If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
10 These authors contributed equally. 11 Present addresses: IRCCS Regina Elena National Cancer Institute, Rome, Italy 12 Present addresses: Enara Bio Oxford OX4 4GA, UK 13 Present addresses: Department of Pediatrics, Ludwigs-Maximilians Universität, Munich, Germany. 14 These senior authors contributed equally.
Theoretical work suggests that collective spatiotemporal behaviour of integral membrane proteins (IMPs) should be modulated by boundary lipids sheathing their membrane anchors. Here, we show evidence for this prediction whilst investigating the mechanism for maintaining a steady amount of the active form of IMP Lck kinase (LckA) by Lck trans-autophosphorylation regulated by the phosphatase CD45. We used super-resolution microscopy, flow cytometry, and pharmacological and genetic perturbation to gain insight into the spatiotemporal context of this process. We found that LckA is generated exclusively at the plasma membrane, where CD45 maintains it in a ceaseless dynamic equilibrium with its unphosphorylated precursor. Steady LckA shows linear dependence, after an initial threshold, over a considerable range of Lck expression levels. This behaviour fits a phenomenological model of trans-autophosphorylation that becomes more efficient with increasing LckA. We then challenged steady LckA formation by genetically swapping the Lck membrane anchor with structurally divergent ones, such as that of Src or the transmembrane domains of LAT, CD4, palmitoylation-defective CD4 and CD45 that were expected to drastically modify Lck boundary lipids. We observed small but significant changes in LckA generation, except for the CD45 transmembrane domain that drastically reduced LckA due to its excessive lateral proximity to CD45. Comprehensively, LckA formation and maintenance can be best explained by lipid bilayer critical density fluctuations rather than liquid-ordered phase-separated nanodomains, as previously thought, with “like/unlike” boundary lipids driving dynamical proximity and remoteness of Lck with itself and with CD45.
Cell responses to environmental cues initiate by events choreographed at the plasma membrane by integral membrane proteins (IMPs). IMPs are embedded in the membrane lipid bilayer via hydrophobic moieties (e.g., transmembrane domains) or covalently-bound lipids or combinations of both. IMPs are sheathed by lipids (called boundary lipids, or lipid shell) that allow for solvation in the lipid bilayer and can contribute to IMPs’ structure and function (
). Molecular dynamics simulations (MDS) provide an increasingly realistic representation at the molecular scale of IMPs’ boundary lipids and contribute to understand IMPs’ individual behaviour and lateral organisation (
) for optimising solvation and function. This combinatorial distribution of boundary lipids predicts that each IMP can be surrounded by a lipid fingerprint of unique physical and chemical properties. Such diverse assortment of IMPs’ immediate lipids in natural membranes is likely to impact on their IMPs’ thermodynamic parameters, including lateral interactions (
). These studies have lent support to models of biomembranes organised into liquid-ordered (Lo) phase-separated nanodomains buttressed by actin-regulated cortical membrane proteins and capable of trapping IMPs to regulate membrane functions (
). Lck enzymatic activity is controlled by the cytoplasmic-resident C-terminal Src kinase (Csk), by Lck autophosphorylation and by the IMP tyrosine phosphatase (PTP) CD45. Csk and CD45 are constitutively active (Fig. 1A) (
) (Fig. 1A). LckP is competent to autophosphorylate in trans Y394 in the activation loop of the kinase domain, a modification that promotes major allosteric changes resulting in LckA (pY394/Y505-Lck) (Fig. 1A). Structural studies predict that LckA possesses optimal enzymatic activity and access to substrates (
) (Fig. 1A), playing therefore the dual role of inducer and controller of LckA. LckA can be phosphorylated in part at Y505 (Fig. 1A), forming a pool of double-phosphorylated Lck (pY394-Lck/pY505-Lck or LckADP) (
) have suggested that Lck might be dynamically entrapped within Lo-phase-separated nanodomains (or raft). CD45 experiences instead random diffusion, occasionally halted by interactions with membrane cortex proteins (
). This scenario suggests that Lck is intermittently sequestered within Lo membrane rafts, where CD45 access is partially forbitten, hence favouring LckA formation and maintenance.
We investigated the validity of this model by genetically swapping Lck membrane anchor with structurally divergent ones borrowed from other IMPs, including single-pass helical transmembrane domains (TMDs) of bitopic IMPs. Such radical structural changes of the membrane anchor implied substantial alteration of Lck boundary lipids (
). Surprisingly, only small differences in steady LckA were observed. However, swapping Lck membrane anchor with that of CD45 drastically reduced LckA, due to augmented lateral proximity between Lck and CD45. We discuss how our data cannot be easily explained by Lo phase-separated membrane domains. However, steady LckA can be explained by well-grounded theoretical predictions, whereby boundary lipids modulate Lck lateral distribution without requiring phase-separated membrane domains.
Dynamic maintenance of steady LckA
We first assessed the spatiotemporal backdrop for the generation and maintenance of LckA, as schematised in Fig. 1A. Lck and CD45 quantitative subcellular distribution was examined in primary T cells and JCaM1.6 cells (a convenient T-cell surrogate model) reconstituted for Lck (hereafter referred to as JCaM1.6-Lck) by super-resolution microscopy using three-dimensional structured illumination microscopy (3D-SIM) (
) (for the advantages of using 3D-SIM, see Experimental procedures). Permeabilised primary T cells (Fig. 1B, upper panel) and JCaM1.6-Lck (Fig. 1B, middle panel) showed that CD45 staining (red) neatly defined the PM, with almost undetectable signal (< 3 %) in cytoplasmic membrane compartment (CP) (Fig. 1B and Fig. S1A). The demarcation of the PM at high resolution together with nuclear staining by DAPI (blue) conveniently framed the exiguous CP space (see enlargements in Fig. S1A) and allowed computing PM/CP ratios to obtain relative PM and CP distribution for Lck (see Experimental procedures for masks’ drawing). In T cells and JCaM1.6-Lck, PM/CP for Lck (green) scored ≈ 2.2-2.3 (Fig. 1B and negative control Fig. S1B, upper panel), indicating that ≈ 70 % of total Lck (LckT) is PM-resident. CP detection of Lck (Fig. S1A, upper panel) was presumably associated with Golgi and recycling compartments (
). As expected, a mutant lacking the membrane anchor, LckΔSH4 (Fig. S1A, lower panel), was mostly in the CP and scored PM/CP of 0.6 (Fig. 1B, bottom panel and histogram and enlargement in Fig. S1A). Membrane unevenness, spatial resolution limits and weak interaction of Lck modular domains with the PM (
See also: Molecular dynamics simulations reveal membrane lipid interactions of the full-length lymphocyte specific kinase Lck. D.Prakaash, G.P. Cook, O. Acuto and Antreas C. Kalli. BioRxiv, doi: https://doi.org/10.1101/2022.05.10.491278
may explain the non-null score for LckΔSH4. The almost exclusive PM staining of CD45 helped tracing a reliable mask for ImageStream, which has lower resolution than 3D-SIM but higher statistical robustness (10,000 events recorded). ImageStream detected in JCaM1.6-Lck ≈ 80 % of Lck as PM-resident (Fig. S1C, see Methods for details), in good agreement with 3D-SIM (Fig. 1B) and previous estimates of Lck subcellular distribution (
). The virtually exclusive PM localisation of CD45 indicated that this compartment is likely to be where LckI is dephosphorylated at pY505 to be converted into LckP, and also where LckP autophosphorylation in trans at Y394 generates LckA (Fig. 1A) and where CD45 dephosphorylates LckA at pY394 (
) to reverse it to LckP (Fig. 1A). The net output of this natural condition in unperturbed T cells should be a steady pool of PM-resident LckA. Remarkably, this pool is established despite CD45:Lck stoichiometric ratio being ≈ 10:1 (
), a condition that could annihilate LckA, unless partially protected from CD45 action.
