Phosphatidylserine exposure in B lymphocytes: a role for lipid packing

The anionic phospholipid phosphatidylserine (PS) is largely confined to the inner leaflet of the plasma membrane of healthy cells, and its flopping to the outer leaflet and exposure to the extracellular face of the membrane is an early event in apoptosis, the safe removal of cellular debris, and initiation of the clotting cascade.¹  The mechanisms by which PS distribution is altered are poorly understood but are generally assumed to involve proteins facilitating either inward or outward PS transport (“flippases” or “floppases,” respectively), and/or scramblases mediating bidirectional, headgroup-nonspecific phospholipid transport.^2,3  The rate of PS translocation has, for example, been found to be sensitive to the altered expression of the ATP-binding cassette transporter ABCA1, suggesting that this protein might either directly transport PS or regulate another transporter.⁴  PS exposure in apoptosis is also accompanied by a decrease in the normal packing of plasma membrane phospholipids, which can be detected by increased binding of the lipophilic dye MC540.⁵  While it is not known whether the 2 phenomena are mechanistically related, exposure of PS (detected by binding of annexin V [AV]) and decreased lipid packing have been reported to identify equivalent populations of early apoptotic cells.^6,7 

Loss of PS asymmetry can also occur in the absence of apoptosis, for example, in T cells stimulated via the ATP receptor P2X₇,⁸  and in a population of nonapoptotic CD4⁺CD45RB^lo activated/memory T cells.⁹  PS has also been reported to be constitutively exposed on most or all viable murine B cells,^10,11  with a level of surface PS of 10 to 100 times that of T cells that does not increase upon stimulation, indicating complete loss of lipid asymmetry.^10,11  In contrast to these reports, we show that PS distribution on murine B lymphocytes, as for most cells, is asymmetric, although PS is rapidly exposed ex vivo in the absence of exogenous signal. PS exposure was accelerated by stimulation with a calcium ionophore and was preceded by and dependent on a decrease in plasma membrane lipid packing. This change in lipid packing could not be dissociated from cell shrinkage and membrane deformity, suggesting that loss of PS asymmetry may occur at sites of membrane-bending at which lipid packing is decreased in one leaflet.

Friend virus B-type (FVB/n) and nonobese diabetic (NOD) mice were bred at the Biological Services Unit at Imperial College, Hammersmith Campus, London, United Kingdom. Mice lacking CD45 have been described elsewhere¹²  and were compared with age-matched nontransgenic mice from the same breeding program at The Babraham Institute, Babraham, Cambridge, United Kingdom. Mice lacking ABCA1 have been described elsewhere⁴  and were maintained on a DBA/1J background in a pathogenfree facility at Charles River Laboratories in Lyon, France. Where studied, ABCA1-deficient mice were compared with wild-type mice from the same breeding colony. All other strains were from Harlan-Olac (Bicester, United Kingdom). All home office and local ethical guidelines for the care of laboratory animals were followed.

Mesenteric lymph node cells were prepared from adult mice. Cell suspensions in phenol red-free Dulbecco modified Eagle medium (DMEM; Sigma, Poole, United Kingdom) were stained with CD19^PE, CD19^PERCP, CD19^APC, CD4^APC, or CD4^CYCHROME (Becton Dickinson, San Jose, CA) antibodies, as indicated. The use of differently labeled antibodies enabled B cells from 2 mouse strains to be detected differentially, and hence studied simultaneously in the same tube. Cells were washed and resuspended in DMEM and, where indicated, equilibrated with AV^FITC or AV^CY5 (AV; Becton Dickinson), or with 0.05 μM merocyanine 540 (MC540; Sigma) for 3 minutes. Where cells were stimulated, baseline fluorescence was established for 30 seconds to 1 minute prior to addition of 2 μMto5 μM calcium ionophore (calcimycin, A23 187; Sigma). Cells were monitored for PS exposure continuously in real time by flow cytometry on a FACScalibur machine and analyzed using CellQuest (Becton Dickinson) or FlowJo (Treestar.com, Ashland, OR) software. Forward light scatter (FSC) was used as a measure of the volume of spherical cells,¹³  its sensitivity being greatest when light is collected over an angle of less than 10 degrees¹⁴  as in the FACScalibur. Distortion of the plasma membrane was measured by an increase in side light scatter (SSC), which has previously been shown to be concomitant with cell shrinkage (decrease in FSC) early in apoptosis.^14-16 

