Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
We provide evidence that Salmonella typhimurium kills phagocytes by an unusual proinflammatory mechanism of necrosis that is distinguishable from apoptosis. Infection stimulated a distinctly diffuse pattern of DNA fragmentation in macrophages, which contrasted with the marked nuclear condensation displayed by control cells undergoing chemically induced apoptosis. In apoptotic cells, DNA fragmentation and nuclear condensation result from caspase-3-mediated proteolysis; caspases also subvert necrotic cell death by cleaving and inactivating poly ADP-ribose polymerase (PARP). Caspase-3 was not activated during Salmonella infection, and PARP remained in its active, uncleaved state. Another hallmark of apoptosis is sustained membrane integrity during cell death; yet, infected macrophages rapidly lost membrane integrity, as indicated by simultaneous exposure of phosphatidylserine with the uptake of vital dye and the release of the cytoplasmic enzyme lactate dehydrogenase. During experimentally induced necrosis, lethal ion fluxes through the plasma membrane can be prevented by exogenous glycine; similarly, glycine completely blocked Salmonella-induced cytotoxicity. Finally, inhibition of the interleukin (IL)-1-converting enzyme caspase-1 blocked the death of infected macrophages, but not control cells induced to undergo apoptosis or necrosis. Thus, Salmonella-infected macrophages are killed by an unusual caspase-1-dependent mechanism of necrosis.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Apoptosis and necrosis are two forms of eukaryotic cell death that can be distinguished biochemically and morphologically (Majno and Joris, 1995). Apoptosis requires ATP and the activation of specific proteases. Dying cells display membrane blebbing, nuclear condensation and, eventually, cell shrinkage, yet maintain membrane integrity until late in this process. Necrosis results from membrane damage either directly or indirectly because of energy depletion, and yields swollen cells that leak their cytoplasmic contents. As a result, necrosis is strongly proinflammatory in vivo, whereas apoptotic cells are rapidly phagocytosed and thus generate minimal inflammation (Majno and Joris, 1995).
Because the mechanism of host cell death can influence the inflammatory response, discriminating apoptosis from necrosis may provide insight into the pathophysiology of infectious diseases caused by intracellular pathogens. The nuclear changes considered characteristic of apoptotic cells include condensation of nuclei and DNA fragmentation (McGahon et al., 1995). However, the latter event was recently also observed to occur during necrosis (Dong et al., 1997; de Torres et al., 1997; Orita et al., 1999), such that detecting DNA cleavage in cells does not necessarily identify apoptosis.
Caspases are a family of cysteine proteases important in apoptosis (Cohen, 1997). Caspase-3 activates ‘executioner’ caspases and cleaves other substrate proteins during the apoptotic death programme (Porter and Janicke, 1999). Although caspase-3-independent pathways of apoptosis exist, this enzyme is essential for nuclear condensation and DNA cleavage (Janicke et al., 1998; Woo et al., 1998). Salmonella infection provokes macrophage DNA fragmentation (Chen et al., 1996a; Lindgren et al., 1996; Monack et al., 1996; Richter-Dahlfors et al., 1997), but it is unknown whether caspase-3 is required for this event. Caspase-1, an interleukin (IL)-1-beta-converting enzyme required for Salmonella-induced macrophage cytotoxicity, interacts with the Salmonella secreted protein SipB (Hersh et al., 1999). The caspase-1 protease was originally classified as a proapoptotic enzyme based on homology to CED-3, a protein important in Caenorhabditis elegans cell death (Miura et al., 1993). Although caspase-1 is required for inflammation, it has been shown to be dispensable for apoptosis (Colussi and Kumar, 1999): caspase-1-deficient mice fail to produce IL-1β and resist endotoxic shock, but demonstrate no significant defects in apoptosis (Kuida et al., 1995; Li et al., 1995; 1997).
