Microbial entry through caveolae: variations on a theme

Authors


*For correspondence at Duke University Medical Center, Box 3020, Durham, NC 27710. E-mail soman.abraham@duke.edu; Tel. (+1) 919 684 3630; Fax (+1) 919 684 2021.

Summary

Caveolae and lipid rafts are increasingly being recognized as a significant portal of entry into host cells for a wide variety of pathogenic microorganisms. Entry through this mechanism appears to afford the microbes protection from degradation in lysosomes, though the level to which each microbe actively participates in avoiding lysosomal fusion may vary. Other possible variations in microbial entry through caveolae or lipid rafts may include (i) the destination of trafficking after entry and (ii) how actively the microbe contributes to the caveolae lipid/raft mediated entry. It seems that, though a wide variety of microorganisms are capable of utilizing caveolae/lipid rafts in various stages of their intracellular lifestyle, there can be distinct differences in how each microbe interacts with these structures. By studying these variations, we may learn more about the normal functioning of these cellular microdomains, and perhaps of more immediate importance, how to incorporate the use of these structures into the treatment of both infectious and non-infectious disease.

Introduction

Many and varied pathogens are known to enter into host cells, either as a means of avoiding the host's immune system, or as an integral part of their replicative cycle. One major hurdle intracellular pathogens must overcome is degradation in the classical endosomal/lysosomal fusion pathway. Some pathogens avoid this by escaping from their phagosomes before lysosomal fusion occurs, whereas others actively modify their compartment to prevent lysosomal fusion, and still others have evolved defences which allow them to survive in the harsh lysosomal environment (Small et al., 1994). It is being discovered that some pathogens have evolved a means of entering cells in a way that completely sidesteps the endosomal/lysosomal pathway altogether by utilizing discrete plasma membrane microdomains located on the host cell surface. These microdomains, commonly referred to as caveolae or lipid rafts, are enriched in cholesterol, glycosphingolipids and glycosylphosphatidylinosotol (GPI)-anchored molecules (Anderson, 1998). Caveolae were initially described as cave-like invaginations of the plasma membrane 50–100 nm in diameter which contained a distinctive protein, caveolin. Later study has revealed the necessity for a broader definition of caveolae because (i) caveolae are capable of assuming flat, vesicular, tubular and more ‘dynamic’ forms, in addition to the distinctive flask shaped invaginations and (ii) the protein caveolin is not found in all microdomains with the biochemical properties of caveolae (Schnitzer et al., 1996; Anderson, 1998). For the purposes of this review, we will define caveolae as pleomorphic lateral assemblies containing the protein caveolin-1 which are enriched in cholesterol, glycosphingolipids and GPI-anchored molecules (Anderson, 1998). A modifier such as flat, vesicular or flask-shaped (the different shapes are depicted in Fig. 1) will be used for clarification. Lipid rafts will be defined as pleomorphic lateral assemblies enriched in cholesterol, glycosphingolipids and GPI-anchored molecules, without the protein caveolin-1 (Harder and Simons, 1997; Simons and Ikonen, 1997; Kurzchalia and Parton, 1999). For ease of reading, except when it is necessary to make a clear distinction between caveolae and lipid rafts, the term caveolae will be used without a modifier to indicate both of these structures.

Figure 1.

Model depicting major defining characteristics of entry of various microbes utilizing caveolae and/or lipid rafts.

It would at first seem unusual that such a wide range of microbes would use the same endocytic pathway to gain entry into host cells. Indeed, that most of these pathogens bind distinct receptors on the host cell surface could lead one to envision multiple endocytic mechanisms for the various microbes. Another explanation for the observed phenomenon is that the various receptors recognized by each microbe are located within caveolae or move to caveolae after microbial contact. This has, in fact, been found to be the case for the receptors of many of the microbes and bacterial toxins that use caveolae (Lencer et al., 1999; Shin et al., 2000). The congregation of many different signalling molecules within caveolae, including molecules implicated in modulating the actin cytoskeleton, makes this an ideal pathway for microbial entry. That, coupled with the fact that microbes entering through this pathway avoid the potentially fatal consequences of lysosomal fusion, allows a greater understanding of why such varied microorganisms have evolved to utilize various host receptors that target the microbe to entry through caveolae. This review will highlight both the similarities and the differences in caveolae mediated entry of various pathogenic microorganisms into host cells, but will first look at the normal physiological role caveolae are thought to play in cellular function.