To investigate further the molecular basis of this natural setting, we used anti-pY416-Src Ab staining that recognizes pY394 and allowed to quantitate by 3D-SIM and flow cytometry (FCM) LckA subcellular localisation and dynamic equilibrium. Anti-pY416 reliability for detecting specifically LckA in 3D-SIM (Fig. S1B) and FCM (Fig. S1D and Fig. S1E) was corroborated by various controls (for details, see Experimental procedures). 3D-SIM showed a PM/CP ratio of LckA in T cells and JCaM1.6-Lck of 2.0 and 2.5 (Fig. 1C), respectively, indicating that ≈ 66 -71 % of LckA is PM-resident. CP-resident LckA (Figs. 1C and S1B) is presumably in a recycling compartment (
), Csk and ZAP-70, respectively (Table S1). FCM showed that blocking Lck activity by A770041, hence the autophosphorylation at Y394 in trans, reduced anti-pY416 staining to background level (Fig. S1E) due to the CD45 constitutive activity that negatively controls pY394 (
), this corresponds to a conversion of ≈ 4 LckA molecules into LckP per ms, revealing the rapid turnover of Y394 phosphorylation controlled by the opposite action of CD45 and Lck. Consistent with this idea, CD45 inhibition by catalase-treated pervanadate (PV) rapidly increased LckA by 50 % up to a ceiling (Fig. 1E). This revealed the presence of a PM-resident pool of LckP being ≈ 50 % of LckA and ≈ 30 % of total PM-Lck, in close agreement with previous estimates (
). In contrast, LckΔSH4 formed only negligible amounts of LckA as compared with intact Lck (cf. Figs. S1F and S1D, right panels), with a small percentage of LckA positive cells with much lower fluorescence intensity per cell. Together, these data indicated that most, if not all LckA must originate at the PM, where > 97 % of CD45 resides.
Surprisingly, A770041 reduced also pY505-Lck by ≈ 60 % of (Fig. 1F). Since A770041 cannot inhibit Csk (Table S1), these data indicate that a considerable proportion of PM-resident pY505-Lck must be produced by Lck itself and not by Csk. This occurs presumably by trans-autophosphorylation of LckA at pY505 to yield double phosphorylated Lck isoform (LckADP) (Fig. 1A), consistent with in vitro or in cellulo data that Lck (
). LckADP belongs therefore to the PM pool of LckA but its functional role was not explored as beyond the scope of this investigation. Fig. 1A illustrates the commonly held notion that Csk keeps Src-family kinases inactive at the PM by directly opposing a membrane phosphatase. However, according to this model, A770041 should have increased and not reduced Lck-pY505 as we observed (Fig. 1F). These data suggested therefore that the proportion of PM-resident LckI, presumably in dynamic equilibrium with LckP and LckA, should be considerably lower than previously thought. Consistent with this prediction, 3D-SIM revealed that, contrary to LckA, PM/CP ratios of pY505-Lck in T cells and JCaM1.6-Lck scored only 0.7 and 0.8, respectively (Fig. 1G and see Fig. S1G for detection of pY505 by FCM and Fig. S1H for anti-pY505 Ab specificity control). Moreover, pY505 PM/CP ratio for LckΔSH4 was only slightly lower than wild type Lck (Fig. 1G). These data indicate that a sizable proportion of PM-resident pY505-Lck is generated by LckA, and not by Csk (Fig. 1F). These observations lessen the role of the Csk in opposing LckA generation at the PM and in its contribution to LckP ⇌ LckA equilibrium. Csk would therefore primarily control Y505 in the CP, keeping Lck in check as LckI, presumably in exocytic compartments en route to the PM (Figs. 1G and S1I).
Fig. S1I shows a summary scheme of Lck isoforms cellular localisation and regulation in unperturbed cells, as suggested by our data. It highlights that the PM is the primary site where LckI incoming from the CP membrane compartments is largely converted into LckP by CD45 almost unopposed by Csk. The PM appears therefore as the compartment where most, if not all LckA and LckP reside in a highly dynamic equilibrium governed by Lck trans-autophosphorylation and CD45 dephosphorylation at Y394. Our data suggested also the existence of an underlying mechanism that allows Lck to partially elude CD45 overwhelming activity in order to ensure LckA generation and steady maintenance.
LckA dependence on LckT
Testing the general validity of these conjectures required an accurate quantitation of LckA as a function of LckT input in intact cells. To this purpose, we set up a two-colour FCM-based assay that concomitantly detected and quantitated with LckA and LckT with high accuracy (Fig. 2A). See “Two-colour FCM for LckA vs. LckT 2D plots” in Experimental procedures for assessing anti-LckT Ab epitope mapping (Fig. S2A), anti-LckT and anti-LckA Abs specificity (Figs. S2B, S2C, S2D and S2E) as well as the procedure to extract LckA and LckT fluorescence values to obtain the line of best fit (Fig. 2B). Consistently, this assay showed a direct dependence of LckA on LckT (Fig. 2B, right panel). The line of best fit showed two components in the 2D plot (Fig. 2B, right panel). At low LckT concentration, LckA formation was less than proportional to Lck input that fitted a second-order function, whereas at higher LckT concentration LckA increase was linear (Fig. 2B, right panel). This trend could be explained by Lck trans-autophosphorylation being accomplished more efficiently by LckA ⇔ LckP interaction as compared with LckP ⇔ LckP (
), respectively (Fig. 2C), the latter becoming less significant when LckA become >> LckP. The linear trend of LckA vs. LckT indicated that CD45 constitutive activity was not regulated by a LckA-driven feedback mechanism and was overly robust as it was able to rapidly revert a large fraction of LckI to LckP and of LckA to LckP, at low and high Lck levels of expression (see also next chapter). This suggested that CD45 activity might be a hidden variable in the LckP ⇌ LckA dynamic equilibrium. The validity of these assumptions was tested by a numerical simulation of a simple phenomenological model. The model assigned a probability (P) of converting LckP to LckA from reaction (
) PAA (Fig. 2C) with P allowed to vary between 0.1 and 1.00 (with incremental steps of 0.05) (Fig. 2D, and see Experimental procedures for details of the modelling). We found that the best fit (p < 10-5) of the simulation to the experimental data was obtained for PPA and PAA of 0.3 and 0.1, respectively (Fig. 2D and insert). This result agrees with LckA generated more efficiently by LckA ⇔ LckP than by LckP ⇔ LckP, with increasing Lck concentration. Importantly, this data did not conflict with the scheme of Fig. S1I. Independently of potential differences in structural details of trans-autophosphorylation for LckA⇔ LckP or LckP ⇔ LckP pairs explaining the two regimens of LckA generations (see Discussion), the modelling generally agreed with the supposed spatiotemporal membrane context where Lck and CD45 operate, as depicted in Fig. 2E. It shows a qualitative model of a ceaseless “Lck cycle”, in which LckA and LckP are in dynamic equilibrium maintained at the PM by the antagonism of CD45 and Lck for Y394 phosphorylation, with CD45 continuously igniting, rescinding and refuelling LckA formation. As alluded above, LckA formation might require a Lo phase-separated membrane nanodomain (or raft) (Fig. 2E). To verify this hypothesis experimentally, we asked whether LckA output varied upon moderate or drastic changes of Lck hydrophobic anchor, hence of its immediate lipid environment.