DIDS (Sigma) was dissolved in water. Tamoxifen (Sigma) was dissolved in ethanol. Calcium ionophore, glybenclamide (Sigma), and the Ca²⁺ indicator Fluo-4AM (used at 0.25 μM; Molecular Probes, Leiden, The Netherlands) were dissolved in DMSO. All results are representative of at least 3 independent experiments.

Because calcium ionophore (A23 187) is highly lipophilic, the concentration of ionophore required to induce PS flop (or MC540 uptake) is, in our experience, more dependent on the amount of cell membrane in the tube (and this is likely to include red cell contamination/debris/dead cells gated out during analysis) than on volume of the reaction mix. For this reason, responses of cells from different sources were compared in a single tube (to ensure identical ionophore to membrane concentrations in a given experiment). Changing ionophore and cell concentrations in a reaction mix have equivalent (but opposite) effects on responses.

Murine mesenteric lymph node cells, freshly dissected, were adhered to glass cover slips coated with poly-l-lysine (Sigma). Cells were exposed for 1 minute to 10 μM calcimycin in the presence or absence of 1 μM MC540, successively fixed in a solution of 4% formaldehyde, 4% sucrose in phosphate-buffered saline (PBS), and, where indicated, stained with AV-Alexa Fluor 568 (Molecular Probes). All cells were also stained with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI; 15 mg/mL; 1:10 000 dilution; Molecular Probes, Eugene, OR) for nuclear DNA. Cells were imaged with a Leica SP confocal microscope equipped with a 100×/1.4 PlanApoChromat oil-immersion objective lens (Leica, Wetzlar, Germany). Merocyanine 540 was excited with the 488-nm line of an argon laser and AV-Alexa Fluor 568 with the 561-nm line of a solid state laser; the emitted fluorescence collected through a triple-dichroic mirror (488/568/663) with a 498-nm to 700-nm bandpass filter and 575-nm to 700-nm bandpass filter, respectively. DAPI staining of nuclear DNA was excited with the 351-nm line of an ultraviolet (UV) laser, and emission fluorescence collected with a 396-nm to 508-nm bandpass filter. Stacks of confocal sections separated by 0.3-μm increments were taken and images analyzed with Metamorph 5.0v1 software (Universal Imaging, Downington, PA). Figures were prepared with Adobe Photoshop 6.0 software (Adobe Systems, San Jose, CA).

It has been reported that PS is exposed at high levels on unstimulated murine B cells and that further exposure does not occur on stimulation.^10,11  We assessed levels of cell-surface PS on rapidly isolated lymphocytes from C57BL/10 mice by real-time flow cytometry in the continuous presence of AV^FITC. After basal binding of AV^FITC was established, cells were treated with calcium ionophore (calcimycin, A23 187) to stimulate nonspecific uptake of Ca²⁺ and consequent PS exposure (Figure 1). Unstimulated B and T cells studied in the same tube exhibited comparably low-level binding of AV, indicating restriction of PS to the inner leaflet of the plasma membrane in both cell types. Stimulation with calcium ionophore resulted in marked PS exposure in both lymphocyte populations. Thus, as for most cell types and in contrast to previous reports,^10,11  unstimulated B lymphocytes (like T lymphocytes) possess an asymmetric lipid bilayer with PS largely restricted to the inner leaflet. Stimulation resulted in PS “flopping” to the outer leaflet: the time at which PS exposure was detectable and the peak AV-binding per cell depended on the concentration of calcimycin added (Figure 1B).