Macrophage apoptosis can be prevented by interactions with activated CD4+ T cells expressing the CD40L surface receptor and secreting macrophage-activating cytokines interferon (IFN)-γ and IL-12 (Suttles et al., 1996; Estaquier and Ameisen, 1997); therefore, CD4+ T cells responding to Salmonella (Nauciel, 1990; Cookson and Bevan, 1997) may potentially help infected macrophages to avoid apoptosis. To begin investigating this possibility, we studied Salmonella-infected macrophages using methods that discriminate between apoptotic and necrotic cell death. In contrast to chemically induced apoptosis, infected macrophages fragment their DNA but do not activate caspase-3, and rapidly sustain membrane damage that releases their cytoplasmic contents. Exogenous glycine prevents pathological ion fluxes through the plasma membranes of necrotic cells, and correspondingly blocks the death of Salmonella-infected macrophages. In addition, a key enzymatic substrate, normally cleaved during apoptosis to avoid the stimulation of necrotic cell death, remains in its active state in infected cells. However, unlike apoptosis or necrosis, death of infected macrophages requires caspase-1. Thus, Salmonella-infected macrophages are killed by an unusual caspase-1-dependent mechanism of necrosis.
DNA fragmentation without nuclear condensation in Salmonella-infected macrophages
Salmonella infection of macrophages induces rapid membrane damage
Fluorescence microscopy is a powerful tool for discerning cellular morphology, but it does not necessarily reveal features of the entire host cell population, because analyses are limited to relatively small numbers of adherent cells. To evaluate large numbers of Salmonella-infected macrophages, we used a flow cytometric assay that discriminates between apoptosis and necrosis: a fluorescently labelled phospholipid-binding protein, annexin V, was used specifically to detect phosphatidylserine (PS), a membrane lipid that is normally localized to the inner leaflet of the cellular plasma membrane (van Engeland et al., 1998). Early apoptotic cells display PS in their outer leaflet while maintaining membrane integrity. In contrast, necrotic cells expose PS as a result of membrane damage, and therefore exhibit simultaneous uptake of membrane-impermeant dyes such as 7-aminoactinomycin D (7-AAD) (van Engeland et al., 1998). Both Salmonella infection and gliotoxin treatment stimulated macrophage PS exposure to similar levels, as detected by staining with fluorescein isothiocyanate (FITC)-annexin V (Fig. 2A). Unlike apoptotic macrophages, however, Salmonella-infected cells positive for annexin V binding also stained with 7-AAD (Fig. 2B), which indicates membrane disruption of the infected host cells. Examination of Salmonella-infected cells at earlier time points (1 h, 2 h and 3 h) after invasion did not reveal apoptotic populations (FITC-annexin V positive, 7-AAD negative; data not shown), and macrophages infected with a prgH mutant did not show annexin V or 7-AAD staining (Fig. 2).
Although membrane damage eventually typifies both apoptosis and necrosis in vitro, the temporal association of the stimulus initiating cell death and detectable membrane damage is distinct. Necrotic cells lose membrane integrity early after injury, whereas apoptotic cells delay this until later stages in their programme of death. Indicative of membrane damage, infected and gliotoxin-treated macrophages released similar levels of the cytoplasmic enzyme lactate dehydrogenase (LDH) during a 6 h time course (Fig. 3). Rapid LDH release by infected macrophages shows that significant cytoplasmic leakage occurred within 2 h after bacterial invasion (Fig. 3), and similar kinetics were observed using infected thioglycollate-elicited macrophages (data not shown). This was not simply a result of bacterial infection, as the prgH mutant failed to stimulate LDH release (Fig. 3), which confirms the requirement of the SPI1 TTSS for Salmonella to cause macrophage death (Chen et al., 1996a; Hersh et al., 1999). In contrast to infected macrophages, LDH release by apoptotic macrophages was markedly slower and did not reach similar levels until 6 h after gliotoxin treatment (Fig. 3). These observations further define cytotoxicity as a process characterized by diffusely distributed DNA fragmentation and rapid membrane damage, which results in exposure of PS, uptake of vital dye and release of cytoplasmic contents.
