Cold-induced coalescence of T-cell plasma membrane microdomains activates signalling pathways

The plasma membrane of mammalian cells contains several types of microdomains, of which lipid rafts, themselves a mixture of small domains with similar properties, have received a lot of attention lately (Edidin, 2003; Lai, 2003; Munro, 2003). It is suggested that lipid rafts form by the self-aggregation of cholesterol and sphingolipids (Simons and Ikonen, 1997). As a result of their lipid composition, they should be more ordered than the bulk membrane and are believed to exist in a liquid ordered (lo) state that resembles both the liquid disordered state, in that the lipids are fluid, and the gel state, in that the lipids are highly organised. Recently, evidence has been presented that favours membrane regions with different fluidity in live cells (Gaus et al., 2003; Kindzelskii et al., 2004).

One biochemical definition of lipid rafts is their insolubility in non-ionic detergents at 4°C – a procedure that generates detergent-resistant membranes (DRMs). DRMs prepared by Triton X-100 (TX-100) extraction (TX-DRMs) are enriched in cholesterol, glycosphingolipids, sphingomyelin and saturated glycerophospholipids (Fridriksson et al., 1999). These lipids can by themselves form a lo-like state at 37°C where acyl chains are tightly packed, highly ordered and extended, supporting the raft hypothesis (Schroeder et al., 1994). The use of different detergents when preparing DRMs makes results hard to compare and interpret. Out of a wide range of detergents tested in a systematic study, the only two that gave a DRM fraction enriched in the proposed lipid raft constituents, namely cholesterol and sphingomyelin, were CHAPS and TX-100 (Schuck et al., 2003). Most detergents were shown to result in the transferrin receptor, the archetypal non-lipid raft marker, appearing in the DRM fraction. Although TX-100 is more selective than other detergents, it is not clear that even TX-DRMs represent a physiological entity (Edidin, 2003; Magee and Parmryd, 2003).

Ligation of the T-cell antigen receptor (TCR) results in the rapid tyrosine phosphorylation of multiple intracellular proteins, mediated by the membrane-associated Src-family protein tyrosine kinases (PTKs) Lck and Fyn and the soluble PTKs ZAP-70 and Syk. The tyrosine phosphorylation triggers downstream signalling pathways including Ca²⁺ mobilisation, activation of the Ras/extracellular-regulated kinases (ERK) and hydrolysis of phosphoinoisitide polyphosphates (Cantrell, 1996). All these pathways can be activated by the aggregation of the ganglioside GM1 using cholera toxin B subunit (CT-B) and anti-cholera toxin (Janes et al., 1999; Parmryd et al., 2003) thus linking reorganisation of lipids to T-cell signalling. However, the role of lipid rafts in TCR signalling has been questioned (Glebov and Nichols, 2004; Pizzo et al., 2002). Formation of the immunological synapse occurs after the initial signalling events (Lee et al., 2002) and it has recently been demonstrated to play an important role in signal down-regulation (Lee et al., 2003).

In model membranes of the mixture of phospholipids:sphingolipids:cholesterol (1:1:1), derived from the lipid composition of TX-DRMs, micron-scale phase separation has been observed at temperatures below 25°C (Dietrich et al., 2001; Veatch and Keller, 2002), but not at higher temperatures. Using fluorescence resonance energy transfer (FRET), it has been shown that phase separation also occurs at physiological temperatures, but on the smaller scale of tens of nanometres (Silvius, 2003). It has long been known that human blood platelets are activated upon chilling, which limits their storage lifetime (Zucker and Borrelli, 1954). This activation has been correlated with membrane phase transitions (Gousset et al., 2002) and it was suggested that raft aggregation is triggered by cell activation.

In this study, the distribution of both lipid and protein lipid raft markers in the plasma membrane of Jurkat T cells at different temperatures was investigated as were early T-cell signalling events in Jurkat as well as primary T cells. The data presented support both the existence of microdomains of unique compositions in cell membranes and a role for them in TCR signalling. Furthermore, they suggest that TCR signalling, characterised by increased tyrosine phosphorylation, could be a consequence of the separation of kinases and phosphatases sustained by microdomain coalescence.