Physiological role of caveolae

Caveolae have been implicated in many cellular functions, one of which is acting as a conduit for transmembrane signal transduction because of the variety of receptors and signalling molecules that are concentrated in caveolae (Okamoto et al., 1998). For example, molecules found in caveolae include the receptors for epidermal growth factor (Mineo et al., 1996), platelet-derived growth factor (Liu et al., 1996), as well as the insulin receptor (Yamamoto et al., 1998). Signalling molecules such as H-Ras, protein kinase C and Src family protein tyrosine kinases are also found in caveolae (Okamoto et al., 1998). The protein caveolin-1, comprising 178 amino acids, has itself been implicated in signal transduction because of its direct interaction with various known signalling molecules through the caveolin scaffolding domain (amino acids 82–101) (Okamoto et al., 1998) and, in fact, caveolin-1 has been shown to be tyrosine phosphorylated during certain signalling events (Mastick et al., 1995; Mastick and Saltiel, 1997; Ushio-Fukai et al., 2001).

Caveolae have been implicated in mediating several different endocytic events, including the internalization of macromolecules such as nutrients, chemokines, hormones, etc. One such event, called potocytosis, is a possible method for pumping ions and small molecules into the cytoplasm of a cell. This in exemplified by the uptake of folate, whose GPI-anchored receptor is concentrated in caveolae (Anderson et al., 1992). Caveolae internalized ligands can also be shuttled to specific sites within the cell, as in the case of cholesterol which is translocated to the endoplasmic reticulum or some hormones which are translocated to the nucleus (Anderson, 1998; Harada et al., 1999). In still another event, the caveolae internalized ligand can be translocated across the cell and released outside at the basolateral surface, as happens to albumin and certain chemokines (Schnitzer et al., 1994; Middleton et al., 1997), or can be translocated in the reverse direction, basolateral to apical, as happens to IgA antibodies (Hansen et al., 1999).

Caveolae seem to be ideally suited as endocytic vehicles, based on a number of observations. One is the previously mentioned clustering of receptors and signalling molecules within caveolae. Another is the observa-tion that caveolae seem to contain all the machinery necessary for vesicle budding, docking and fusion (Schnitzer et al., 1995). The discovery that dynamin, a GTPase involved in the budding of clathrin-coated pits, is also required for caveolar budding is also indicative of the endocytic function of caveolae (Oh et al., 1998). Caveolae also seem to be linked to the actin cytoskeleton based on the localization of actin mobilizing/modulating molecules to caveolae and the recent report that the F-actin cross-linking protein filamin is a ligand for the protein caveolin-1 (Rozelle et al., 2000; Stahlhut and van Deurs, 2000). Although there are still many questions regarding the exact physiological role of caveolae, there is an increasing agreement that they play an important part in the everyday signalling and endocytic functions of mammalian cells.

That such vastly different pathogens such as the human immunodeficiency virus type-1 (HIV-1) (Bomsel, 1997; Alfsen et al., 2001), Chlamydia trachomatis (Norkin et al., 2001) and the malarial parasite Plasmodium falciparum (Lauer et al., 2000) all gain entry into host cells through a pathway involving caveolae or lipid rafts, or use these structures to aid in intracellular survival, could lead to the assumption that each microbe interacts with caveolae or lipid rafts in the same manner. The fact that HIV-1 and the infectious form of C. trachomatis (the metabolically inert elementary body) enter in a passive manner, whereas the malarial parasite enters by active penetration independent of host cell endocytic machinery is just one indication that, although there are distinct similarities in their interactions with caveolae and lipid rafts, distinct differences exist as well. Regardless of any existing differences, it is becoming clear that several pathogenic microorganisms have recognized the physiological endocytic functions of caveolae and lipid rafts as a flexible and exploitable means of gaining entry into host cells in a manner which aids in the avoidance of lysosomal fusion.

