Na⁺-Independent Transport (Uniport) of Amino Acids and Glucose in Mammalian Cells

Transport proteins regulate the movement of lipid-insoluble nutrients across the hydrophobic domain of the plasma membrane. The capacity to transport amino acids and glucose is required by all cells and is essential for protein synthesis, for the supply of metabolic energy, for maintaining and replacing structural components and for other key physiological functions. This paper addresses amino acid (and to a lesser extent glucose) transport that is largely independent of co-substrate requirements (uniport).

Individual cells express a variety of distinct transport systems, each facilitating the flux of several different amino acid or sugar substrates. Furthermore, most amino acids can be transported by several different transport systems, some expressed in distinct organs and tissues. The broad substrate-specificity and overlapping tissue and cellular distribution patterns of amino acid transporters posed an analytical challenge which was innovatively addressed beginning in the 1950s and 1960s in work pioneered by H. N. Christensen and investigated by many others (for reviews, see Christensen, 1989; Van Winkle, 1993; Kilberg and Häussinger, 1992; Kilberg et al. 1993; McGivan and Pastor-Anglada, 1994; Bertran et al. 1994). Numerous transport systems were identified and characterized (Table 1). This research established methods and chemical analogs useful for discriminating the transport systems and allowed the initial analysis of the newly cloned transporters to progress quickly. We use the terms transport system to designate a functionally distinct transport process and transporter to designate a protein that catalyzes amino acid transport across a biomembrane (after Van Winkle, 1993).

Transport processes are classified according to structural or energetic aspects. Carrier-mediated transport can be distinguished from simple diffusion by the specific kinetic properties originally used to characterize enzymes (Krämer, 1994). It is limited by the number of available cell surface substrate binding sites and is demonstrated by transport rates measured when all binding sites are occupied (V_max); in contrast, simple diffusion is non-saturable and limited only by substrate concentration. The most widely accepted classification, introduced by Mitchell (1967), is based on the utilization of energy sources for transport. Primary transport involves vectorial solute translocation directly coupled to chemical or photochemical reactions. Secondary transport is driven solely by the electrochemical energy of a given solute, which can drive the uphill transport of some nutrients against their own concentration gradient and is achieved by cotransport (symport) or countertransport (antiport, Fig. 1).

The transport of cationic amino acids that is independent of co-substrate requirements (uniport) is driven simply by its own electrochemical gradient (Krämer, 1994; Kavanaugh, 1993) in a process often termed facilitated diffusion. This a misleading term since carrier-mediated transport is fundamentally different from diffusion (Krämer, 1994). Terms such as uniport and symport are confusing when describing amino acid transport systems because the transport of some substrates is ion-independent but that of others is ion-dependent (see Table 1). For example, amino acid transport system y⁺ requires Na⁺ for zwitterionic amino acid influx (symport), but it mediates cationic transport (lysine and arginine) in a Na⁺-independent manner (uniport; Christensen and Antonioli, 1969; White et al. 1982). The problem of naming cloned amino acid transporters in a manner recalling their precedent transport systems, substrate recognition and Na⁺ requirements is discussed by Christensen et al. (1994).

The Na⁺-independent transport of cationic amino acids is mediated by at least three distinct mCAT proteins (murine cationic αmino acid transporter) (for a review, see MacLeod et al. 1994) and also by a new class of proteins sometimes referred to as accessory proteins (for reviews, see Hediger et al. 1993; Bertran et al. 1994; Palacín, 1994). Cationic amino acid influx mediated by these transporters is driven by both substrate concentration and cellular voltage gradients. The negative resting membrane potential of cells favors the inward movement of cations. Kavanaugh (1993) used voltage-clamp experiments to determine the voltage-and concentration-dependence of arginine flux mediated by the mCAT-1 protein expressed in Xenopus oocytes. The voltage-sensitive behavior of the transporter was fitted with the iso-uni-uni kinetic model of facilitated transport (White and Christensen, 1982). In voltage-ramp experiments, arginine flux was examined. The I_max for arginine influx increased with hyperpolarization whereas the K_m value increased with depolarization. The data support the idea that the arginine binding site is located within the membrane electric field, resulting in voltage-dependent binding and unbinding. This model requires movement of negative charge from the inside to the outside during the transition of the unliganded transporter for the zero-trans influx cycle (Kavanaugh, 1993).