Host cell death is independent of apoptotic caspases
Apoptosis and necrosis use different enzymatic pathways to execute death, and the caspase-3 protease activated during many apoptotic programmes (Porter and Janicke, 1999) is required for DNA fragmentation (Janicke et al., 1998; Woo et al., 1998). Because infection stimulated macrophage DNA cleavage (Fig. 1), we examined the role of caspase-3 during Salmonella-induced cell death. In contrast to gliotoxin-treated apoptotic cells, infected macrophages did not activate caspase-3 at any time during a 6 h infection (Fig. 4A). Similar results were obtained with infected peritoneal macrophages (data not shown). However, Salmonella-infected cells remained competent for caspase-3 upregulation by gliotoxin treatment (Fig. 4B), demonstrating that infection neither inhibits activated caspase-3 nor obscures its experimental detection. This suggests that infection-stimulated DNA cleavage and cytotoxicity are caspase-3-independent events. As caspase-3 functions downstream of caspase-9 and is also required for caspase-2, -6, -8 and -10 activation during apoptosis (Slee et al., 1999), the absence of caspase-3 activation in infected macrophages strongly suggests that Salmonella induces cell death by a mechanism distinct from apoptosis.
During apoptosis, cleavage and inactivation of poly ADP-ribose polymerase (PARP) by one or more caspases (Rosen and Casciola-Rosen, 1997) is critical for preserving cellular ATP and therefore the energy required to complete programmed cell death (Ha and Snyder, 1999). DNA damage, including strand breakage, activates PARP to consume substrate NAD+, which leads to ATP depletion and cellular necrosis (Ha and Snyder, 1999). In contrast to apoptotic macrophages, PARP remained in its active, uncleaved state in infected cells (Fig. 5). This suggests that the apoptotic pathways leading to PARP inactivation do not function during Salmonella-induced macrophage death. Together with the kinetics of membrane damage and release of cytoplasmic contents, and the distribution of DNA fragmentation without detectable caspase-3 activation, the mechanism of Salmonella-induced cytotoxicity is most consistent with necrosis.
Salmonella-induced macrophage death is prevented by inhibition of either caspase-1 or non-specific ion fluxes through the plasma membrane
Necrotic cell death resulting from ATP depletion can be prevented by treatment with exogenous glycine (Dong et al., 1997), which blocks the formation of non-specific plasma membrane leaks for small ions such as sodium (Frank et al., 2000). Glycine treatment also blocks necrosis induced by mitochondrial poisons, cold ischaemia and hypoxia (Weinberg et al., 1987; Dong et al., 1997; Frank et al., 2000). Our observations suggest the possibility that the attendant DNA damage of Salmonella infection activates PARP to deplete ATP and subsequently cause death by necrosis. Consistent with this hypothesis, glycine completely prevented cell death induced by Salmonella infection or ATP depletion, but negligibly affected apoptosis induced by gliotoxin (Fig. 6).
Recent work has demonstrated that Salmonella-induced cytotoxicity requires active caspase-1 (Hersh et al., 1999), which, unlike other homologous caspases, is not required for apoptosis (Kuida et al., 1995; Li et al., 1995; 1997). Pretreatment with a caspase-1 peptide inhibitor, YVAD, prevented infection-induced death of macrophages (Fig. 6), including thioglycollate-elicited peritoneal macrophages (data not shown). This confirmed a caspase-1-dependent mechanism of cytotoxicity in our experimental system, which was observed previously by Hersh et al. (1999) using RAW 264.7 and peritoneal macrophages. Significantly, caspase-1 inhibition did not prevent experimentally induced apoptosis or necrosis (Fig. 6). We conclude that infected macrophages release their cytoplasmic LDH and probably die as a result of non-specific leaks in their plasma membrane, which can be blocked by exogenous glycine. Unlike conventional necrosis, this process requires the IL-1-converting enzyme caspase-1, and thus Salmonella induces cytotoxicity by a unique mechanism.
We demonstrated that S. typhimurium induces macrophage death via an unusual caspase-1-dependent mechanism of necrosis. The evidence supporting this conclusion is presented in Table 1. An intact SPI1 TTSS system, responsible for secreting SipB, was required for Salmonella to stimulate macrophage DNA fragmentation. Caspase-3, the protease required for DNA fragmentation during apoptosis, was not activated after Salmonella infection, nor did infected macrophages display the condensed nuclei typical of apoptotic cells. Salmonella-infected macrophages demonstrated rapid loss of membrane integrity, release of cytoplasmic LDH and uptake of vital dye concomitant with PS exposure. Infected cells retained PARP in its active uncleaved state, suggesting the possibility that PARP activation by damaged macrophage DNA caused ATP depletion and death by necrosis (Ha and Snyder, 1999). Experimental ATP depletion causes necrotic membrane damage, which is blocked by exogenous glycine (Dong et al., 1997), and glycine treatment also blocked Salmonella-induced cytotoxicity. Conversely, gliotoxin treatment of macrophages induced bona fide apoptosis, as demonstrated by DNA fragmentation, nuclear condensation, caspase-3 activity, PS exposure without membrane damage and PARP proteolysis. Finally, Salmonella-induced cytotoxicity required active caspase-1, unlike apoptosis induced by gliotoxin or necrosis induced by ATP depletion.