Caveolae-mediated viral entry

Many viruses have been found to require endocytosis through clathrin-coated pits to gain entry into their host cells. This leads to fusion with lysosomes, which has functional significance for many viruses, because the acidic pH of the lysosomal compartment usually leads to viral uncoating and entry into the cytoplasm (Marsh and Helenius, 1989). In contrast to this, a number of viruses enter into host cells in a manner that is inhibitable by cholesterol sequestering agents. In fact, many of these viruses have been specifically associated with flask-shaped caveolae by transmission electron microscopy (Table 1) (Anderson et al., 1996; Richterova et al., 2001; Marjomaki et al., 2002). The most thoroughly studied of these is simian virus 40 (SV40). According to one model, after binding to its receptor, major histocompatibility complex (MHC) class I, there appears to be a recruitment of caveolin-1 to the site of viral attachment from preformed caveolae, which ultimately leads to the formation of caveola around the virus (Stang et al., 1997), whereas another model suggests that virus bound MHC class I associates with pre-existing caveolae (Chen and Norkin, 1999). Regardless of the exact method, SV40 associates directly with flask-shaped caveolae leading to a loss of actin stress fibres and the appearance of actin tails emanating from the virus containing caveolae (Pelkmans et al., 2002). The SV40 containing flask-shaped caveolae then pinch off from the plasma membrane and transport their load to the ER via an organelle termed the caveosome (Pelkmans et al., 2001). It is unknown how the viral genome reaches the nucleus from the ER but it would seem to involve transport through the cytosol and the nuclear pore complex (Kasamatsu and Nakanishi, 1998).

Table 1. . Characteristics of microbial entry through caveolae and/or lipid rafts.
PathogenSite of entryNature of entry and persistenceReceptorDestinationReferences
  1. NA, Non-applicable.

Viruses
Simian virus 40Flask-shaped
 caveolae
NAMHC I is not associated with
 regular caveolae and does
 not enter with the virus
ER by way of
 caveosome
Stang et al. (1997)
Chen and Norkin (1999)
Polyoma virusFlask-shaped
 caveolae
NAUnknownER (by way of
 caveosome?)
Richterova et al. (2001)
Echovirus 1Flask-shaped
 caveolae
NA α2β1 integrin may be associated
 with caveolae and is internalized
 with the virus
Perinuclear region that
 may not be ER
Marjomaki et al. (2002)
RSVCaveolae (shape
 unknown)
NAUnknownIn dendritic cells entry
 leads to antigen
 presentation
Werling et al. (1999)
EBOV/MBGVCaveolae (shape
 unknown)
NAGPI anchored folate receptor-αUnknownBavari et al. (2002)
Empig and Goldsmith
 (2002)

Chan et al. (2001)
HIV-1
(Productive infection
 of CD4+ T cells)
Lipid rafts
(no reported
 association
 with caveolin)
NACD4, CCR5, CXCR4UnknownLiao et al. (2001)
Popik et al. (2002)
HIV-1
(Epithelial/endothelial
 cell entry)
Lipid rafts (no
 reported association
 with caveolin)
NAEpithelial receptor GalCer is
 normally associated with
 lipid rafts
TranscytosisAlfsen et al. (2001)
Bomsel (1997)
Liu et al. (2002)
Bacteria
FimH-expressing E.coliFlat and/or vesicular
 caveolae
Passive entry and passive persistenceCD48NAShin et al. (2000)
C. trachomatisFlat and/or vesicular
 caveolae
Passive entry and active persistenceUnknownNANorkin et al. (2001)
C. jejuniLipid raft (no caveolin
 association)
Active entry and active/passive
 persistence
UnknownNAWooldridge et al. (1996)
MycobacteriumLipid raft? Associated
 with cholesterol after
 entry
Passive entry and
 active persistence
CR3 and an unknown
GPI anchored molecule
NAGatfield and Pieters (2000)
Peyron et al. (2000)
Brucella spp.Lipid raft? Associated
 with cholesterol and
 GPI-anchored proteins
Active/passive entry and active
 persistence
GPI-anchored protein?NANaroeni and Porte (2002)
Watarai et al. (2002)
S. typhimuriumLipid raft? associated
 with cholesterol and
 GPI-anchored proteins
Active entry and active persistenceUnknownNACatron et al. (2002)
Garner et al. (2002)
Parasites
P. falciparumNAActively penetrates host cell membrane,
 acquires lipid raft associated
 molecules, and cholesterol is
 required for intracellular persistence
UnknownNALauer et al. (2000)
T. gondiiNAActively penetrates host cell membrane,
 and acquires lipid raft associated
 molecules
UnknownNAMordue et al. (1999a)
Mordue et al. (1999b)