To assess the possibility that mouse strain-dependent differences in B-cell distribution of PS account for differences between our data and those published previously, we studied lymphocytes from a variety of normal and autoimmune strains. To ensure direct comparison of responses, B cells from strains being studied were stained with differently labeled anti-CD19 antibodies (to allow subsequent discrimination during analysis) and mixed before stimulation. Responses were compared by real-time flow cytometry in a single tube. PS was restricted to the inner leaflet of B cells from all mice, although significant strain-dependent differences in sensitivity to stimulation were apparent. For example, responses of B cells from the autoimmune-prone New Zealand white (NZW) and NOD strains of mice were comparable (not shown), but responded more rapidly than those from C57BL/10, DBA/2 (Figure 2), and 6 other nonautoimmune prone strains (129/Ola, BALB/c, C3H, CBA/Ca, FVB/n, NIH; not shown). That responses were more pronounced in cells from autoimmune-prone mouse strains is consistent with our recent observation that calcium ionophore-stimulated lymphocytes from patients with systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA) externalize PS more readily than do control cells.¹⁷ 

We suggest that the apparent discrepancy with previous reports arises from the fact that, in contrast to T cells, PS in B cells becomes exposed to the surface ex vivo within 2 hours even in the absence of exogenous stimulation (Figure 1C). Moreover, this exposure of PS on B cells was associated with their relatively rapid death ex vivo, as evidenced by a steadily falling proportion of CD19⁺ cells within the live gate population (Figure 1C). The disproportionate death of CD19⁺ B cells ex vivo is routinely observed in our hands, although the rate appears to be affected by factors such as temperature and serum concentration (data not shown). In previous studies,^10,11  B lymphocytes appear to have been maintained for a relatively long period ex vivo as this population was enriched over magnetic bead columns prior to analysis. In our experiments, by contrast, B cells were not physically isolated, but instead gated electronically on a flow cytometer on the basis of fluorescent antibody binding.

The rate of calcium ionophore-stimulated PS exposure has previously been reported to be decreased in erythrocytes and fibroblasts lacking the ABC transporter ABCA1, and increased in HeLa cells overexpressing ABCA1 following transfection,⁴  leading to the suggestion that ABCA1 might transport PS to the cell surface or regulate a heterologous PS transporter. However, we showed that the rates of PS exposure by ionophore-stimulated B lymphocytes from parental and from ABCA1^-/- mice could not be distinguished (Figure 3), indicating that ABCA1 is not involved in calcium ionophore-stimulated PS transport in murine B cells.

MC540 is a fluorescent, lipophilic dye that binds preferentially to the outer leaflet of plasma membrane bilayers with relatively loosely packed lipids,^5,18-20  including those of apoptotic cells.^6,7  We compared the kinetic relationship between MC540-binding (to measure lipid packing) and AV-binding (to measure PS exposure). Following calcium ionophore-stimulation in the presence of MC540, dye uptake exhibited a rapid increase (Figure 4A). Increased MC540 uptake preceded loss of PS asymmetry and, indeed, occurred at concentrations of ionophore below that required to stimulate detectable PS exposure (Figure 4A). Hitherto, the only event (other than calcium uptake) shown to precede PS translocation following calcimycin-stimulation is cell shrinkage due to K⁺ efflux via K_Ca3.1 (formerly IKCa1).²¹  Cell shrinkage and consequent buckling of the cell membrane can be measured in flow cytometry by a decrease in forward light scatter (FSC) and concomitant increase in side light scatter (SSC).^14-16  The rates of MC540 uptake (Figure 4B-C and Figure S1, which is available on the Blood website; see the Supplemental Figures link at the top of the online article) and cell shrinkage (Figure S1), as for PS externalization (AV-binding), were strain dependent with B cells from NOD and NZW mice exhibiting a relatively rapid response, suggesting that strain-dependent variation in rates of PS translocation reflects upstream differences in rates of cell shrinkage and decrease in lipid packing. Strain-dependent variations did not, however, appear to reflect differences in calcium uptake (Figure 4D). Decreased B-cell lipid packing (onset of MC540 uptake), shrinkage, and membrane buckling were concomitant (Figure 5A-B). Both membrane distortion and decreased lipid packing depended on cell shrinkage, as evidenced by their inhibition by clotrimazole, which block K_Ca3.1-dependent K⁺ efflux, and also by tamoxifen, which inhibits volume-regulated chloride channels.^22,23  DIDS and glybenclamide, which are thought to inhibit loss of lipid asymmetry downstream of cell shrinkage, had no effect on MC540 uptake (not shown). Together, these data strongly suggest that decreased lipid packing requires cell shrinkage and membrane buckling, and that all precede PS exposure.