Fluorescein- and horseradish peroxidase-conjugated secondary antibodies were from Sigma and cholera toxin B subunit (CT-B)-rhodamine conjugate from List Biological Laboratories. CT-B-Alexa Fluor 594, anti-mouse-Alexa Fluor 594, anti-mouse-Alexa Fluor 488 and anti-rabbit-Alexa Fluor 488 conjugates were from Molecular Probes. Anti-CT-B was from Calbiochem-Novabiochem. Anti-Lck rabbit antiserum (2166) has been described previously (Kabouridis et al., 1997) and rabbit anti-transferrin receptor (TfR) was a kind gift from E. Lennox (MRC LMB, Cambridge, UK). Anti-CD3 monoclonal antibodies (OKT3 and UCHT1), anti-ZAP-70 rabbit antiserum (ZAP-4) and anti-pTyr monoclonal antibody (4G10) were generous gifts from S. Ley and 145-2C11 was a kind gift from B. Stockinger (NIMR, London, UK). Anti-LAT rabbit antiserum (M41) was from M. Turner (The Babraham Institute, Cambridge, UK). Anti-p44/42 MAP kinase and anti-phospho-p44/42 MAP kinase were from Cell Signaling Technology. Anti-Lck, clone 3A5, was from Santa Cruz Biotechnology. Rainbow molecular mass markers were from Amersham. Unless otherwise stated, chemicals were from Sigma.

E6.1 Jurkat T cells and derivative cell lines were cultured in RPMI 1640 medium supplemented with 5% FCS, 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin.

Src-family tyrosine kinases were inhibited by incubation for 45 minutes at 37°C with 10 μM PP1 (Calbiochem) from a stock solution in DMSO. Control cells were supplemented with the same amount of DMSO.

Cells were resuspended at 10⁷ cells/ml and split into 250 μl aliquots on ice. The cells were incubated at the indicated temperatures for 10 minutes with anti-CD3 (OKT3) at 1 μg/ml present for the last 5 minutes. Cells were pelleted by brief centrifugation followed by prompt addition of 125 μl ice-cold lysis buffer (50 mM Hepes pH 7.5, 150 mM NaCl, 1.5 mM MgCl₂, 1 mM EGTA, 1% Triton X-100, 20 mM NaF, 10 mM Na₄P₂O₇, 1 mM PMSF, 1 mM Na₃VO₄, 5 μg/ml each of chymostatin, leupeptin, antipain and pepstatin). Cell suspensions were vortexed and incubated on ice for 5 minutes. Cell debris was removed by centrifugation at 14,000 g for 5 minutes at 4°C.

Cells were suspended at 10⁷ cells/ml (5 ml) in RPMI containing 25 mM Hepes and incubated at either 10 or 37°C for 10 minutes, of which for the last 5 minutes the cells were stimulated with anti-CD3 (OKT3) at 1 μg/ml. 45 ml ice-cold PBS was added and the cells were pelleted at 0°C. The cells were resuspended in 1 ml MNE (25 mM Mes pH 6.5, 150 mM NaCl, 2 mM EDTA) containing 1% Triton X-100, protease inhibitors (5 μg/ml each of antipain, leupeptin, chymostatin and pepstatin and 1 mM PMSF), 5 mM NaF and 1 mM Na₃VO₄. The lysate was incubated on ice for 15 minutes after which DNA was sheared by passage through a 23G needle 10 times. The lysate was mixed with 1 ml 80% sucrose/MNE and overlaid with 2 ml 30% and 1 ml 5% sucrose/MNE. Centrifugation was carried out in a SW55 rotor at 200,000 g for at least 16 hours. The Triton soluble fraction was defined as the bottom 0.8 ml of the 40% sucrose layer. The DRM fraction was collected from the 5-30% sucrose interface, diluted with MNE and pelleted by centrifugation at 100,000 g for 1 hour. The pellet was rinsed with MNE and solubilised in RIPA and sample buffers.

Peripheral blood mononuclear cells were isolated from fresh blood from healthy donors on a Histopaque gradient. Cells were mixed with CD4-depletion beads (Miltenyi Biotec., Cologne, Germany) and negatively selected on a MACS column. The cells were diluted to 10⁶/ml, split into two aliquots and cultured in RPMI medium supplemented with 10% human serum, 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin ± 2 μg/ml phytohaemagglutinin for 3-6 days.