Both the polyoma virus and echovirus 1 (EV1) also associate directly with flask-shaped caveolae (Richterova et al., 2001; Marjomaki et al., 2002). Polyoma virus entry shows many similarities to the entry of SV40 including an association with the protein caveolin-1 after entry, a possible association with ‘caveosomes’ and trafficking to the ER (Richterova et al., 2001). EV1 entry, however, shows some dissimilarities one being that it is transported to perinuclear structures negative for markers of ER (Marjomaki et al., 2002). Possible differences could be determined by the receptors used by the two viruses. SV40 utilizes MHC class I, whereas EV1 uses α2β1 integrin (Atwood and Norkin, 1989; Bergelson et al., 1992; Breau et al., 1992). MHC class I is not believed to normally associate with caveolae and is not internalized with SV40 particles (Anderson et al., 1998). On the other hand, α2β1 integrin remains associated with EV1 after entry and both the α2 and β1 integrin subunits have been reported to associate with caveolin-1 (Wei et al., 1996; Wary et al., 1998; Marjomaki et al., 2002). The importance of the caveolin-1 protein in viral entry remains uncertain, though the expression of a dominant negative form of caveolin can block the entry of both SV40 and EV1 (Roy et al., 1999; Marjomaki et al., 2002). Caveolae may also be involved in the uptake of respiratory syncytial virus (RSV) by dendritic cells (Werling et al., 1999) and the filoviruses Ebola (EBOV) and Marburg (MBGV) by a variety of cell types (Bavari et al., 2002; Empig and Goldsmith, 2002), where internalized virus is found associated with caveolin-1, although no direct association with a particular shape of caveolae has been reported. Caveolin-1 is thought to be required, but not solely responsible, for the formation of flask-shaped caveolae (Vogel et al., 1998) and distinct differences in signalling mechanisms have been observed between flask-shaped caveolae and flat caveolae, which could potentially effect how a microbe interacts with a host cell (Sowa et al., 2001).

Recent work has shown that receptors for the human immunodeficiency virus (HIV) (CD4, CCR5 and CXCR4) are associated with lipid rafts in T cells and the disruption of lipid raft integrity by cholesterol depletion using methyl β-cyclodextrin greatly inhibits productive infection (Liao et al., 2001; Nguyen and Taub, 2002; Popik et al., 2002). In a markedly different interaction between a virus and host cell lipid rafts, HIV interacts with galactosyl ceramide (GalCer) on the apical surface of epithelial cells leading to internalization and the direct transcytosis of the virus through the cell and its release outside the basolateral surface, an interaction that does not lead to the infection of the epithelial cells (Bomsel, 1997; Alfsen et al., 2001; Campbell et al., 2001). Lipid raft mediated transcytosis of HIV may also occur in brain microvascular endothelia (Liu et al., 2002). In addition to the use of lipid rafts for entry into CD4+ T cells and penetration of the epithelial/endothelial barrier, it is thought that HIV also utilizes lipid rafts for its replication in and escape from permissive host cells (Nguyen and Hildreth, 2000), thus implicating lipid raft involvement in almost every stage of the HIV lifecycle.

The differences seen in viral entry through caveolae or lipid rafts may be, at least in part, mediated by the respective receptors utilized by the different viruses. It is also conceivable that two or more molecules can interact with one another differently in the context of flask-shaped caveolae versus flat caveolae versus lipid rafts. This was demonstrated recently by Sowa et al. (2001) where they showed a differential in the interaction between endothelial nitric oxide synthase (eNOS) and caveolin-1 in the context of flask-shaped caveolae versus flat caveolae. This raises the possibility that a similar set of signal transduction machinery could be sending different signals depending on the architecture of the local environment. Further research should be performed to determine if there is a requirement for a particular shape of caveolae and/or the protein caveolin-1 in viral entry.

Caveolae and lipid raft mediated entry of bacteria

Arguably, the major hurdle to a bacterial infection is the host's immune system. Many mechanisms have evolved to allow pathogenic bacteria to avoid host defences and one of the most studied is the entry of bacteria into host cells. Existing intracellularly affords the bacteria protection from the humoral arm of the immune system, giving them a protected niche in which to survive. Bacterial entry can be separated into two groups, active entry and passive entry. A paradigm of active entry is that of Salmonella typhimurium, where the bacterium is known to actively secrete various effector proteins into the host cell cytoplasm through its type III secretion machinery, thereby inducing uptake (Collazo and Galan, 1997; Suarez and Russmann, 1998). An example of passive bacterial entry is that of Chlamydia trachomatis where the infectious chlamydial elementary body (EB) is metabolically inert and yet is able to gain entry (Norkin et al., 2001).