Subcellular distribution of Lck with non-native membrane anchors
Myristoylation and di-palmitoylation at LckSH4 (Fig. 3A) provide attachment of Lck to the inner leaflet of the PM (
). Thus, swapping LckSH4 with structurally diverse IMPs’ membrane anchors, including removal of palmitoylation, should inform about the role of Lck-contiguous lipid milieu required for LckA formation and maintenance. To test this idea, Lck lacking SH4 (LckΔSH4) was fused to disparate membrane anchors (Fig. 3A). SrcSH4 was chosen as it is myristoylated but not palmitoylated and, contrary to LckSH4, SrcSH4 contains several basic residues (Table S2). We selected also the helical TMDs of the bitopic membrane proteins LAT and CD4, both featuring two palmitoylation sites, and a palmitoylation-defective CD4 TM mutant (CD4C/S). These membrane anchors diverged for lipid adducts, amino acid composition, sequence, length and membrane-juxtaposed segments (Table S2). Consequently, they should considerably alter the composition and topology of the natural Lck immediate lipid milieu (
), making unlikely that they could favour Lck ⇔ Lck via TMD-dependent protein-protein interactions. The three residues-long extra-cellular sequence of LAT was added to each helical anchor to facilitate similar expression of the Lck chimeras. All chimeras were expressed similarly to Lck (Fig. 3B), with only SrcSH4-Lck expressing about twice as much and all cell lines maintaining identical amounts of endogenous CD45 (Fig. S3A). PM/CP ratios determined by 3D-SIM for LAT-Lck, CD4-Lck and CD4C/S-Lck chimeras (Fig. 3C) indicated them to be very similar to native Lck. Only SrcSH4-Lck showed a PM/CP of about 1.00 (i.e., even PM and CP distribution), perhaps reflecting Src higher propensity to localise in recycling membranes (
). However, SrcSH4-Lck reduction at the PM should be compensated by its higher expression (Fig. 3B), resulting in PM-resident SrcSH4-Lck absolute amount similar to the other chimeras. Thus, all non-native membrane anchors conferred PM residency similar to native Lck, guaranteeing a fair comparison of their capacity to form LckA.
Moderate impact of different membrane anchors on LckA formation
To augment robustness and precision in detecting differences in LckA, we bar-coded and mixed together before dox-induction two cell lines expressing each a different chimera and one expressing native Lck (Fig. S4A and Experimental procedures). For every chimera, LckA increased linearly even at LckT expression ≥ 10-fold higher than in Cln20 (blue box superimposed to 2D FCM in Fig. 4A and S4B), indicating a considerable reservoir of CD45 enzymatic activity to effectively oppose increasing LckI and LckA. Such LckA scalability made also less likely the existence of a potential PM-resident regulator, such as a dedicated membrane scaffold protein, which should be expected to be a limiting factor. LckT and LckA increase did not correlate with cell size (Fig. S4C), excluding that their increase per cell basis did not reflect mainly cell size. We restricted our analysis of LckA generation for LckT values of Cln20, as this was considered physiological and was less penalising computationally and more robust statistically (see Experimental procedures). 2D FCM plots were densely binned and the values of LckA for each LckT bin extracted within the LckT range of Cln20 (Figs. S4A and 4B, left panels and Experimental procedures) and subjected to best fit line regression analysis (Fig. 4B, right panels and Experimental procedures). Surprisingly, the data showed only small differences in LckA formation by SrcSH4-Lck, LAT-Lck (Fig. 4B upper panels), CD4-Lck and CD4C/S-Lck (Fig. 4B bottom panels), as compared to native Lck. Regression analysis showed that none of the curves reporting LckA generation by the Lck chimeras was overlapping with native Lck and with each other (Fig. 4B, right panels), indicating that such relatively small differences in LckA were significant. Similar results were obtained by plotting LckA normalised to LckT for each bean (LckA/LckT vs. LckT plots in Fig. S4D), that better captures the two regimens of LckA yield at low and high LckT, as observed for Cln20. Predictably, LckΔSH4 showed severely reduced LckA (Figs. S1E, 4C and S4E), despite being expressed at higher amounts than Lck (Fig. S4F) and for equal CD45 expression (Fig. S4G), consistent with LckΔSH4 being not PM-anchored and therefore escaping CD45 regulation required to generate LckP (Fig. S1I). Notably, palmitoylation was not essential, nor provided an advantage for LckA generation. If anything, LAT-Lck and CD4-Lck performed slightly worse than native Lck (Fig. 4B) and Src-Lck and CD4C/S-Lck that are not palmitoylated (Fig. 4B). The similar behaviour of the Lck chimeras was unexpected in view of the substantial physicochemical divergence of the hydrophobic anchors. One explanation could be that highly different membrane anchors provide Lck with similar trapped diffusion within distinct phase-separated (rafts) nanodomains and result in apparently similar lateral behaviour. Alternatively, LckA might form independently of membrane rafts. In this scenario, direct protein-protein interaction would dominate Lck interactions with itself and with CD45, with their respective immediate lipid environment playing a mild modulatory effect. Being both explanations unsatisfactory (see Discussion), we sought to test an alternative hypothesis that could provide a more adequate explanation of these apparently puzzling results.
Impact of Lck membrane anchor on lateral interactions
To provide a plausible explanation for our data, we considered an alternative model of IMPs lateral behaviour that does not necessarily require IMPs trapping in Lo phase-separated nanodomains. Theoretical studies, including MDS (
), indicate that the boundary lipids surrounding IMPs have an average composition and spatial arrangement distinct from bulk lipids and from IMPs with different membrane anchors. This condition can reduce miscibility of boundary lipids of different IMPs, implying the presence of free-energy barriers theoretically estimated to be of few Kcal/mole, comparable to or larger than the thermal energy (
) and therefore unlikely to result in phase-separation of IMPs. Such barriers should reduce the likelihood of IMPs dynamical lateral proximity, without however forbidding it. However, energy barriers should be much lower or even vanishingly small for identical IMP’s anchors (i.e., identical boundary lipids). According to this proposition, the probability of dynamical self-proximity for Lck chimeras and for native Lck should be similar, despite highly divergent hydrophobic anchors (i.e., boundary lipids) so to achieve similar trans-autophosphorylation ability (i.e., LckA formation). However, this should be less so for LckA maintenance which depends on some level of dynamical remoteness from CD45, which can be ensured by the structural divergence between the anchors of CD45 and Lck or Lck chimeras tested. Such condition would result in small but significant differences of steady LckA (even of different sign) as observed for the Lck chimeras (Figs. 4A - C). A distinctive prediction of this idea is that Lck endowed with CD45 TMD (CD45-Lck) (Fig. 5A) should exhibit trans-autophosphorylation capacity (i.e., LckA generation) similar to native Lck, despite CD45 TMD having no propensity for trapped diffusion in a Lo phase-separated lipid nanodomain (
). However, CD45-Lck should have a higher likelihood of dynamic proximity to endogenous CD45 and consequently experience reduction or annihilation of steady LckA. To test this prediction, LckΔSH4 was fused to CD45 helical TMD (CD45-Lck) (Fig. 5A and Table S2) and conditionally expressed in JCaM1.6 at similar levels as native Lck (Fig. S5A). 3D-SIM for CD45-Lck showed a PM/CP ratio of 1.7 (Fig. 5B), only slightly lower than native Lck (i.e., 63 % vs. 68 % PM-resident for CD45-Lck and Lck, respectively). In agreement with the above prediction, CD45-Lck yielded drastically lower LckA formation than native Lck (and the other Lck chimeras) and was virtually indistinguishable from LckΔSH4 (Figs. 5C and S5B), which presents in our experimental system a bare minimum of LckA generation though for opposite reasons. Expression of endogenous CD45 was identical to cells expressing native Lck (Fig. S5C), excluding that changes in CD45 explained LckA reduction. To test the prediction that the striking reduction of LckA was due to accrued capacity of endogenous CD45 to dephosphorylate CD45-LckA, and not to defective LckA formation by CD45-LckA, we acutely inhibited CD45 enzymatic activity by PV. This showed that PV induced immediate recovery of CD45-LckA (Fig. 5D) and is schematised in Fig. 5E. LckA increment induced by PV for native Lck and CD45-Lck above their respective basal LckA values reached similar levels (Fig. 5D), further excluding alterations of CD45-Lck trans-autophosphorylation ability. Thus, CD45-Lck can accomplish trans-autophosphorylation but it experiences a dephosphorylation rate of pY394 by endogenous CD45 considerably higher than native Lck. Note that PV treatment showed poor recovery of LckA for LckΔSH4 (Fig. 5D), indicating different causes for reduced LckA of CD45-Lck and LckΔSH4, namely, poor trans-autophosphorylation capacity and accrued dephosphorylation by CD45 rate, respectively. Thus, an apparently simple rule for dynamical lateral proximity and remoteness driven by membrane anchor identity and divergence, respectively, can explain our data (see Fig. 5E).