Jurkat T cells incubated at 10°C for 5 minutes were stimulated with 1 μg/ml OKT3 for 5 minutes at 10°C and lysed in 1% NP40-containing buffer. Precleared lysates were used for immunoprecipitation with either anti-ZAP-70, anti-LAT or anti-Lck rabbit antisera coupled to protein A-Sepharose beads with dimethylpimelidate. Precipitated proteins were recovered by glycine elution and analysed by SDS-PAGE and western blotting for tyrosine phosphorylation. Protein concentration of total lysates was estimated visually by Coomassie Brilliant Blue staining of filters containing aliquots of the lysates alongside BSA standards. Gels were loaded on an equal protein basis. Proteins were resolved on 11% acrylamide gels and wet-blotted to nitrocellulose membranes. Blots were developed using enhanced chemiluminescence. Films were scanned on a ScanMaker 4 (Microtek) and the intensity of different regions of the blots was quantified using Quantity One 4.2.1 software (Bio-Rad).

For fluorescence confocal microscopy experiments, cells were washed in PBS and attached to TESPA-coated coverslips by incubation on ice for 45 minutes (2.5 ×10⁵ cells/coverslip). CT-B-rhodamine or Alexa Fluor-594 conjugate (5 μg/ml) in 0.1% bovine serum albumin (BSA)/PBS was used to label GM1 by incubation on ice for 30 minutes. CT-B patching was induced as described previously (Janes et al., 1999). The cells were fixed with 4% paraformaldehyde (PFA) in PBS for 60 minutes at 20°C.

For immunofluorescence staining of endogenous proteins, fixation was performed in 4% PFA/PBS at 0°C and 20°C for 60 minutes and for 20 minutes at 37°C. The fixation was always preceded by a pre-incubation of at least 5 minutes at the desired temperature. Fixation at 20°C consistently resulted in more intact cells with discernible nuclei than did fixation at the other temperatures. Where indicated, cholera toxin B-Alexa Fluor 594 was added to the cells at the respective temperatures prior to fixation. Cells were permeabilised with 0.1% NP40/PBS on ice for 5 minutes after PFA fixation if intracellular staining was desired. Where indicated, extraction was performed on fixed cells with 1% TX-100 on ice for 5 minutes. Cells were then blocked with 2% BSA/PBS on ice for at least 15 minutes followed by incubation with primary antibodies (1-10 μg/ml in 2% BSA/PBS) at room temperature for 30 minutes. After washing in PBS, the cells were incubated with fluorescein- or Alexa-conjugated secondary antibodies in 2% BSA/PBS at room temperature for 15 minutes followed by washing in PBS.

Confocal microscopy was performed with a Leica TCS SP1 confocal microscope with a 100 × objective lens (NA 1.4) and a pinhole of 1.0 Airy disc. Sequential laser excitation at 488 and 568 nm and emission collection between 495-530 nm and 590-680 nm for fluorescein/EGFP/Alexa Fluor 488 and Alexa Fluor 594, respectively, were used to minimise bleedthrough between the two fluorophores. The zero offset was set with the laser turned off and the photomultiplier gain was adjusted to keep the intensity range, from background to the most intense, within the eight bit digital output range. Cells selected for imaging met the criteria of having no near neighbours and clearly discernible nuclei with phase contrast. Imaging was performed at the equatorial plane of the cells. Each experiment was repeated at least twice and more than 100 cells meeting the above stated criteria were visually examined from each population. The images were acquired blindly to minimise operator bias. Displayed images were prepared using Adobe Photoshop 7.0 software.

The Pearson (Manders, 1993) correlation coefficient was used to quantify the degree of overlap of two fluorophores in the plasma membrane. Background subtraction and plasma membrane demarcation was performed as described earlier (Parmryd et al., 2003).

Calculation of the fraction of the fluorescence at the plasma membrane was based on a binary mask defining the membrane (Parmryd et al., 2003) and then using an image processing `fill' operation to delineate the total area of the cell in the image. The fraction of fluorescence in the plasma membrane compared with the total fluorescence in the cell was calculated for equatorial optical sections. While the absolute fractions cannot be extrapolated from a single image to the whole cell, the use of similar equatorial slices permits comparisons between cell populations.

The plasma membrane was delineated manually with sequentially marked points that were then joined automatically. The accuracy of the demarcation was improved by using software that allowed for the movement, insertion or deletion of individual points in the sequential point list (Parmryd et al., 2003). Once delineated, the fluorescence intensities around the membrane were extracted to generate perimeter trace images. When fluorescent vesicles were in contact with the plasma membrane these portions of the membrane were excluded from further analysis. To permit comparisons of the pattern of distribution of fluorescence in the plasma membrane between populations, the standard deviations of the intensity were expressed as a percentage of the mean perimeter fluorescence. This removes the effect of PMT gain and gives equal weight to each dataset when calculating population statistics. A single frequency distribution histogram, representing a whole population, was made by summing the individual perimeter intensity distribution histograms after rescaling to give each an equal weight in the final histogram. This was changed into an intensity distribution that shows the fraction of the total intensity present at each fluorescence intensity and emphasizes the contribution of a small number of high intensity pixels.