The intracellular lifestyle is not without its own set of dangers, chief among them the potentially lethal effects of fusion of the bacteria containing endosome with the lysosomal compartment. Many bacteria have evolved mechanisms to avoid this fate. For example, S. typhimurium actively modifies its compartment to avoid fusion with lysosomes (referred to as active persistence), whereas Listeria monocytogenes, escapes from its phagosome and survives in the host cell's cytoplasm (Small et al., 1994). Another mechanism to avoid the endosomal/lysosomal compartment is to gain entry via caveolae or lipid rafts, which are believed to naturally avoid fusion with lysosomes (Table 1). As will be discussed below, some bacteria may merge the active avoidance of lysosomal fusion with entry through caveolae or lipid rafts.

One of the best-characterized interactions between bacteria and caveolae is the entry of the normally non-invasive type 1 fimbriated Escherichia coli into murine bone marrow derived mast cells (BMMC) (Shin et al., 2000). The interaction is mediated by the type 1 fimbrial adhesin, FimH and its cognate receptor on mast cells, the GPI-anchored protein CD48 (Malaviya et al., 1999). Escherichia coli is taken up into BMMC in a tight fitting compartment that is positive for various markers of caveolae such as cholesterol, the ganglioside GM1 and the protein caveolin-1 (Shin et al., 2000). In a more definitive experiment, FimH-mediated bacterial entry was found to be susceptible to the caveolae-disrupting agents filipin and methyl β-cyclodextrin, unlike opsonin mediated bacterial entry which was not inhibitable by these substances. As a result of utilizing the caveolar pathway to gain entry, the bacteria were able to avoid the potent bactericidal functions of the mast cells and survive intracellularly. It is believed that entry through caveolae is responsible for this avoidance of lysosomal fusion, as the laboratory strain of E. coli used in these experiments is not thought to modify its compartment after entry (referred to as passive persistence). Although no flask-shaped caveolae were found in mast cells, it is thought that many plasmalemmal (flat) and vesicular caveolae are recruited to sites of bacterial attachment to form large bacteria-encapsulating chambers and the protein caveolin-1 was visualized around internalized bacteria by immune electron microscopy (Shin et al., 2000).

Chlamydia trachomatis is also thought to exploit caveolae as a means of entry into both phagocytic and non-phagocytic cells (Norkin et al., 2001). Like the entry of E. coli into mast cells, C. trachomatis entry is inhibited by caveolae-disrupting agents and the bacteria containing compartments react with antisera against the protein caveolin-1. A significant difference is that C. trachomatis plays a much more active role in avoiding lysosomal fusion at later stages than E. coli, as the chlamydial vesicle traffics to the Golgi region and actively intercepts sphingolipid, and presumably caveolin-1, containing vesicles from the Golgi (Hackstadt et al., 1995; 1996) and some serovars of Chlamydia are believed to actively modify their compartment to avoid lysosomal fusion (Scidmore et al., 1996).

Other bacteria that have been potentially linked to entry via caveolae or lipid rafts include Campylobacter jejuni (Wooldridge et al., 1996), Mycobacterium bovis and Mycobacterium kansasii (Gatfield and Pieters, 2000; Peyron et al., 2000), Brucella suis and Brucella abortus (Naroeni and Porte, 2002; Watarai et al., 2002) and Salmonella typhimurium (Catron et al., 2002; Garner et al., 2002). The entry of these bacteria into various cell types has been shown to require cholesterol and in the cases of M. bovis, B. abortus and S. typhimurium, cholesterol accumulates at sites of bacterial entry. An association with the caveolar marker caveolin-1 has not been reported for any of these pathogens and the involvement of caveolin-1 is unclear, except in the entry of C. jejuni into the Caco-2 intestinal cell line which has been shown not to express caveolin-1 or form flask-shaped caveolae (Mirre et al., 1996; Vogel et al., 1998). Both Mycobacteria and Salmonella are thought to also actively modulate their compartments to avoid lysosomal fusion, again in contrast to the entry of E. coli into mast cells (Garcia-del Portillo, 1996; Ferrari et al., 1999).