The GLUT (glucose transport) proteins are encoded by six functional genes that mediate Na⁺-independent uniport of glucose and fructose (for reviews, see Bell et al. 1993; Mueckler, 1994). For these hexoses, and zwitterionic amino acids that lack a net charge, their Na⁺-independent transport is independent of membrane potential and is driven only by substrate concentration gradients. The GLUT proteins mediate the exchange of glucose between the blood and cytoplasm of cells in this manner. The direction of glucose transport is highly regulated and depends on the cell type and the metabolic status of the cell. However, the restricted expression of glucose-6-phosphatase endows only a subset of mammalian cells with the capacity to dephosphorylate, and hence export, glucose (e.g. liver and muscle, for a review, see Mueckler, 1994). Glucose transport is also mediated by a Na⁺-dependent process that utilizes the electrochemical gradient to drive hexose influx (Wright et al. 1994).

Facilitated transport is bidirectional, and the direction of transport depends on several factors. In the case of cationic amino acids, cellular membrane potential strongly favors influx. Similarly, zwitterionic amino acid and glucose influx is favored when co-transported with Na⁺. The asymmetrical recognition of substrate by a transporter may influence the preferred direction of transport. This property is exhibited by the y⁺ transport system with the analog GPA (4-amino-1-guanylpiperidine-4-carboxlyate). GPA is recognized, transported and capable of lysine trans-stimulation only when present on the extracellular face of the membrane (White et al. 1982; White and Christensen, 1982). Trans-stimulation is defined by an increase in transport when trans-side substrate concentration is elevated. Trans-stimulation by arginine was observed in mCAT-expressing Xenopus oocytes (Closs et al. 1993). Similarly, glucose transporters interact asymmetrically with ligands such as cytochalasin B (an inhibitor of hexose transport) and IAPS-forskolin (Wadzinski et al. 1988) or ATB-BMPA (Holman et al. 1990). Perhaps asymmetrical substrate recognition is a general property of transport proteins.

In addition to their constituent role in proteins, amino acids function as neurotransmitters, sources of metabolic energy and precursors to many biologically important molecules (Fig. 2). Nutrient homeostasis results from complex trafficking between organs and involves specific transport proteins (Christensen, 1982, 1990). Several different transporters mediate the flux of the cationic amino acids arginine, ornithine and lysine (Table 1). Although arginine is not an essential amino acid, only liver and kidney cells produce useful amounts. The kidney is the major arginine source via synthesis from citrulline (Barbul, 1990), whereas the high arginase activity in liver prevents its export (Herzfeld and Raper, 1976). All other cells import arginine from plasma for their metabolic requirements. Arginine is the immediate precursor in the synthesis of nitric oxide (NO), a short-lived metabolite involved in vasodilatation, neurotransmission, tumor immunity and non-specific host defense to foreign antigens (for a review, see Moncada and Higgs, 1993). Arginine is also the precursor of creatine, which functions as a high-energy phosphate carrier during muscle contraction. Ornithine is the precursor of polyamines (putrescine, spermidine and spermine), which are present in all cells and participate in cellular division and proliferation (Janne et al. 1991).

In mammals, lysine in an essential amino acid required for protein synthesis and is a precursor for trimethyllysine and carnitine synthesis, which are essential for the translocation of long-chain fatty acids into the mitochondrial matrix for β-oxidation. Since fatty acid oxidation is the major energy source used by cardiac and skeletal muscle, deficiencies in carnitine are associated with cardiac, liver and kidney disease (Borum, 1983; Bieber, 1988; Nyhan, 1988).