When performing immunofluorescence with antibodies in fixed cells, there is always a risk that any domains observed are the result of antibody-triggered redistribution facilitated by inadequate fixation. To evaluate this risk a protocol was developed to assess how well the cells were fixed. Jurkat T cells were fixed at 0, 20 or 37°C followed by labelling for the ganglioside GM1 with cholera toxin B-subunit (CT-B), and then treating with anti-CT-B to induce cross-linking of GM1-enriched domains. Surprisingly, since lipids are not fixed with paraformaldehyde, the fixation protocol proved to be sufficient to prevent any reorganisation of GM1, visible at the resolution of the light microscope, upon addition of the patching reagents (Fig. 1A). This suggests that fixed proteins form obstacles for lipid diffusion. After fixation at 20 and 37°C, GM1 staining is prominent in regions of the plasma membrane rich in membrane activity. Cells where GM1-CT-B patching was performed prior to fixation are shown for comparison (Fig. 1B).

Lck is a Src-family tyrosine kinase that is modified by saturated acyl chains in its N terminus, which can target it to specific signalling domains in the plasma membrane. At 0°C, Lck displayed punctate staining with many regions either devoid of or clearly enriched in protein (Fig. 2A, first row). In cells fixed at 20°C, the plasma membrane staining intensity variations were less pronounced and they were even further reduced at 37°C. CD3ϵ forms part of the T-cell receptor (TCR) complex and is a transmembrane polypeptide with no known targeting motifs. In cells fixed at 0°C, there were both small and large regions of the plasma membrane devoid of CD3 staining (Fig. 2A, second row). At 20 or 37°C, there was a more continuous, fine punctate CD3-staining along the plasma membrane. The transferrin receptor (TfR), a type II membrane protein, is the archetypal marker for non-ionic detergent soluble (non-raft) membranes. For the TfR, the intensity around the plasma membrane at all temperatures was more homogenous than for Lck and CD3, but the TfR also showed occasional more intense regions at 0 and 20°C (Fig. 2A, third row). Perinuclear staining, most likely of the endocytic recycling compartment, could be seen in some cell sections.

To better visualise the variations in intensity of the plasma membrane staining, perimeter traces are displayed in Fig. 2B. From cells fixed at 0°C, both the CD3 and Lck traces have many clear peaks as well as troughs, confirming that these proteins cluster at reduced temperatures. To assess the level of clustering in a cell population, perimeter traces from ten or more cells were combined in single intensity distribution histograms (Fig. 2C). For all three proteins, the distribution was more skewed towards higher intensities in the populations fixed at 0°C. The populations fixed at 37°C had almost normal distributions and those fixed at 20°C showed an intermediate distribution. The level of clustering was quantified using the spread of intensities about the mean intensity. Table 1 shows the arithmetic mean standard deviation of a population along with the standard error of the population standard deviations. The higher the standard deviation, the less uniform the staining. Lck at 0°C was the most clustered, followed by CD3 at 0°C. The TfR was no more clustered at 0°C than Lck or CD3 at 37°C, but all three proteins displayed the same trend in that the level of clustering increased as fixation temperature decreased.

In previous studies we have demonstrated that the TfR does not coaggregate with patched GM1-enriched domains whereas Lck does (Janes et al., 1999; Parmryd et al., 2003). Since Lck has a punctate distribution at 0°C this represents an opportunity to assess colocalisation without inducing domain aggregation by cross-linking proteins or lipids. At 0°C, the intensity of TfR staining was clearly reduced in regions where Lck staining was intense (Fig. 3, first row) suggesting that they are present in different membrane domains. This was further emphasised by the minimal Pearson correlation coefficient of 0.010 (s.e.m. ±0.025, n=10) of the two fluorophores, showing that there is no relationship between the distributions of the two proteins. Lck and CD3, however, had a Pearson correlation coefficient in the plasma membrane of 0.44 (s.e.m. ±0.067, n=10) at 0°C, indicative of considerable colocalisation (Fig. 3, second row).