Our quantitative appraisal of CD45 and LckA subcellular location and of LckA steady maintenance provides a spatiotemporal view of LckA origin and persistence in unperturbed T cells and compellingly suggests that LckA arises from highly dynamical interactions of Lck with itself and CD45 (Fig. S1I). Specifically, CD45 constitutive activity initiates and maintains at the plasma membrane a self-perpetuating LckA precursor-product cycle, almost unopposed by Csk. To consolidate this model, we conceived an FCM-based assay, whose data fit to an empirical model indicating the occurrence of two possible trans-autophosphorylation reactions, one being favoured and prevailing with increasing Lck. The crystal structure of a dimer of IRAK4 unphosphorylated (inactive) catalytic domain shows one partner to be in a stereochemical configuration that mimics phosphorylation in trans of the other partner (
). This example suggests a plausible configuration for LckP ⇔ LckP trans-autophosphorylation. However, this configuration must be different from that of LckA ⇔ LckP, in which accommodation of tyrosine Y394 of LckP into catalytically active site of LckA (
) should be favoured, making trans-autophosphorylation in LckA ⇔ LckP to proceed more efficiently than in LckP ⇔ LckP. Hence, accumulation of LckA over LckP should prevail with increasing Lck and result in an overall augmented Lck trans-autophosphorylation with Lck increase as our data indicate. The linear correlation between LckT and LckA with increasing LckA, is incompatible with CD45 being regulated by a LckA-dependent feedback loop. Rather, the considerable dynamic range of LckA generation indicates a formidable capacity of CD45 to convert LckI into the LckP, the precursor of LckA, and to control LckA over a wide scale of Lck expression. This setting makes CD45 formally a hidden variable not made explicit in our phenomenological model.
The overwhelming power of CD45 activity begged the question as whether LckA generation and/or maintained occurred in a specialised lipid environment of the PM where Lck could be dynamically segregated. Drastic changes in Lck membrane anchor would necessarily change Lck boundary lipids and alter its dynamic location into such specialised environment. We found a surprising tolerance of Lck regulation to those changes, as the Lck chimeras generated LckA steady levels similar, though not identical to native Lck. Allegedly, these results suggested that Lck membrane anchor, and consequently its immediate lipid environment plays only a modest, if any modulatory role in LckA formation and/or maintenance. In this scenario, Lck regulation in unperturbed cells should largely rely on differential rates of protein-protein interaction and of catalysis for Lck ⇔ Lck and Lck ⇔ CD45 interactions. However, if so the CD45-Lck chimera should behave similar to the other Lck chimeras. The apparent odd behaviour of CD45-Lck was anticipated by considering instead that boundary lipids do play a key role for highly dynamical lateral interactions of IMP such as for enzyme/substrate. This proposition was based on the intuitive idea that both Lck ⇔ Lck and Lck ⇔ CD45 interactions could be also governed by a simple “like/unlike” rule of their respective boundary lipids, akin to the “like-like/like-unlike” rule applied to phase separation in lipid bilayers (
) who found that Lck tyrosine phosphorylation and TCR-proximal signalling were vigorously inhibited in T cells expressing the intracellular domain of CD45 anchored to the PM via Lck-SH4, - i.e. CD45 and Lck shared the same membrane anchor. This swap of membrane anchors is symmetrical to the one made in our investigation - i.e., Lck anchor appended to CD45 and vice versa – and yielded very similar results. More generally, Tsien and co-workers (
) found that mutated GFP and YFP (mGFP and mYEF), which cannot form dimers in solution, exhibited Förster Resonance Energy Transfer, (FRET), (i.e., requiring no protein-protein direct contact by proximity of a few nm) when anchored to the PM via the same membrane anchor, being either dual-acylation or prenylation. However, FRET was markedly reduced when mGFP and mYEF were membrane-anchored by dual-acylation and prenylation, respectively, and vice versa (
). These and our studies agree in that membrane anchor likeness and unlikeness can confer to IMPs a probability of lateral proximity and remoteness, respectively, with presence or absence of protein-protein interaction being not a prerequisite to observe such a lateral behaviour. Both earlier studies concluded that each lipidated membrane anchor conferred bestowed confinement (i.e., concentration) in the same or different Lo membrane raft, favouring therefore proximity or remoteness, respectively (
However, our data showed that membrane anchor palmitoylation is not necessary for steady LckA formation. Moreover, the considerable scalability of steady LckA generation by Lck or Lck chimeras (> 1.5 orders of magnitude above physiological Lck levels (Fig. 4A) were difficult to reconcile with Lo membrane domains being mandatory for LckA generation. Such an important scalability entails the unlikely scenario of a PM populated by different subsets of Lo phase-separated membrane nanodomain, each one represented in high numbers and endowed with similar efficacy of trapping Lck or different Lck chimeras and excluding CD45. Alternative mechanisms can explain ours and previous observations (
From a theoretical perspective, different physical mechanisms can account for membrane lateral organisation at the nanometric scale under conditions of thermodynamic equilibrium. Those agreeing best with experimental observations are related to phase separation of a membrane molecular mixture characterized by a de-mixing critical point (see Supporting information and Fig. S6A) discussed for example in (
). The raft hypothesis posits that below the critical temperature (Fig. S6A), stable, relatively long-lived ∼ 100 nm nanodomains gather specific lipid and protein species (Fig. S6B). This is called the strong segregation limit (
), occurs above, though close enough to the critical temperature (Fig. S6A). It stipulates that more diffused and elusive density fluctuations of lipid and protein species suffice to promote some molecular encounters while making others less probable, consequently giving rise to membrane organisation. Criticality has been observed in realistic membrane mixtures, such as giant plasma membrane vesicles (GPMVs) (
) (see Supporting information). Since our data suggest that the first mechanism is less likely, we favour the second one as a plausible alternative to rationalise the role of boundary lipids in Lck and CD45 lateral interaction. As explained in more detail in the Supporting information, critical density fluctuations lead to the formation of transient nanodomains of molecular composition different from the bulk. The typical size of these nanodomains is set by the so-called correlation length (ξ), much larger than the molecular scale (Fig. S6C). If an IMP has a marked energetic preference for the lipid phase constituting these fluctuating domains, it acts as a condensation nucleus that gives rise to a long-lived lipid annulus around it, the lateral size of which is set by ξ (Fig. S6C). Two IMP anchors that localise in “like” and/or miscible boundary lipids will tend to encounter with a higher probability because this condition reduces the interfacial energy cost at the external boundary lipids (
). In contrast, if they localize in “unlike” and poorly miscible boundary lipids, their close encounter will be less probable. Figure 6 illustrates a simplified view of these two situations applied to Lck and CD45. A fundamental difference with phase separated domains is that such a mechanism can explain why so disparate membrane anchors do not impede formation of LckA (i.e., accomplish similar trans-autophosphorylation and CD45 avoidance). Even though this idea will have to be confirmed by additional experiments in the future, our observations are fully compatible with these theoretical predictions, whereas the more traditional raft theory hardly accounts for them.