Cystine is not an essential amino acid. Nevertheless, the control of cystine distribution is interesting in that this amino acid is toxic to many cellular enzyme systems. Two cycles interchange cystine and cysteine within the cytoplasmic and lysosomal compartments and involve specific pH-dependent transport systems in lysosomes. Cystine transported into cells is used in protein or glutathione synthesis or is rapidly reduced to cysteine. The cysteine is transported into lysosomes, where it may react with disulfide bridges of proteins undergoing proteolysis. Cystine is exported from the lysosome back to the cytosol to complete the cycle. Glutathione is an essential antioxidant that plays a key role in oxidative stress responses (for a review, see Pisoni and Schneider, 1992). There are two common hereditary defects in cystine transport, Fanconi syndrome and hereditary cystinuria. Strong evidence for the role of rBAT in the latter condition is presented by Palacín (1994).

Amino acid reabsorption by the kidney is so efficient that more than 99% is recovered in the glomerular filtrate. Heritable defects in intestinal and renal amino acid transport have been detected. For example, lysinuric protein intolerance appears to be a transport dysfunction in the basolateral excretion of lysine from kidney tubule cells (Rajantie et al. 1981) and intestinal mucosa (Rajantie et al. 1980). Patients with cystinuria hyper-excrete cystine, arginine, lysine and ornithine in their urine (Crawhall et al. 1967; Whelan and Scriver, 1968; Simell, 1989). Based on rBAT (related B system αmino acid transporter) expression in the microvilli of proximal straight tubules (Furriols et al. 1993), and the observation that anti-sense rBAT RNA reduced cystine uptake in Xenopus oocytes (Bertran et al. 1993), studies were undertaken to examine the rBAT gene in patients with cystinuria. Recent work from Palacín’s group revealed one common and several other isolated missense mutations in the rBAT gene of individuals with heritable forms of cystinuria (Calonge et al. 1994; Palacín, 1994).

The role of glucose and the regulation of its transport in metabolism and homeostasis are better understood than the more complex problem of transporting the wide variety of amino acids. The study of glucose transport benefited from the early cloning of the GLUT family of uniporters and the intense research effort directed to their analysis (over 1000 papers in 5 years; for a review, see Mueckler, 1994). Glucose is required by many cells for oxidative and non-oxidative ATP production. In mammalian cells, both a Na⁺-dependent cotransporter (Hediger et al. 1989; Wright et al. 1994) and at least six functional Na⁺-independent uniporters regulate glucose fluxes.

The appropriate distribution of whole-body glucose appears to be controlled by tissue-specific expression and regulation of several transporter isoforms with distinct kinetic properties that together play a major role in glucose homeostasis. Expression of GLUT1, GLUT2 and GLUT4 is regulated by numerous endogenous and xenobiotic factors. GLUT 4 regulation, the most intricate, illustrative and controversial of the GLUT isoforms, is discussed by Mueckler (1994).

GLUT4 is expressed only in adipocytes and muscle cells, both of which are highly ‘insulin-sensitive’ and respond to insulin with a rapid and reversible increase in glucose transport. The protein is sequestered intracellularly and, upon stimulation by insulin, it is recruited to the cell surface. This recruitment results in immediately increased glucose transport and occurs in the absence of de novo transcription or translation (for reviews, see Birnbaum, 1992; Bell et al. 1993). Skeletal muscle glucose transport accounts for 20% of basal whole-body glucose uptake and for 75–95% during hyper-insulinemia (Baron et al. 1988). GLUT4 regulation in these insulin-sensitive tissues is important for glucose homeostasis (Mueckler, 1994).

The intracellular sequestration of GLUT4 proteins presumably results from the presence of target sequences within the protein. Chimeric constructs formed between GLUT1 and GLUT4 were tested in several systems in combination with deletion mutations to identify the regions responsible for intracellular GLUT4 targeting. Results so far are directly conflicting. For example, Piper et al. (1992, 1993) have reported that the amino terminus of GLUT4 was necessary and sufficient for intracellular sequestration, whereas other groups report that the intracellular targeting sequence is within the carboxyl terminus (Verhey et al. 1993; Czech et al. 1993). To complicate matters further, data from Asano et al. (1992) support the idea that two internal domains are responsible for sequestering GLUT4 in the trans-Golgi. Unfortunately, no studies were performed in adipocyte or muscle cells; the investigations were carried out in Chinese hamster ovary, COS, NIH 3T3 cells or Xenopus oocytes. It is possible that tissue-specific factors involved in intracellular trafficking of GLUT4 may vary by cell type and modify the behavior of GLUT 4 in a cell-specific manner.