Next, the distribution of the ganglioside GM1, widely used as a lipid raft marker, was investigated using CT-B as a probe. CT-B staining of the plasma membrane was even and continuous without any apparent temperature-related differences when cells were fixed before staining (Fig. 4, top row and Table 2). However, when CT-B was added to the cells at the different temperatures prior to fixation the staining was more clustered at all temperatures (Table 2). This suggests that protein-lipid and/or lipid-lipid interactions are lost upon fixation and/or that some lipid diffusion occurs in the fixed cells. That GM1 was more clustered in cells that were imaged immediately after preparation than those imaged after storage in the fridge overnight (not shown) also supports this interpretation.

An operational definition of lipid rafts is that they are more resistant to extraction with non-ionic detergents at low temperature than are other regions of the plasma membrane (Brown and Rose, 1992). Therefore, cells were subjected to extraction with 1% TX-100 on ice, after fixation at various temperatures, to investigate if the distribution of the GM1 was altered by this treatment. GM1 cannot be directly fixed by PFA and could therefore potentially be susceptible to TX-100 extraction. In cells fixed at 0°C, GM1 was depleted from a few regions of the plasma membrane after TX-100 extraction. In cells fixed at 20 or 37°C, TX-100 extraction did not result in plasma membrane regions depleted of GM1 but it caused a less even plasma membrane distribution (Fig. 4, bottom row). In fixed cells, the effect of 1% TX-100 extraction on GM1 did not differ from that of 0.1% NP-40 permeabilisation (Fig. 4, bottom row and Fig. 6).

Signalling can be stimulated by GM1 aggregation with CT-B cross-linking (Janes et al., 1999; Parmryd et al., 2003) and various lipid raft-partitioning molecules displayed a patchy distribution at 0-20°C. This raises the possibility that T-cell signalling might be stimulated at low physiological temperatures, although counterintuitive considering the effect that reducing the temperature generally has on the kinetics of biological processes. To test this, Jurkat cells were incubated at various temperatures and analysed for phosphotyrosine (pTyr) content since this is increased early in the T-cell signalling response. At 10°C, there was an increase in the level of pTyr, particularly in polypeptides with molecular masses of 35-38, 56-58, 71 and about 105 kDa (Fig. 5A). These masses correspond to LAT, Lck and Fyn Src-family PTKs, ZAP-70 (Fig. 5C) and probably to Vav, all known to participate in T-cell signalling. Sequential 5-minute incubations at 10°C and then at 37°C or vice versa resulted in pTyr patterns identical to those at 37°C and 10°C, respectively (not shown), showing that the low temperature-induced pTyr-increase is reversible. Furthermore, the cold-induced increase in pTyr was observed in cells that had been incubated for up to 6 hours at the respective temperatures before the addition of anti-CD3 (not shown). The ERK MAP kinase, which is further downstream in the T-cell signalling cascade, was also activated at 10-20°C (Fig. 5B).

When the T cell stimulatory anti-CD3 antibody OKT3 was added, there was a 2.3-fold increase in the total level of pTyr in cells incubated at 10°C (Fig. 5A; Table 3) and the pTyr level of LAT was increased 6-fold. The pattern of the polypeptides was not changed, but a band at 23 kDa, probably corresponding to TCRζ, now became clearly visible. Very little difference could be seen after stimulation of cells incubated at 37°C except for some increase in the pTyr level of LAT. In some experiments, the temperature range in which activation was seen varied between 0°C to 20°C, but the highest level of pTyr was consistently observed at 10°C; the reason for this shift is unclear. In contrast to the pTyr increase, ERK activation at low temperatures was not substantially increased by OKT3, whereas at 30-37°C it was (Fig. 5B).

At 20 and 37°C, the majority of pTyr staining was intracellular but at 0°C staining of the plasma membrane was clearly visible (Fig. 6). The Pearson correlation coefficient for pTyr and GM1 at 0°C was 0.488, which means that they are convincingly colocalised. In cells fixed at 0°C, 3.5 times more of the pTyr staining was in the plasma membrane than in cells fixed at 37°C (Table 4), which further supports the idea that the pTyr bands that increased in intensity at 0°C (Fig. 5A) are plasma membrane proteins involved in T-cell signalling. The intracellular pTyr staining that appears to be more pronounced at 20 and 37°C, is likely to originate from both soluble and cytoskeletal-associated proteins not involved in T-cell signalling. TX-DRMs of cells preincubated at 10 and 37°C were prepared; those from cells stimulated with OKT3 at 10°C had slightly higher levels of pTyr than those stimulated at 37°C (Fig. 7A), consistent with increased preorganisation of the signalling complexes at the lower temperature. However, since cells are cooled during the preparation of TX-DRMs, cells stimulated at 37°C are likely to show similar features to those at lower temperatures. Interestingly, the 23 kDa band was detectable in TX-DRMs from cells stimulated with OKT3 as well as unstimulated cells. Only a subset of tyrosine phosphorylated proteins from cells incubated at 37°C were found in the TX-DRMs (Fig. 7B). The majority were found in the cytoskeletal pellet or the Triton-soluble fractions. This may explain the substantial intracellular pTyr staining observed at this temperature (Fig. 4).