From a molecular perspective, experimental and theoretical data (e.g., MDS of IMP-containing lipid bilayers) support the idea that different IMPs are surrounded by different lipid annuli or “lipid fingerprints” to minimise free energy of solvation. This multilayer sheath of a few nm exhibits spatial distribution and dynamics distinct from bulk-solvent around the IMP (
). The structure and dynamics of a lipid fingerprint surrounding IMPs necessarily leads to an interaction energy between them, determined by the sign and value of lipid mixing free energy, resulting from the competition between lipid-lipid affinities and mixing entropy (
), nonetheless they are sufficient to reduce, though not abolish IMPs close proximity for immiscible boundary lipids (Fig. S6C and 6). Conversely, two IMPs exhibiting the same boundary lipids (i.e., each and every IMP with respect to itself) should experience a moderate attractive interaction resulting in a higher probability for dynamic proximity (Fig. S6C and 6). This general property could prime formation of IMP short-lived homo-clusters eventually reinforced by specific protein-protein interactions when proteins arrive at contact.
In the context of our data, it is interesting to note that recent studies have shown that Ras alone forms dimers without direct protein-protein interaction (
). Lack of experimental evidence for the exact nature of the bouquet of boundary lipids of different IMPs prevents predicting the free-energy landscape that modulates IMP lateral proximity and distancing. Determination of the chemical composition of boundary lipids remains a difficult technical challenge. Recent progress in MS-based lipidomics of IMPs in native nanodisks (
) are promising avenues for experimentally define lipid fingerprints. Such knowledge, together with powerful MDS settings should allow to calculate free-energy differences between different boundary lipids.
Comprehensively, our data suggest that remoteness and close proximity of Lck and CD45 is modulated by their immediate lipid environment in order to generate the “right” amount of steady LckA required for effective T-cell activation.
Cell lines were maintained at 37 ⁰C with 5 % CO2 in a humidified incubator (Heraeus). Human embryonic kidney epithelial Lenti-X293T (Clontech) cells were cultured in complete DMEM (Sigma Aldrich) supplemented with 15 % foetal bovine serum (FBS) (Clontech). Jurkat cells were used as a convenient T-cell surrogate. Jurkat Clone 20 (Cln20) (
), a Lck-deficient Jurkat cell variant (Cln20 and J.CaM.1 are both CD4- and CD8-negative) and JCaM1.6-derived cell lines were cultured in RPMI 1640 supplemented with 10 % FBS up to maximum concentration of 3 - 4 x 105 cells/ml. JCaM1.6-derived cell lines with tetracycline-inducible gene expression system were maintained in RPMI 1640 supplemented with 10 % tetracycline-negative FBS (Clontech). Cells were routinely tested and found negative for mycoplasma and were not STR profiled but were routinely checked by FCM for specific cell surface markers. Primary human CD4+ T cells (> 95 % pure) were isolated by negative selection from whole blood of healthy donors (National Blood Service, Bristol, UK) using the Dynal CD4 negative isolation kit (Thermo Fisher). Cells were routinely maintained in culture overnight (ON) in RPMI-1640, 10 % FBS before being used for experiments. For Lck inhibition, cells were treated with 2 or 5 μM A770041 (Axon) at 37 ⁰C for 30 sec, 1 min or 5 min, as specified in the corresponding figure legend. For protein tyrosine phosphatase (PTP) inhibition, cells were treated at 37 ⁰C for 1 or 3 min with 100 μM catalase-treated pervanadate (PV), as specified in the corresponding figure legend.
Antibodies and reagents
Rabbit anti-Lck mAb-PE (73A5) mAb, rabbit anti-pY505-Lck (#2751) and rabbit anti-pY416-Src (#2101) polyclonal Abs were from Cell Signaling Technology (CST). Rabbit anti-Lck (NBP1-85804) was from Novus Biologicals; mouse anti-pY505-Lck mAb-PE (BD Biosciences); rat anti-human CD45 (YAML 501.4) Ab (Santa Cruz Biotechnology); mouse anti-human CD45-AF647 (HI30) mAb (BioLegend). For FCM and 3D-SIM Abs were: AlexaFluor 647 goat anti-rabbit IgG; AlexaFluor 594 donkey anti-rat IgG and AlexaFluor 488 goat anti-rabbit IgG and (Thermo Fischer). A770041 (Axon Medichem), Sodium Orthovanadate (Vanadate) New England BioLabs (NEB), catalase and hydrogen peroxide (30 %) from Sigma-Aldrich.
Catalase-treated pervanadate (PV) solution was freshly prepared prior to each experiment as previously described (
). Briefly, PV stock solution (1 mM) was prepared by adding 10 μl of 100 mM Sodium Orthovanadate and 50 μl of 100 mM hydrogen peroxide (diluted from a 30 % stock in 20 mM HEPES, pH 7.3) to 940 μl of H2O. Reagents were gently mixed and incubated for 5 min at room temperature (RT). Excess of hydrogen peroxide was removed by adding 200 μg/ml of catalase and the resulting solution was used shortly after to minimize decomposition of the vanadate-hydrogen peroxide complex.
Specificity controls of Abs used for FCM and 3D-SIM
The specificity of the anti-pY416, anti-pY505 Abs has been extensively tested previously for immunoblot and for tissue staining (
). Here, we analysed further the reliability of the aforementioned Abs and of anti-Lck 73A5 for flow cytometry and/or 3D-SIM. Induced or non-induced JCaM1.6 cells expressing Lck were stained either by rabbit anti-Lck 73A5-PE (FACS analysis) or rabbit anti-Lck (NBP1-85804, 3D-SIM) or rabbit anti-pY416 polyclonal Ab (FACS and 3D-SIM) or rabbit anti-pY505 (3D-SIM) or mouse anti-pY505-Lck mAb-PE (FACS analysis), followed when necessary by secondary anti-rabbit AF-647 Ab. Fig. S1B and S1D shows that anti-Lck 73A5-PE mAb, rabbit anti-Lck (NBP1-85804) polyclonal Ab and pY416 polyclonal Ab exclusively reacted with dox-treated cells, which specifically express the Lck protein by 3D-SIM and FACS respectively. Furthermore, Fig. S1E shows that the reactivity of anti-pY416 Ab, which specifically recognises pY394 of Lck in immunoblot (
), was lost after treatment of the induced cells with 2 μM A770041 or when the Ab was previously incubated with a synthetic peptide containing phospho-Y394. Similar controls for the anti-pY505 Ab are shown in Figs. S1G and S1H.
