0.05 and significant when *P $<$ 0.05, **P $<$ 0.01, and ***P $<$ 0.001. SEM is calculated as the SD divided by the square root of n. Additional data and text can be found in SI Appendix.