The Jurkat-derived cell line J.RT3-T3.5 does not express CD3 on its surface and does not produce the TCR α/β heterodimer (Weiss and Stobo, 1984). Tyrosine phosphorylation was increased in J.RT3-T3.5 cells when they were incubated at 0-10°C (Fig. 8). However, the increased phosphotyrosine was seen only in bands that were the size of Src-family kinases and not in other bands that become tyrosine phosphorylated during normal TCR-induced activation, such as LAT, ZAP-70 and Vav. This suggests that the Src-family kinases are separated from phosphatases in J.RT3-T3.5 cells also and that tyrosine phosphorylation dependent on CD3/TCRζ ITAM phosphorylation, but not that independent of it, is defective in this cell line. Unlike in the parental Jurkat cells, this increase could not be further enhanced by the addition of anti-CD3, confirming that OKT3 is mediating its effect through the TCR (not shown). In the Lck-deficient Jurkat-derived cell line JCaM1.6 (Straus and Weiss, 1992) hardly any pTyr was observed (Fig. 8). Moreover, pTyr was not induced by either temperature decreases (Fig. 8) or anti-CD3 stimulation (not shown). As expected, the Src-family PTK inhibitor PP1 abolished both the cold-induced activation of ERK and the increase in pTyr in Jurkat cells (not shown). This further supports the need for Lck to obtain the cold-induced activation.

Jurkat T cells respond more readily to various stimuli than do primary T cells and they are likely to have a more saturated lipid mix in their membranes because of the lipid composition of tissue culture media. It was therefore important to investigate the effect of temperature in primary T cells. Although primary cells never produce blots with the same clarity as established tissue culture cell lines, it is evident that the addition of anti-CD3 to CD4+ primary human T cells gives the strongest pTyr response at 10-20°C (Fig. 9A). The most prominent band of ∼36 kDa is likely to be LAT. The ERK activation in primary T cells also mimics that in Jurkat cells with the highest level of activation observed at 20°C (Fig. 9B). This result supports the conclusion that signalling complexes are preorganized at low temperatures. A similar result was obtained for primary mouse T cells (not shown).

Lipid rafts are predicted to be more ordered than the bulk membrane and to exist in a liquid ordered (lo) state. For a pure lipid, the temperature at which a phase transition from the gel state to the liquid disordered state occurs, the melting temperature (Tm), is clear-cut. For mammalian cell membranes, Tm usually occurs between 10-20°C (Tablin et al., 1996) and the transition is not sharp because of differences in Tm of all the hundreds of lipid species present and high concentrations of cholesterol that inhibits phase transitions. It is possible that the portion of the membrane that is occupied by lipid rafts gradually increases as the temperature is decreased because more and more lipids could partition into a lo-phase at temperatures below their Tms. This increase would stop when the temperature drops below the Tm of all participating lipids and the membrane reaches the gel state where not much diffusion takes place. In macrophages, the portion of the membrane occupied by the more ordered of the two phases was higher at 22°C than at 37°C, supporting this theory (Gaus et al., 2003). Also in plasma membrane vesicles from mast cells, the more ordered component occupied a larger area when the temperature was decreased (Ge et al., 2003). It is equally possible that temperature decrease promotes phase separation in cell membranes similar to that observed in model membranes (Dietrich et al., 2001). Either of these mechanisms would represent a reinforcement of the plasma membrane organisation at 37°C and not the artifactual formation of domains with no physiological relevance, as was proposed recently (Kusumi et al., 2004). In the present study it was found that certain T-cell signalling molecules cluster at low temperatures and, remarkably when taking reaction kinetics into account, this correlated with increased T-cell activation.