Table 1. Summary of observations.
Yes (late in death)
Yes (early in death)
Yes (early in death)
Blocked by glycine
The mechanism of Salmonella-induced cytotoxicity was examined in a macrophage cell line and primary macrophages at several time points during infection using multiple criteria to distinguish between apoptosis and necrosis. Controls for both types of cell death were included, as were experiments with a non-cytotoxic bacterial mutant to exclude the possibility that our observations were simply the result of bacterial invasion or infection. Others have also observed features of Salmonella-infected macrophages that were consistent with apoptosis (Chen et al., 1996a; Lindgren et al., 1996; Monack et al., 1996; Richter-Dahlfors et al., 1997; Hersh et al., 1999), such as DNA cleavage and caspase-1-dependent cell death; however, neither of these events are apoptosis specific, as revealed in our investigation (Table 1) and supported by a growing body of evidence in the literature (Kuida et al., 1995; Li et al., 1995; 1997; Dong et al., 1997; de Torres et al., 1997; Orita et al., 1999). Further, it is unlikely that small populations of cells in the early or late stages of apoptosis would go undetected in our experiments using sensitive methods such as flow cytometry (Fig. 2) and Western blotting for PARP cleavage (Fig. 5). Thus, our results extend previous studies by demonstrating that infected macrophages die by a caspase-1-dependent mechanism of necrosis, a conclusion at odds with the current model of Salmonella-induced apoptosis.
Our data indicated that necrotic death of infected macrophages can be prevented by glycine (Fig. 6), which blocks non-specific ion fluxes through the plasma membrane (Frank et al., 2000). We conclude that an unusual mechanism accounts for Salmonella-induced cytotoxicity: necrotic cell death experimentally induced by ATP depletion was also prevented by glycine, but did not depend upon the activity of caspase-1 (Fig. 6). Inflammation at sites of infection in vivo (Jones and Falkow, 1996) may therefore result from both the production of IL-1β and IL-18 by caspase-1 and the release of cytoplasmic contents from infected, necrotic cells. Collectively, these data suggest that caspase-1 activation triggers non-specific plasma membrane leaks, possibly by using effector pathways common to other forms of necrosis, such as cold ischaemia, hypoxia, mitochondrial poisons and ATP depletion (Weinberg et al., 1987; Dong et al., 1997; Frank et al., 2000).
The requirement for other enzymes during Salmonella-induced macrophage death remains largely unknown. For example, are serine proteases required for infected cell death? DNA fragmentation during necrosis results from serine protease activation, in contrast to caspase (cysteine protease)-mediated DNA fragmentation during apoptosis (Dong et al., 1997). Interestingly, exogenous glycine also inhibits degradative intracellular proteolytic activity in necrotic cells during ATP depletion (Dickson et al., 1992). PARP may also function during Salmonella-induced cytotoxicity. This polymerase acts in a polar manner during cell death: inactivation of PARP by cleavage preserves cellular energy levels and allows apoptosis to proceed, whereas damaged DNA activates uncleaved PARP to deplete ATP, leading to membrane damage and necrosis (Ha and Snyder, 1999). The Salmonella-induced DNA damage may therefore activate PARP, resulting in ATP depletion and the formation of plasma membrane leaks that can be blocked by glycine (Fig. 6). Fibroblasts from PARP-deficient mice are protected against ATP depletion and necrotic death, but remain sensitive to apoptotic stimuli (Ha and Snyder, 1999). Thus, PARP–/– macrophages may also be resistant to Salmonella-induced cytotoxicity if death requires polymerase activity.