Immunostaining and 3D-SIM image acquisition and analysis
Initial experiments showed that 3D-SIM super-resolution microscopy improved segmentation at regions of interest for PM and CP and confidence for a quantitative assessment of sub-cellular distribution of Lck and CD45. This is because 3D-SIM doubles lateral and axial resolution (i.e., 8-fold in x, y, z) and considerably enhances image contrast over conventional fluorescence microscopy (
). For 3D-SIM, single-cell suspensions were immobilized on poly-L-lysine (Sigma-Aldrich)-coated high No. 1.5H precision glass coverslips (Marienfeld-Superior) in PBS containing CaCl2 and MgCl2 for 15 min at 37 ⁰C, in a cell culture incubator. Cells were fixed for 10 min with 4 % formaldehyde/PBS at 37 ⁰C and washed once with PBS. In a few experiments, BD PhosFlow Fix Buffer (BD Biosciences) was used and similar results were obtained. Permeabilization was performed with ice-cold 0.1% Tx-100, 0.5% (bovine serum albumin, Sigma) in PBS for 5 min and washed once with PBS. After blocking with PBS/1 % BSA for 15 min, cells were stained for 1 h at RT with rabbit anti-Lck Ab (NBP1-85804) 1:100 for Jurkat and 1:50 for primary human CD4 T cells. Rat anti-human CD45 Ab (YAML 501.4, SC) at 1:100 for both Jurkat and primary CD4 T cells. Anti-pY416 (rabbit) (CST) was diluted 1:100 and 1:50 for Jurkat and primary human CD4 T cells. Mouse anti-pY505 (BD) was diluted 1:50 for Jrkat and primary human CD4 T cells. Fluorochrome-conjugated secondary antibodies: AlexaFluor 594 donkey anti-rat IgG and AlexaFluor 488 goat anti-rabbit IgG Alexa were added for 1 h. Nuclei were counterstained with 1 μg/ml DAPI (Sigma-Aldrich) and coverslips were mounted to microscopy slides with ProLong Gold anti-fade reagent (Thermo Fisher). 3D-SIM was performed on an OMX V3 Blaze microscope (GE Healthcare) using 405-, 488- and 592-nm laser lines and a 60x/1.42 oil UPlanSApo objective (Olympus). Multi-channel images were captured sequentially by sCMOS cameras (PCO). 1 μm stacks were acquired at 125 nm z-distance, with 15 raw images per plane (three angles, five phases) resulting in 120 raw images in total, for each sample. Calibration measurements of 0.2 μm diameter TetraSpeck fluorescent beads (Thermo Fisher) were used to obtain alignment parameters subsequently utilized to align images from the different colour channels. Image stacks were computationally reconstructed from the raw data using the SoftWoRx 6.0 software package (GE Healthcare) to obtain super-resolution image with a resolution of wavelength-dependent 100-130 nm in x and y and 300-350 nm in z. Raw and reconstructed image data quality was confirmed using SIMcheck ImageJ plugin (
). Image processing and evaluation was performed using in-house ImageJ scripts: 32-bit reconstructed image stacks were thresholded to the modal intensity value (defining the centre of noise) and converted to 16-bit composites. The central four image planes were then average projected and Gaussian blurred (sigma 3 pixel). Regions of interest (ROI) covering the nuclear and plasma membrane (PM) were defined by “Otsu” auto-thresholding in the DAPI and anti-CD45 channel, respectively, and applying further processing steps (“Binary mask”, “Fill holes” and “Erode”). The area between the PM and nuclear ROI was defined as the cytoplasmic ROI. Measurements of the average fluorescence intensity within the respective PM and cytoplasm ROIs were used to calculate the plasma membrane/cytoplasm (PM/CP) ratios for the staining of anti-Lck, anti-Src, anti-pY416 and anti-pY505 antibodies. Lck subcellular localisation observed using the cell fixation and permeabilization procedure described above for 3D-SIM and for ImageStream (see below) were very similar to the subcellular localisation reported previously in live primary T cells using Lck-GFP (
). This indicates that our protocols for cell fixation and permeabilization do not significantly modify the native sub-cellular distribution of Lck. Note that experiments comparing Lck wild type subcellular localization in JCaM1.6-Lck and chimeras/mutants were performed in bulk (i.e., Lck and mutants compared in the same experiment) to guarantee the most homogeneous conditions and reduce variability. Therefore, the same representative images for JCaM1.6-Lck were shown in Figs. 1B, 3C and 5B as they come from the same in bulk experiment.
Flow cytometry (FCM)
Single-cell suspensions were transferred into a 96-well V-bottom plate, washed once with 100 μl FACS buffer (0.5 % BSA) in PBS). After spinning, supernatants were removed and cell pellets re-suspended in 50 μl staining solution containing fluorescence-conjugated primary Ab diluted in FACS buffer and incubated for 20 min at RT. Cells were then washed twice and either acquired immediately in a FACS Calibur flow cytometer (BD Biosciences) or BD LSR Fortessa X20 (BD Biosciences). Alternatively, cells were fixed with a pre-warmed fixation solution (BD Cytofix®, BD Biosciences) for 10 min at 37 ⁰C. Cells were then washed twice in 150 μl permeabilisation buffer (BD Perm/Wash I, BD Biosciences), re-suspended in 150 μl permeabilisation buffer and incubated at 4 ⁰C for 30 min. Primary antibodies, diluted in permeabilisation buffer, were added to the cells for 1 h, followed by three washes in permeabilisation buffer and the addition of the corresponding secondary antibodies (in permeabilisation buffer). After 3 washes, cells were analysed in a FACS Calibur flow cytometer or BD LSR Fortessa X20. Acquired data were analysed by FlowJo (FlowJo Software part of BD). Counts, percentages or median intensity fluorescence values (MFI) were extracted from FlowJo as excel files.
Imaging Flow Cytometry (ImageStream)
Samples were stained for Lck, CD45 and DAPI according to the general protocol for intracellular staining described above for FCM. After staining, cells were re-suspended at 1*107 cells per ml for loading onto the ImageStream instrument. Samples were run on a 2 camera, 12 channel ImageStream X MkII (Amnis Corporation) with the 60X Multimag objective, the extended depth of field (EDF) option providing a resolution of 0.3 μm per pixel and 16 μm depth of field. Bright field images were captured on channels 1 and 9 (automatic power setting). At least 10,000 images per sample were acquired using INSPIRE 200 software (Amnis Corporation) and then analysed using the IDEAS v 6.2 software (Amnis Corporation). A colour compensation matrix was generated for all the fluorescence channels using samples stained with single colour reagents or antibody conjugate coated compensation beads, run with the INSPIRE compensation settings, and analysed with the IDEAS compensation wizard. Images were gated for focus (using the Gradient RMS feature) on both bright field channels (1 and 9) followed by selecting for singlet cells (DNA intensity/aspect ratio). A mask depicting the plasma membrane (PM) was defined from the anti-CD45 staining, used as a membrane marker, and a ratio between the Median FI of Lck at the PM and the Median FI of Lck in the rest of cell was calculated.
