How PFN delivers GzmB into cells without causing detectable damage to the cell membrane is a matter of intense interest especially since an understanding of this phenomenon will allow the development of more rational strategies for intra-cellular protein delivery. The evidence presented here suggests that the mechanism involves lipid components of the membrane themselves, as well as PFN monomers inserted into the membrane. For example, single channel conductance measurements show that the physical characteristics of the pore including its size, its stability and its mechanism of formation are affected by the degree of ordering in the membrane lipids 
. Pore diameters formed by PFN appear to vary depending on membrane composition ranging from homogenous phospholipids of a single headgroup to erythrocyte membrane and the plasma membrane of nucleated cells 
. Despite stringent efforts to uncover the 20 nm pores, when PFN is applied to target cells at concentrations that contribute to granzyme delivery, even low molecular fluorophores such as PI do not enter the target cells. Instead, we present evidence that PFN reconfigures membrane lipids on live cells that results in unanticipated phospholipid flip-flop. We show by Ann-V and lactadherin staining that PS appears in the outer leaflet of the bilayer, and that this marks cells that are prepared for GzmB delivery. PS flip-flop is not caused by PFN-induced calcium influx and seems to increase when cells are exposed to an oLPC that enhances positive membrane curvature and by cholesterol depletion.
PFN appears to induce flip-flop of PS from the inner to the outer membrane leaflet that does not rely on the induction of intracellular translocases and flippases. The externalization of PS we observe can most easily be explained by the fusion of the inner and outer leaflets of the bilayer and the lipid's migration over the resulting lip, a process that does not occur during the formation of transmembrane proteinaceous channels. A related toroidal structure is known to arise during electroporation 
as well as during membrane fusion events that occur upon viral cell entry and transfer of intracellular cargo between vesicular compartments 
PFN monomers have been reported to clearly oligomerize into a cylindrical pore-forming proteinaceous channel which span the membrane bilayer to form a pore and this mechanism must contribute to certain biological effects mediated by the toxin 
. Many other proteins are known to form pores using a similar mechanism of oligomerization and insertion. Nevertheless, biophysical data have suggested that bacterial peptides such as melittin 
, toxins including Equinatoxin II 
and small proteins like Bax 
, all may induce pores consisting of a mixture of incompletely-oligomerized protein and the membrane phospholipids themselves. Whether larger toxins such as bacterial CDCs and the related MACPF domain proteins such as PFN form these structures is highly controversial 
but must be considered possible given the apparent effects of the membrane lipid components on pore structure, stability and lifetime 
. Pores formed by such matrices of protein and lipid would have the necessary characteristics to allow PS migration between PM leaflets and would be similar to pore structures induced during membrane fusion 
. Using a variety of biophysical approaches including electron microscopy 
, atomic force microscopy 
and X-ray diffraction 
, the existence of these unusual pores in model membranes has been argued to exist. However, even if evidence for their formation in vitro
is accepted, their appearance in biological membranes has remained unproven, due to their presumed evanescence as shown by transient Ca influx and the lack of technologies capable of imaging complex membranes at the required resolutions.
One of the major concerns in understanding how a membrane is altered by an applied protein is whether its insertion achieves the biological effect observed in vivo. PFN may be added to target cells at concentrations that induce widespread necrosis but such toxic levels neither mirror the conditions necessary for granzyme delivery nor reflect the levels encountered by a target cell after CTL engagement (see ). Our data establish for the first time guidelines that allow the identification of physiologically relevant PFN concentrations for cellular and biophysical studies. We hypothesized previously that PFN oligomers might disrupt the plasma membrane allowing granzymes to cross the bilayer in the absence of discrete pores 
. The phenomena we see evidenced in the movement of PS to the upper leaflet of the membrane would correlate with such an event.