We have recently shown that PS translocation in T cells is negatively regulated by the tyrosine phosphatase CD45.⁹  Consistent with this finding (and in contrast to parental B cells) we found PS to be constitutively exposed on CD45^-/- B cells (Figure 6). One possible explanation for the elevated exposure of PS on B cells from CD45^-/- mice is that these cells might lack normal signaling pathways mediating survival, thereby leading to increased apoptosis. However, we have found no evidence for increased apoptosis in peripheral T or B cells from CD45^-/- mice, assessed using 4 different assays (Figure 6, and Figure S2). It therefore seems unlikely that the high PS expression on more than 90% CD45^-/- B cells could be significantly explained by apoptosis, particularly as B-cell numbers are higher in CD45^-/- mice than the parentals.²⁴  Thus, B cells from CD45^-/- mice are viable but have exposed PS. Importantly, unstimulated CD45^-/- B cells exhibit higher SSC and MC540 binding than do parental B cells stained in the same tube (Figure 6), consistent with the hypothesis that membrane curvature and decreased lipid packing promote PS exposure.

The finding that PS exposure depends on prior cell shrinkage, an increase in membrane buckling, and decreased lipid packing suggests a model whereby calcium ionophore-stimulated PS exposure might occur at energetically favorable sites on developing membrane blebs (see “Discussion”). We therefore analyzed MC540 binding and PS exposure by calcium ionophore-stimulated lymphocytes by confocal microscopy. As we have found MC540 to reduce the binding of AV per cell, the 2 stains were analyzed separately. Figure 7A shows a stack of confocal sections through calcium ionophore-stimulated lymphocytes stained with AV. Blebs are visible as small protrusions at the cell surface. Consistent with the hypothesis, binding of both AV and MC540 occur preferentially at sites of membrane blebbing (Figure 7). Bleb localization is somewhat more apparent for AV than MC540, consistent with the fact that MC540 exhibits greater background binding to unstimulated cells.

In most cells, phospholipids are distributed asymmetrically in the plasma membrane, with the great majority of PS in the inner leaflet. PS translocation to the outer leaflet is required for the immunologically silent removal of apoptotic cells.²⁵  PS exposure can also occur on nonapoptotic cells, for example on murine T lymphocytes following stimulation of the ATP receptor P2X₇ (note: murine B cells do not express P2X₇).⁸  Moreover, PS is constitutively exposed at the surface of many cells of the CD45RB^lo subpopulation of viable CD4⁺ activated/memory lymphocytes. Importantly, PS distribution between inner and outer leaflets modulates the activity of several membrane proteins.⁹ 

Reports that viable murine B cells, in contrast to T cells, do not exhibit lipid asymmetry have been suggested to reflect a key aspect of their biology.^10,11  Here, we show that, in contrast to published findings,^10,16  PS in B cells, as in T cells, is in fact largely confined to the inner leaflet of the plasma membrane and surface exposure can readily be induced following stimulation with a calcium ionophore. A likely explanation for the previous report that PS distribution is not asymmetric is the observation that B cells spontaneously translocate PS to the cell surface relatively rapidly ex vivo. In contrast to previous studies, we did not physically isolate B cells, but instead gated cells electronically during analysis of flow cytometric data, thereby reducing time spent by cells ex vivo.