Determination of A770041 IC50 for Lck, Csk, Src and ZAP-70
For Lck inhibition, we used A770041, which has a high affinity and specificity for Lck (
). The IC50 of A770041 for Lck, Csk, Src and ZAP-70 were determined by incubating serial dilution of A770041 with 1 μM of either one of recombinant Lck, Csk, Src and ZAP-70 in the presence of 1 μM ATP and 1 μM substrate, as previously reported (
). Data were obtained from MRC PPU Reagents and Services, School of Life Sciences (University of Dundee) and are shown in Table S1
LckT, LckA two-colour FCM
We opted for a two-colour FCM-based assay that concomitantly detected LckA and LckT on a per cell basis. An anti-Lck Ab (73A5) raised against Lck C-terminal tail, was found to be most adequate for this purpose. 73A5 showed an excellent FCM signal-to-noise ratio and epitope mapping by non-phosphorylated overlapping peptides revealed it to recognise Lck C-terminal end including Y505 (Fig. S2A). Treatment by A770041 or PV, both of which can change Y505 phosphorylation and conformers level, left 73A5 reactivity largely unaffected (Fig. S2B and S2C), indicating that 73A5 does not discriminate among Lck isoforms. 73A5-PE and anti-pY416 Abs were used at saturating concentrations with negligible effect on signal-noise and no hindrance to one another for Lck binding was observed (Fig. S2D). Moreover, plots of LckT and LckA amounts vs. forward scatter (FSC) indicated that LckT and LckA density/cell in Jurkat Cln20 was not linearly related to cell size (Fig. S2E), making unlikely that Lck concentration/cell was constant and indicating therefore that detection of LckA increase was indeed concentration-dependent on LckT. Together, these features allowed to unambiguously quantitate LckA as a function of LckT per cell basis and over a considerable LckT dynamic range (see Results).
LckT vs. LckA 2D plots
Cln20 or dox-induced JCaM1.6 expressing either wild type Lck, or Lck-chimeras or ΔSH4-Lck mutant were concomitantly stained for LckA and LckT as described above in “LckT, LckA two-colour FCM”. Double staining followed by FCM provided 2D plots (Figs. 2A and 2B) that described the dependence of LckA as a function of LckT. Indeed, Lck distribution in Cln20 was normal (Figs. 2B and S2A) and increase of LckT was minimally influenced by cell size (Fig. S2E). These features made our assay effectively reporting the increase Lck concentration per cell basis and therefore derive a genuine dependence of LckA on LckT. For our modelling, we used the data obtained in Cln20 cells as their average concentration of LckT can be considered close to physiological. This is justified by Cln20 expressing levels of Lck ≈ 5 times higher than T cells (
) but having an average diameter ≈ two-fold that of a T cell (Fig. 1B), hence a cell surface 4 times larger than T cells. This means that Cln20 and T cells have on average similar Lck concentration of LckT. Moreover, Cln20 and T cells have very similar PM/CP ratio for Lck (Fig. 1B) making their Lck concentration at the plasma membrane very similar. When comparing LckA generation by Lck and the Lck chimeras, we present in Fig.4A the full range of LckA expression upon dox-induction (without any evident sign of saturation). However, only the range of LckA generated within Cln20 range (blue box superimposed to each 2D FCM plot) was considered for the comparisons. This considerably reduced the burden of data collection and analysis without sacrificing to the validity of the data. Indeed, no Lck chimera showed major deviations in LckA dependency on LckT beyond the Cln20 range (Fig. 4A). The geometric median ± SD for LckA and LckT was calculated for each bin and background was subtracted (e.g., A770041-treated Cln20 or dox-untreated JCaM1.6). The resulting values were subjected to regression analysis to obtain the line of best fit (Fig. 2B, right panel). Non-linear regression and statistical analysis and were performed with Prism (GraphPad Software) or R software standard libraries.
Construction of chimeric or mutated proteins and cloning
LckSH4 provides firm attachment of Lck to the plasma membrane. LckSH4 is eleven amino acid-long and devoid of secondary structure (Fig. 3A), away from folded Lck SH domains. As such, LckSH4 is unlikely to have a critical influence on Lck allosteric regulation and catalytic activity. The cDNA of human Lck wild-type (Lck) was used to generate all Lck chimeras and the cytoplasm-resident mutant LckΔSH4. All Lck constructs were cloned in the expression vector pLVX-Tight-Puro (Clontech Laboratories, Inc.), between 5’ NotI and 3’ EcoRI restriction sites. The SrcSH4-Lck chimera was generated by PCR using an oligonucleotide juxtaposing human SrcSH4 to human Lck. Specifically, the oligonucleotide used comprised the nucleotide sequence encoding amino acids 1-11 of human SrcWT, followed by amino acids 11-18 of Lck (Table S2). LckΔSH4 was obtained by PCR using a 5’ primer corresponding to amino acids 11-19 of Lck. To facilitate the generation of the LAT-, CD4-, CD4C/S- and CD45-Lck chimeric proteins, an XbaI restriction site was introduced prior to triplet coding for Asp11 of Lck. Then, NotI-XbaI fragments comprising the nucleotide sequences coding for the selected anchors were ligated to Lck XbaI-EcoRI fragment, lacking the SH4 domain (coding for residues 11-509) (see Table S2). The chimeras LAT-Lck and CD45-Lck were generated with cDNA of human LAT and human CD45 of our laboratories. For the CD4-Lck chimera we used as a template a cDNA of murine CD4 graciously provided by Prof Simon Davis’ laboratory. The CD4C/S-Lck chimera was generated in our laboratory by site-directed mutagenesis of our CD4-Lck construct. All chimeric and mutant constructs were verified by DNA sequencing.
Production of lentiviral particles
Lentiviruses were generated using the packaging cell lines Lenti-X293T. The culture medium was exchanged with RPMI supplemented with 10 % FBS just prior to transfection. Lenti-X293T at 80 % confluence were transfected using PEIpro (Polyplus) according to the manufacturer’s instructions. The packaging plasmids pVSVG and pSPAX2 were mixed with the lentivirus expression vectors containing the gene of interest. PEIpro solution was added to the plasmids mix and immediately vortexed, left 15 min at RT and then added dropwise to the cells by gently swirling the plate. Supernatant containing lentiviral particles was collected after 48 h and filtered through a 0.45 μm sterile filter (Sartorius Stedim). Lentivirus supernatants were concentrated with PEG-itTM (SBI) concentration kit according to the manufacturer’s instruction. Briefly, lentiviral supernatants were mixed with Virus Precipitation Solution (SBI) to a final concentration of 1X Virus Precipitation Solution and incubated overnight at 4 ⁰C followed by a centrifugation at 1,500 x g for 30 min at 4 ⁰C. Pellets containing lentivirus particles were re-suspended in 1/100 of the volume of the original cell culture using cold RPMI. Aliquots were immediately frozen in cryogenic vials at - 80 ⁰C and stored until use. Aliquots of each lentivirus batch were routinely pre-tested by serial dilution titration. Frozen aliquots were thawed only once and used immediately with minimal loss of virus titre as determined by FCM.
Generation of Tet-On inducible cell lines
Stable, inducible cell lines were generated using the Lenti-X Tet-On-Advanced Inducible Expression System (Clontech Laboratories, Inc.) according to the manufacturer’s instructions. Briefly, JCaM1.6 were transduced with lentiviral particles (as described above) containing the PLVX-Tet-On-Advanced vector, which constitutively expresses the tetracycline-controlled trans-activator rtTA-Advanced. 48 h after transduction, the cells were subjected to selection by Geneticin (1 mg/ml) to generate a stable JCaM1.6 -TetON cell line. This parental cell line was then transduced with lentiviral particles of pLVX-Tight-Puro containing the Lck constructs and, 48 h after transduction, subjected to selection by Puromycin (10 μg/ml) and Geneticin (1 mg/ml) to generate the respective stable cell line. Expression of the Lck constructs was induced by 1 μg/ml doxycycline (dox, Sigma-Aldrich) added to the cell culture medium, routinely 14 - 18 h prior to each experiment. Potential phenotypic drift of cell cultures was reduced by conditionally expressing Lck or chimeras in JCaM1.6 by doxycycline induction for 14-16 h.