Numerous bacterial pathogens, including Salmonella, Shigella and Yersinia spp. (Zychlinsky et al., 1992; Monack et al., 1996; 1997; Mills et al., 1997; Ruckdeschel et al., 1997), have been described that induce apoptosis in eukaryotic cells (Weinrauch and Zychlinsky, 1999). These pathogens use TTSS to kill macrophages in vitro: the secreted proteins SipB (Salmonella), IpaB (Shigella), YopJ (Y. pseudotuberculosis) and YopP (Y. enterocolitica) are required to induce the death of infected cells (Zychlinsky et al., 1994; Chen et al., 1996a; Mills et al., 1997; Monack et al., 1997). Caspase-1 stimulation may be a common feature of TTSS-mediated cytotoxicity, as both SipB and IpaB bind and activate caspase-1 (Chen et al., 1996b; Hersh et al., 1999). These similarities suggest that Shigella also kills macrophages by a mechanism distinct from apoptosis: macrophages lacking the apoptosis effector proteins caspase-3, caspase-11 or p53 and macrophages overexpressing the cell death inhibitor proteins Bcl-2 or Bcl-XL remain susceptible to Shigella-induced cytotoxicity, whereas caspase-1-deficient macrophages are resistant (Hilbi et al., 1998). While this manuscript was in preparation, Salmonella was shown to induce cytotoxicity of infected macrophages despite inhibition of caspase-3 (Watson et al., 2000). Collectively, these data and our observations suggest that these Gram-negative bacteria induce cell death by a unique pro-inflammatory mechanism most closely resembling necrosis.
Infections with Salmonella, Shigella and Yersinia result in macrophage death and extensive inflammation (Weinrauch and Zychlinsky, 1999), and the production of IL-1 and IL-18 by caspase-1 is required for host inflammatory responses (Li et al., 1995). This suggests that caspase-1-dependent cell death specifically leads to inflammation, which is central to the pathogenesis of these bacteria in vivo. Supporting this hypothesis, the pathophysiology of disease is altered in caspase-1 knockout mice infected with Shigella or Salmonella: the acute inflammatory response and macrophage death were reduced during Shigella infection (Sansonetti et al., 2000), whereas Salmonella was unable to colonize tissues deeper than Peyer's patches after oral infection (Monack et al., 2000). The strong correlation between bacteria-induced cytotoxicity and inflammation supports our model of cell death, as conventional apoptosis does not stimulate an inflammatory response in vivo (Ren and Savill, 1998). This raises the question as to how to describe bacterial-induced host cell death that appears to be inflammatory in nature. We propose that S. typhimurium kills macrophages by an unusual mechanism: this pro-inflammatory process requiring the Salmonella SPI1 TTSS system and host caspase-1 ultimately gives rise to necrotic cell death resulting from pathological ion fluxes in the plasma membrane.
Bacterial strains, macrophages and growth conditions
S. typhimurium strain SL1344 and its prgH1::TnphoA derivative (Behlau and Miller, 1993) were used for all experiments. Bacteria were grown as described previously (Chen et al., 1996a). Briefly, overnight cultures back-diluted 1:15 into L-broth containing 0.3 M sodium chloride were grown at 37°C with shaking for 3 h, washed and resuspended in cold sterile PBS, and kept on ice before macrophage infections. The macrophage-like cell line J774A.1 was obtained from the American Type Culture Collection. Peritoneal exudate macrophages were selected as the plastic-adherent population from C3H/HeJ mice 3 days after 1 ml of Brewer's thioglycollate intraperitoneally (Cookson and Bevan, 1997). Macrophages were cultured at 37°C in 7% CO2 in Dulbecco's minimal essential medium (DMEM; Gibco BRL) supplemented with 10% FCS, 5 mM HEPES, 0.2 mg ml−1l-glutamine and 0.05 mM β-mercaptoethanol.
Gliotoxin (Sigma) was used at 5 µM to induce macrophage apoptosis (Waring et al., 1988; Monack et al., 1996). To deplete cells of ATP and induce necrosis, macrophages in glucose-free Ringer's buffer were treated for 3 h with ionomycin (Calbiochem) and carbonylcyanide-m-chlorophenylhydrazone (Sigma) at concentrations of 5 µM and 15 µM respectively (Dong et al., 1997). Caspase-1 was inhibited with acetyl-Tyr-Val-Ala-Asp-choloromethyl ketone (ac-YVAD-cmk, effective concentration 100–200 µM; Calbiochem) for 1 h before infection or induction of apoptosis or necrosis. Glycine protection was tested in macrophages pretreated with 5 mM glycine (Fisher Chemicals) for 1 h (Weinberg et al., 1987; Dong et al., 1997) before infection or induction of necrosis or apoptosis.