The rate of PS translocation in B cells, as measured by the proportion of responding cells at a given time, was found to be strain dependent. These strain-dependent events also correlated with differences in rates of MC540 uptake, supporting the hypothesis that decreased lipid packing is required for PS translocation. Significantly, in view of previous findings that lymphocytes from patients with rheumatoid arthritis and systemic lupus erythematosus have relatively high rates of PS translocation,¹⁷  the rate of PS exposure was highest in B cells from the autoimmune-prone strains NZW and NOD. Both NZW and NOD strains are also associated with susceptibility to lupuslike disease.^26-28  In principle, a tendency toward PS exposure might promote susceptibility to SLE through increases in shedding of proinflammatory microvesicles,^29,30  presentation of PS itself as an autoantigen,³¹  or subsequent increased release of other autoantigens in the form of apoptotic debris. As the increased rate of PS translocation in NOD and NZW lymphocytes is downstream of an increased rate of cell shrinkage and concomitant decrease in lipid packing, the data suggest that aberrant volume regulation may contribute toward susceptibility to lupuslike disease.

We have shown previously that ionophore-stimulated PS exposure is subsequent to and dependent on cell shrinkage caused by K⁺ efflux via K_Ca3.1.²¹  We show here that cell shrinkage results in a decrease in lipid packing, as evidenced by increased uptake of the dye MC540, presumably a consequence of membrane buckling. Thus, significant changes in the organization of the lipid bilayer appear to precede PS translocation. Uptake of MC540 and decreased lipid packing occur simultaneously and are both blocked by inhibitors of volume regulatory ion channels (along with cell shrinkage). Hence, PS translocation appears to depend on upstream cell shrinkage and concomitant decrease in lipid packing. Tamoxifen, a potent inhibitor of volume-regulated chloride channels, blocked PS exposure at concentrations as low as 1 μM, making it, to our knowledge, the most potent inhibitor of lipid scrambling described to date.

It is an apparent paradox that lipid packing decreases upon cell shrinkage, whereas an increase in packing might be anticipated. Nevertheless, as has been observed previously,^14-16  distortion of the plasma membrane (measured by increase in SSC) is concomitant with cell shrinkage early in apoptosis. We suggest, therefore, that buckling of the plasma membrane as cells shrink distorts the lipid bilayer locally such that MC540 inserts into the plasma membrane at sites of decreased lipid binding at the apex of blebs (Figure 8). PS⁺ blebs/microvesicles are known to form and be shed following stimulation with calcium ionophore.^2,32,33,35  Our observation that cell shrinkage, which is required for PS exposure, is strongly associated with decreased lipid packing provides a model for headgroup-nonspecific loss of lipid asymmetry (“scrambling”; Figure 8). In this model, bending of the membrane during cell shrinkage creates domains at the base and apex of blebs in which phospholipids in the outer leaflet are in relative excess or deficit, respectively, creating regions favorable for phospholipid flip-flop. Movement of inner leaflet phospholipids such as PS to the outer leaflet would be favored at the apex of blebs and that of PC to the inner leaflet favored at the base. This model predicts that phospholipid movement is bidirectional and headgroup nonspecific (consistent with the phenomenon of phospholipid “scrambling”), and that PS exposure is predominantly found on shed particles and is less apparent on the cell bodies that remain following shedding.^32,33  Consistent with our hypothesis, MC540 binding and PS exposure on calcium ionophore-stimulated lymphocytes both localized to sites of membrane blebbing (Figure 7). The model also implies that a dedicated protein may not be required to mediate PS translocation. The search for a specific protein that mediates bidirectional transport has been a notable failure.^36,37  While it has been shown that the rate of PS exposure is sensitive to altered expression of the ABC transporter, ABCA1, in human B cells,³⁸  murine erythrocytes,⁴  and following ABCA1 transfection,^4,38  the rate of PS exposure by murine ABCA1^-/- and parental B cells could not be distinguished (Figure 3). Hence, whereas ABCA1 can modulate rates of PS exposure, it does not appear to be required. If ABCA1 is a floppase, therefore, its loss can be compensated for by another enzyme in genetically modified murine B lymphocytes. Alternatively, and perhaps more likely, ABCA1 is not itself a floppase but can modulate the rate of PS translocation indirectly.

Based on studies of bacterial membranes, it has recently been suggested that phospholipid flip-flop may be promoted by a wide range of α-helical membrane-spanning proteins that provide energetically favorable sites for charged headgroups to traverse the hydrophobic interior of bilayer.³⁹  This model, which is complementary to our own, removes the need to posit (but does not disprove) the existence of a specific bidirectional phospholipid transporter.