The differences that exist in the entry of various bacteria into host cells through caveolae or lipid rafts could be mediated by a number of factors. One possibility is the receptor(s) on the host cells recognized by the bacteria, at least in cases where a specific receptor is known to mediate entry. For example, type 1 fimbriated E. coli recognize CD48 on BMMC whereas M. bovis entry into macrophages has been shown to involve CR3 and an unknown GPI-anchored molecule (Malaviya et al., 1999; Peyron et al., 2000). Another possibility is the extent to which the bacteria play an active role in their entry into host cells. The entry of E. coli into BMMC is entirely dependent on activity derived from the mast cells, as even dead bacteria can gain entry (Z. Gao and S. N. Abraham, unpublished data), whereas S. typhimurium is known to secrete effector proteins directly into the host cytoplasm to induce bacterial uptake (Collazo and Galan, 1997; Suarez and Russmann, 1998). The possibility arises that the specific way in which bacteria interact with caveolae or lipid rafts could lead to different levels of dependence upon these structures to gain entry into host cells and could potentially effect the level to which the bacteria must actively resist lysosomal fusion. Regardless, further study into the mechanisms that allow these microbes to target, co-opt and, in some cases, modify the caveolar endocytic pathway for their own benefit will allow a greater understanding of how we may do the same.

Interactions of parasites with lipid rafts

The parasites Toxoplasma gondii and Plasmodium falciparum enter into both phagocytic and non-phagocytic cells by active penetration that does not depend on the endocytic machinery of the host cell and is therefore quite distinct from the caveolae or lipid raft mediated entry of viruses and bacteria. The compartment formed by parasite entry, called the parasitophorous vacuole, resists fusion with lysosomes by unknown mechanisms (Mordue et al., 1999b). In the case of T. gondii, avoiding lysosomal fusion may be mediated in part by the route of entry because antibody-opsonized T. gondii are taken into compartments that fuse with endosomes and are eventually acidified (Joiner et al., 1990). Interestingly, the parasitophorous vacuoles of both T. gondii and P. falciparum acquire host GPI-anchored, transmembrane and cytosolic proteins thought to be associated with caveolae or lipid rafts (Mordue et al., 1999a; Lauer et al., 2000). The P. falciparum vacuole also acquires caveolae and lipid rafts associated host glycolipids and cholesterol has, in fact, been found to be vital to the maintenance of the intracellular existence of the parasite in erythrocytes (Lauer et al., 2000). This indicates that, whereas the entry of these parasites is quite distinct from the other microbes mentioned, caveolae and/or caveolar components play an important role in their pathogenic intracellular lifestyle.

Conclusion

Caveolae and lipid rafts are an increasingly recognized portal of entry for a surprising variety of pathogenic microorganisms (Fig. 1). These structures have been linked to the entry and/or intracellular survival of viruses, which have been associated with a particular type of caveolae (flask-shaped), bacteria, whether they gain entry and persist in a passive (E. coli) or active (S. typhimurium) manner and even large parasites (P. falciparum). Although these various microbes have all co-opted caveolae and lipid rafts, they seem to not interact with them in the same manner. Trafficking after entry, and even the level to which the microbes depend on caveolae and lipid rafts for entry and persistence, may vary from microbe to microbe. These differences could potentially be mediated by the specific receptor each microbe recognizes on the host cell surface, the type or shape of the microdomain involved in entry (lipid raft versus flask-shaped versus flat-shaped caveolae), or the particular nature of entry/persistence (passive versus active) each microbe expresses. Microbes need not be the only organisms to utilize the flexibility of caveolae and lipid rafts for targeting and gaining entry into eukaryotic cells. This has been recently demonstrated by McIntosh et al. (2002) who used an antibody which specifically targets lung endothelial caveolae to achieve lung specific delivery and transendothelial transport of the antibody to underlying tissue cells. This adds credence to the belief that further study of the nature of caveolae and lipid rafts, potentially using microbes as the tools of study, could allow new and exciting methods of specifically delivering therapeutic agents to treat a variety of illnesses, even microbial infections themselves.

Acknowledgements

This work was supported by NIH Grants DK 50814 and AI 50021 and an award to S.N.A. from The Sandler Family Supporting Foundation.

Ancillary