Macrophages in supplemented DMEM were allowed to adhere and mature in vitro for 24 h before infection. Peritoneal macrophages were cultured overnight in antibiotic-free media before infection. Using a previously described invasion protocol (Chen et al., 1996a), bacteria were added to macrophages at the indicated multiplicity of infection (MOI) in antibiotic-free media and allowed to invade for 2 h. After washing twice with Hanks' buffered salt solution (HBSS), media containing 15 µg ml−1 gentamicin were added to kill extracellular bacteria. The infected cells were allowed to incubate until the time points specified and then evaluated experimentally. Macrophage infection was confirmed by microscopy or lysing infected macrophages and determining the number of gentamicin-protected intracellular bacteria. At an MOI of 10, multiple experiments demonstrated that 98% of macrophages were infected by SL1344 (average of 10 bacteria/macrophage), whereas 78% were infected by the prgH1::TnphoA derivative (average of 3 bacteria/macrophage).
J774A.1 adherent to 12 mm2 glass coverslips were washed five times with PBS, fixed, permeabilized, and fluoresceinated nucleotide was incorporated enzymatically (In Situ Death kit; Boehringer Mannheim) to label nicked DNA. Cellular fluorescence was evaluated by microscopy (40 × objective), and images captured using a digital camera were equally adjusted for brightness and contrast and reduced in size for publication using Adobe Photoshop.
Adherent J774A.1 cells were removed from 25 mm2 flasks using PBS–EDTA and pooled with any non-adherent cells in the culture media. Cells were analysed for FITC-annexin V and 7-AAD (Sigma) staining as described previously (Li and Tait, 1998). Cells washed twice in annexin V binding buffer (10 mM HEPES-Na, 133 mM NaCl, 5.8 mM KCl, 5 mM glucose, 0.1% BSA, 2.5 mM CaCl2, pH 7.4) were resuspended to 1 × 107 cells ml−1, and 106 cells were incubated with 100 nM FITC-annexin V for 15 min on ice. After two washes in binding buffer, 7-AAD was added to 2 µg ml−1 final concentration, and cells were immediately analysed on a Coulter EPICS MCL flow cytometer.
Lactate dehydrogenase release
Supernatants from macrophages grown in 96-well plates in the presence of 5% FCS were evaluated for the presence of cytoplasmic enzyme LDH using the Cytotox 96 kit (Promega). Percentage cytotoxicity = 100 × (experimental LDH–spontaneous LDH)/(maximum LDH release–spontaneous LDH).
Adherent macrophages removed from 25 mm2 flasks using PBS–EDTA and pooled with non-adherent cells from culture media were washed once in PBS−0.1% BSA, and resuspended to 107 cells ml−1. Lysates from 106 cells were assayed for active caspase-3 using the substrate aspartate-glutamate-valine-aspartate-7-amino-4-tri-fluoromethyl coumarin (DEVD-AFC; ApoAlert Caspase-3 assay kit; Clontech). AFC release was measured fluorometrically at 505 nm. Fluorescent units (FU) fold induction was calculated as the fluorescence of the experimental sample incubated with substrate divided by the fluorescence of the PBS-treated control sample incubated with substrate. Active caspase-3 could be detected in a population with a minimum of 25% apoptotic cells (data not shown).
Adherent J774A.1 cells were removed from 25 mm2 flasks using PBS–EDTA and pooled with non-adherent cells from culture media. Total protein from 106 cells washed once in PBS−0.1% BSA was separated by 10% SDS–PAGE, and PARP cleavage was assessed by Western blotting using anti-PARP monoclonal IgG1 antibody clone C-2-10 (Clontech).
We thank Dr John Tait for generous use of recombinant fluorescein-annexin V and helpful comments regarding the manuscript, Dr Michael Jacobson and Philip Bergman for critical review of the manuscript, Dr Samuel Miller for kindly providing the prgH mutant, and Kathy Allen for excellent technical support in flow cytometry. B.C. was supported as a Pfizer Scholar during this work and by a Pilot Research Grant from the Howard Hughes Medical Institute.