Based on recent structural data, PFN is predicted to oligomerize to a single-sized pore (20 nm) regardless of extracellular conditions or membrane composition that serves as the platform for insertion. However, data reported here and elsewhere suggest that the structures formed by PFN are variable in size 
depending on at least two factors that influence binding density and oligomerization rates. These include 1) The concentration of Ca ions and inhibitory proteins 
that PFN encounters and 2) the composition and organization of the membrane that the monomers bind, oligomerize and then insert within. Together these conditions appear to influence the characteristics of pores with cylinders predominating above a concentration threshold while membrane structures that give rise to PS flip-flop are more common when the level of monomers is limiting (). These latter structures may be incomplete ring oligomers, or arcs, that have been observed previously 
Factors that influence the outcome of PFN interaction with cells.
While PFN is observed to form cylinders and arcs on liposomes and sheep erythrocyte ghosts when evaluated by EM, a systematic description of membrane alterations that might be induced by the toxin has not been undertaken. In this regard, we have learned that PFN induces membrane invaginations in both LUVs and giant unilamellar vesicles (manuscript submitted). Although additional work is necessary to understand how PFN induces such structures, it is intriguing to speculate that PFN arcs may generate positive membrane curvature that partly contributes to the observed invaginations and the counterpart to these morphological alterations is transient membrane fusion and phospholipid flip-flop when PFN is applied to nucleated cells in limiting amounts associated with granzyme delivery. Inasmuch as cylindrical pore structures have not been identified in target cells that would allow passage of the granzymes, the observed flip-flop becomes the first viable pathway that could offer a route for GzmB translocation. However, translocation would depend on the availability of GzmB to interact with membrane phospholipids of the outer leaflet. Since GzmB would preferentially bind the glycosaminoglycan (GAG) side chains of cell surface proteolgycans, a GzmB-phospholipid interaction would require a radical movement of the proteoglycans and their associated Glycosaminoglycan chains outward from the phospholipid binding sites on the leaflet. Such a possibility may occur efficiently in the CTL synapse where major alteration in the topological distribution of integral membrane proteins has been reported.
PFN induced Ca influx is considered to be mechanistically linked to granzyme delivery 
. In this model, PFN triggers a plasma membrane-repair response that depends on inward movement of Ca ions. The response then drives endocytic uptake of GzmB and its delivery, an outcome inhibited by BAPTA-AM. For these studies, an arbitrary sub-lytic dose of PFN was employed without distinguishing the subset of cells that PFN renders susceptible to GzmB. By adding PI before PFN, we could identify and recognize those cells that succumb to GzmB versus targets that became rapidly necrotic. We were thus surprised to observe that intracellular chelation of PFN-induced Ca influx did not reduce granzyme delivery and apoptosis induction when detected by a highly sensitive probe for caspase-3/-7 activation. This issue notwithstanding, Ca influx and PS flip-flop appear to represent entirely separate biological processes where Ca influx indeed may serve to protect cells as the intracellular levels of Ca rise with increasing deposition of cylindrical pores.
PFN-induced phospholipid flip-flop appears to correlate highly with target cells that undergo GzmB-induced apoptosis, but PS translocation may serve other functions. PFN is considered to act in host defense primarily as the delivery agent for the granzymes. However, PFN has been shown to be both necessary and sufficient in the absence of the granzyme A and B to eliminate tumor burden 
. PFN alone would be predicted to act by inducing tumor cell necrosis, which is viewed as deleterious to the host by inducing autoimmune responsiveness. Perhaps, PFN-induced phospholipid flip-flop is instead sufficient to stimulate phagocytic clearance of tumor cells though mechanisms described for apoptotic cells.
In summary, the observation that a 5 min pulse of PFN and GzmB is sufficient for apoptosis suggests that the granzyme may simply translocate through structures that reflect PS flip-flop at the plasma membrane. The validity of these concepts await our ability to image target cells treated with the physiologically relevant concentrations of the pore forming protein, to determine whether mutations that alter the oligomerization of membrane bound PFN encourages formation of the proposed arc-like structures, and to learn whether GzmB exploits these structures to translocate to the cytosol.