The commitment of a T cell to signalling is controlled by the balance between PTKs and protein tyrosine phosphatases (PTPs) (Mustelin and Tasken, 2003). It was therefore a concern that the cold-induced activation was the result of different kinetics of PTKs and PTPs, leaving the former relatively more active and the latter relatively less active at lower temperatures. We find this unlikely for several reasons. Firstly, there is not just one PTK and one PTP involved in the signalling cascade and both groups have representatives from different protein families with different characteristics. Secondly, if PTP inhibitors are not included when cell lysates are prepared on ice, pTyr is rapidly lost from proteins (A.I.M., unpublished observation), which means that PTPs are active at this temperature. Thirdly, the cold-induced pTyr band pattern does not look like the increase obtained in the presence of the general PTP inhibitor pervanadate (Secrist et al., 1993). That cold-induced pTyr was more pronounced in Jurkat cells than in primary T cells may reflect the former lacking at least one protein tyrosine phosphatase involved in signal regulation (Tangye et al., 1998).

To what extent TX-DRMs represent true membrane organisation is a matter of debate (Edidin, 2003; Shogomori and Brown, 2003). Frequently, DRMs are isolated at the 5/30% sucrose interface to which any lipid-rich membrane material would float even if the cells from which they are derived have never been subjected to any detergents. TX-DRMs from Jurkat cells are a mixture of vesicles ranging in diameter between 50 nm and several μm (Magee and Parmryd, 2003). TX-100 extraction of cells is routinely performed on ice because extraction at 37°C gives TX-DRMs devoid of proteins (Schroeder et al., 1994). Since membrane rearrangement occurs quickly after temperature changes (Laggner and Kriechbaum, 1991), TX-DRMs are more likely to represent cells at 0°C and not at 37°C, if there is a difference in the membrane arrangement at the two temperatures. Considering the small size estimated for individual lipid rafts on the cell surface of resting cells at 37°C, it is clear that TX-DRMs are substantially aggregated structures. Our study shows that Lck, which partitions to DRMs, is uniformly distributed in the plasma membrane at 37°C but is clustered at 0°C, when viewed at the level of resolution of the light microscope (200-300 nm). This suggests that plasma membrane domains of unique compositions are stabilised at low temperatures and TX-DRMs are not solely the result of detergent-induced domain formation that can occur when lipid vesicles are mixed with low amounts of TX-100 over long periods (Heerklotz, 2002). However, it is noteworthy that the phospholipids:sphingolipids:cholesterol (1:1:1) mixture is what remains insoluble when model membranes, with a wide range of lipid compositions, are extracted with TX-100 at 4°C (De Almeida et al., 2003; Sot et al., 2002). That the TX-DRM mixture displays a lo-phase behaviour at 37°C might therefore be a coincidence.

Not much GM1 could be extracted by TX-100 from fixed cells, as judged by CT-B staining. Fixation of the cells should reduce the blebbing, fusion and scrambling that occurs when model membranes are subjected to cold TX-100 extraction (Sot et al., 2002). Hence, the GM1 remaining after extraction is not likely to be a consequence of detergent-induced domain formation but rather represents a true detergent-resistant membrane organisation. Aggregation of GM1-enriched domains triggers a T-cell signalling response, but it is weaker than that triggered by direct stimulation of the TCR (Janes et al., 1999). This suggests that GM1-enriched microdomains containing different subsets of proteins exist, which causes a dilution of the signalling molecules upon aggregation. The fact that GM1 has a uniform distribution and covers a larger surface area in cells fixed at 0°C than do either Lck, CD3 or tyrosine-phosphorylated proteins argues for this explanation. If fixation hinders TX-100 extraction of GM1 from the cell membranes analogously to the way fixation prevents patching of GM1, the remaining GM1 could overestimate lipid rafts and, additionally, the possibility that GM1 is not the ideal lipid raft marker cannot be excluded. The basis for considering GM1 a good lipid raft marker is its partitioning to TX-DRMs but it has recently been shown that TX treatment of membranes at 4°C causes the irreversible redistribution of TX-100-soluble GM1 to membranes where GM1 is TX-100-insoluble (Heffer-Lauc et al., 2005). Therefore, the lack of GM1 in TX-100-soluble fractions after TX-100-extraction does not necessarily mean that all GM1 is TX-100 insoluble. In two recent studies on mast cells, it was shown that aggregation of GM1 does not occur as a consequence of aggregation of the FcϵRI at 37°C (Wilson et al., 2004; Wu et al., 2004). However, in both these studies the cells or membranes were subjected to fixation prior to analysis and it is conceivable that interactions between GM1 and specific lipids and/or proteins are lost upon fixation. That GM1 in the present study displayed a moderately more patchy distribution if cholera toxin B was added before rather than after fixation, or if cells were imaged immediately after fixation instead of being stored, support this interpretation.