CellTrace violet labelling
To quantitatively evaluate the formation of LckA depending on LckT and according to different lipid anchor, we employed an FCM-based approach that allows to concomitantly detect LckA and LckT on a per-cell basis. To improve precision and accuracy, we performed double staining of LckA and LckT of two different JCaM1.6 expressing mutated or chimeric-Lck together with JCaM1.6-Lck (used as an internal reference). To this aim, two cell lines were labelled with different concentrations (1 and 0.25 μM) of CellTrace violet (Thermo Fisher) and JCaM1.6-Lck with carrier control (DMSO, Sigma) prior to dox-induction. Specifically, cells were washed once in PBS and adjusted to a final concentration of 106 cells/ml in pre-warmed PBS at 37 ⁰C. CellTrace violet or carrier control DMSO (Sigma) was added at the concentrations indicated above and cells were incubated at 37 ⁰C in the dark. After 20 min, samples were diluted 5-fold in complete medium and incubated for an additional 5 min at 37 ⁰C in the dark. After removal of excess of CellTrace violet, cells were re-suspended in complete medium, counted, mixed in 1:1:1 ratio and induced in the same well by ON-addition of 1 μg/ml dox. In this way, three JCaM1.6 cells were induced at the same time for expressing independently two chimeric-Lck constructs and Lck wild type, respectively, and then subjected to FCM analysis. This stratagem considerably reduced experimental variability and allowed Lck wild type as standard internal control.
Probabilistic model of LckA formation
To investigate LckA formation as a function of LckT, we generated a simple probabilistic model where Lck can assume three different states: the inactive conformation (LckI), the primed conformation (LckP) and the active conformation (LckA). Therefore, the three following reactions occurring at the plasma membrane were considered:
The following assumptions were made in the model:
In the initial state (1), the equilibrium reaction is largely shifted towards LckP conformation.
Two different probabilities (P) are assigned to reactions (2) and (3), while P for reaction (1) is close to 1.00.
The increase of total Lck (Lck T) is included in the model by the presence of an additional parameter.
The contribution of CD45 is not included in the model as it can be considered a hidden variable (see Results)
Starting from these assumptions, we studied the variation of LckA with respect to the amount of LckT. For each cycle, Lck can interact with any other Lck form and this interaction can either lead to: i) un unchanged condition - e.g., LckI interacting with any other Lck conformation or ii) formation of one LckA generated by LckP interacting with LckP. With increasing LckT, the amount of LckA increases and an additional reaction can take place: LckA reacting with LckP, leading to two molecules of LckA. The probabilities associated to these reactions: (2) and (3), PPA and PAA respectively, are optimised to fit experimental data and can vary in the simulation from 0.1 to 1.00 with step increments of 0.05. Our phenomenological approach attempted to describe the experimental data by a simple mode, based on trend of the line of best fit of the experimental data. Occurrence of reactions (2) and (3) leads to the generation of LckA. In this minimalistic phenomenological model P incorporates various factors that may influence positively or negatively LckA formation (e.g., Csk, CD45, Lck intrinsic enzymatic activities and their concentrations, which for CD45 and Lck depend also on their lateral behaviour). As inferred from our own data, Csk contribution to LckP LckA dynamic equilibrium established at the plasma membrane should be minimal (see in the Results section “Dynamic maintenance of steady LckA” and Fig. S1H). This is because in the steady state Csk does not seem to effectively offset CD45 action that converts to LckP most of LckI merging from the cytoplasm into the plasma membrane. Moreover, based on the data presented, CD45 constitutive activity limits LckA amount at the plasma membrane and, in so doing, generates LckP that fuels LckA formation. Hence, CD45 acts on both sides of the LckA formation - i.e., reactions (1), (2) and (3). As such, CD45 can be considered as a hidden variable contributing to P. Such an assumption is justified also a posteriori by the perfect fit of the probabilistic model to the experimental data without explicitly considering CD45 action in the model. For this reason, our phenomenological model is valid for quantifying the two concatenated reactions PA and AA and their relative weight independently of other factors that influence those reactions. The line of best fit and p-value were obtained by R software standard libraries.
Procedure used for the Ising model simulation
We simulated the ferromagnetic Ising model with coupling constant by the Kawasaki-Metropolis algorithm (
) on a square lattice with periodic boundary conditions. The temperature is set to , just above the critical one; is the Boltzmann constant. The concentration is exactly the critical one, i.e. both lipid phases, represented in black and white in Fig. S6C, have equal concentration. The IMP or protein anchor is schematised by a disc imposing a boundary condition as if it were filled with the black phase.
All the experimental data are contained within the article. There are no restrictions on any data or materials presented in this paper. Requests for unique resources and reagents generated in this study should be directed to and will be fulfilled by the lead contact.
This article contains supporting information.
Declaration of Interests
The authors declare no competing interests.
We thank Drs Anna K. Schulze and Thomas Hofer (DKFZ, Heidelberg) for help with initial FCM data analysis and Prof Simon Davis (Oxford University) for donation of a murine CD4 cDNA. We are particularly indebted with Dr Richard M. Parton for initial supervising of 3D-SIM sample preparation and imaging and with Drs Antreas Kalli; Gerhard Schutz; Kai Simons, Ilpo Vattulainen, Rajat Varma, Peter Tieleman; Omer Dushek, Michael Dustin and Andres Alcover for helpful discussions and suggestions. We thank Christine Ralf and Ana Maria Vallés for reading the manuscript.
Conceptualisation by K.N. and O.A. Project supervision by K.N. and O.A. Experiments were performed by N.P., K.N., D.C., G.M. and A-L.L. Supervision of SIM experiments, masks’ design and script for PM/CP by L.S. Supervision of ImageStream experiments, PM/CP script and data analysis by D.H. and S.P.C. Lck empirical model by E.M., A.G. and M.D. Suggestions for a model of critical phenomena in lipid bilayer phase separation and performing Ising model simulation by N.D. Data Interpretation and conceptual elaboration by N.P., H- T.H., N.D., K.N. and O.A. Manuscript writing: original drafts by O.A., edited by N.P., K.N., N.D. and O.A. The complete manuscript was read by all authors.
Funding and additional information
Wellcome Trust Grants GR076558MA and WT094296MA to O.A.; K. Karatheodori Program E609 (University of Patras) to K.N. Wellcome Trust Awards 091911 and 107457 to Micron Oxford; Ministero dell’Istruzione, Universita e Ricerca (R. Levi-Montalcini fellowship) to M.D. CNRS IRP CHOLESTIM to H.-T.H.
Author contributions. Conceptualisation by K.N. and O.A. Project supervision by K.N. and O.A. Experiments were performed by N.P., K.N., D.C., G.M. and A-L.L. Supervision of SIM experiments, masks’ design and script for PM/CP by L.S. Supervision of ImageStream experiments, PM/CP script and data analysis by D.H. and S.P.C. Lck empirical model by E.M., A.G. and M.D. Suggestions for a model of critical phenomena in lipid bilayer phase separation and performing Ising model simulation by N.D. Data Interpretation and conceptual elaboration by N.P., H- T.H., N.D., K.N. and O.A. Manuscript writing: original drafts by O.A., edited by N.P., K.N., N.D. and O.A. The complete manuscript was read by all authors.