Several groups have found phosphorylated TCR in TX-DRMs from activated T cells (Montixi et al., 1998; Xavier et al., 1998), while others have failed to do so (Brdicka et al., 1998; Zhang et al., 1998). Nevertheless, a consensus exists that ligation of the TCR causes its translocation to lipid rafts. Our results imply that this consensus should be reconsidered. Firstly, the addition of anti-CD3 to cells kept at 10°C resulted in a large increase of pTyr and at this temperature the membrane should be in a gel state, thus precluding translocation. Therefore, the TCR must already be present in the same membrane domain as are the membrane-located PTKs Lck and Fyn. The lack of a further response of the cold-induced ERK activation to anti-CD3 is likely to result from reduced kinetics of a kinase in the signalling cascade leading to ERK activation and/or the failure to translocate proteins to the plasma membrane at low temperatures. Secondly, at low temperatures Lck and CD3 have patchy plasma membrane distributions that demonstrate colocalisation, suggesting that they partition to similar domains, whereas the TfR has a more uniform distribution that is not correlated with that of Lck.

Intuitively, one would expect non-raft domains to cluster as a result of exclusion from the clustered raft domains if the two domains occupy similar portions of the membrane. TfR is not seemingly patched to the same extent as are Lck and CD3, which is likely to be an effect of non-raft domains occupying a much larger surface area of the plasma membrane than the Lck- and CD3-containing lipid rafts. Assuming that Lck has an affinity for domains that occupy 25% of the plasma membrane and that the TfR has a preference for domains that occupy the remaining 75%, and that lowering the temperature segregates these domain types from one another, then at 37°C both proteins would appear to have an even distribution. At 0°C they will appear to have a patchy distribution but Lck will show a 4-fold enrichment in its raft domains compared with the remaining membrane, whereas the TfR would only show 1.3-fold enrichment in the non-raft domains.

We favour a model in which rafts in resting cells at 37°C are small, very dynamic and coupled between the two leaflets. Upon coalescence, triggered for instance by receptor ligation/dimerisation or a temperature drop, they become larger and more stable, quite likely as the combined result of stabilisation of a more ordered phase, anchorage to intracellular structures and/or confinement by the assembly of large intracellular complexes (Fig. 10). The confinement is caused by the membrane skeleton that creates diffusion barriers (Fujiwara et al., 2002; Murase et al., 2004). Our results suggest that T cells rapidly establish new pTyr equilibrium when the temperature is altered. Since T cells are more receptive to activation by anti-CD3 at temperatures lower than 37°C, it is tempting to speculate that this is a mechanism to help the immune response at areas prone to infection, such as skin lesions, where the lower than 37°C temperature might usefully boost signalling. In platelets, several signalling proteins have been found in TX-DRMs (Gousset et al., 2004) and membrane raft coalescence has been suggested to be a consequence of cold-induced platelet activation (Gousset et al., 2002). If, instead, activation is a consequence of lipid raft coalescence, as is the case for T cells (Janes et al., 1999), this might represent a mechanism facilitating blood clotting at surface-exposed sites.

The existence and importance of membrane microdomains and their role in T-cell signalling is not universally accepted. For instance, it was concluded that lipid rafts are not involved in T-cell signalling because there was not an increased FRET signal between GM1/CT-B and a GPI-anchored protein at the region of contact between Jurkat T cells and beads coated with stimulatory antibody (Glebov and Nichols, 2004). In order for the FRET signal to increase upon domain coalescence, the density of at least one of the molecules would have to be increased within the aggregated domain. That this should occur upon domain coalescence is not supported by the literature and it is still far from clear how FRET signals can be extrapolated to assess membrane organisation (Kenworthy, 2002). The present study provides evidence both for T-cell membrane heterogeneity and the importance of domain coalescence for overcoming the activation threshold for T-cell signalling. It also stresses that care must be taken when designing experiments to avoid mistaking the cold-induced activation for other stimulation.

We are grateful to colleagues listed under Materials for generous supply of antibodies, as well as all blood donors. We thank the staff at the NIMR imaging facilities for assistance with confocal imaging. We are indebted to P. Kinnunen for valuable advice and discussion. I.P. received a fellowship from the Swedish Cancer Fund (4350-B01-02SAA). The project was supported by grants to A.I.M. from the UK MRC (Programme Grant G0100471), and to I.P. from Jeanssons Stiftelser and Carl Tryggers Stiftelse (CTS 03: 254).