The study of amino acid transporter gene regulation is just beginning. The mCAT-2 gene encodes two distinct proteins that differ in their apparent K_m for arginine by greater than 10-fold (Table 2). These isoforms arise from alternate transcript splicing that involves a 40 amino acid internal domain segment. Additional variability in expression was revealed when mCAT-2 transcripts were found to contain at least four different 5′ untranslated regions (UTRs) which arise from at least three distinct promoters (Finley, 1993; K. Finley, A. Barrieux, J. Kleeman, P. Huynh and C. L. MacLeod, in preparation; for a review, see MacLeod et al. 1994). Hence, the mCAT-2 gene gives rise to two protein isoforms with different kinetic properties and the multiple promoters provide for specific gene regulation with the economy of a single gene. Such features may confer a similar degree of transport flexibility on mCAT-2-mediated amino acid transport as that provided by six different GLUT gene products on glucose transport.

The coregulation of mCAT-2 and 4F2hc occurs in resting and activated lymphocytes and may prove to be useful in examining their possible functional interaction. Resting lymphocytes express mCAT-1 mRNA constitutively, yet they exhibit little arginine transport. Upon activation, T-cells show an absolute requirement for arginine that correlates with mCAT-2 and 4F2hc gene induction and observed increases in arginine transport (Boyd and Crawford, 1992; Parmacek et al. 1989; Devés et al. 1992; Kilberg and Häussinger, 1992). Transcriptional and post-transcriptional regulation of the mCAT-2 gene has been observed in lymphoma cells that differ in mCAT-2 expression and in somatic cell hybrids formed among them (Wilkinson et al. 1991). The 4F2hc gene has been isolated; its structure, sequence and expression have been used to define the promoter and other regulatory elements (Karpinski et al. 1989). The regulation of the rBAT gene has not been investigated.

Precious little is known about the structure of vertebrate transporters in biomembranes. Major hurdles must be overcome before details of structure can be obtained. It is unfortunate that their hydrophobic properties, relatively large size and low abundance pose experimental difficulties that are confounded by the need to prepare large quantities of purified protein from expression systems such as bacteria or baculovirus. So far, these problems have prevented the successful resolution of their three-dimensional structure. Nevertheless, their similarity to the better-studied bacterial and yeast membrane proteins (Reizer et al. 1993), bacteriorhodopsin and porins has considerable value in approaching this problem (for reviews, see Saier, 1994; Cowan and Rosenbusch, 1994). As the isolation of more transporters proceeds, family relationships among the proteins can be more firmly established and structural information on one family member may guide further analysis on other members. Family relationships may reveal common topological motifs among its members (for reviews, see Griffith et al. 1992; Saier, 1994).

The lack of precise information on the three-dimensional structure of transporters has forced researchers to rely heavily on computer modeling using algorithms. These aids are useful as a first approximation and can identify hydrophobic domains, putative α-helical regions and charged residues. But it is clear that computer programs cannot accurately predict membrane-spanning domains since some α-helical hydrophobic domains of greater than 18 residues are not embedded in membranes and the possible presence of β-barrels, for example, is often overlooked. Most researchers use computer predictions to make testable models to locate cytoplasmic and surface domains and to identify key residues for binding and transport.

Because positively charged amino acids in membrane proteins are often found on the intracellular face (von Heijne, 1992), their presence can assist in inside versus outside orientation modeling. An innovative strategy for determining the orientation of membrane proteins involves alkaline phosphatase as a reporter. Fusion constructs of alkaline phosphatase with the bacterial melibiose carrier exhibit enzymatic activity only when orientated towards the periplasm and not towards the cytoplasmic side of the bacterial membrane (Botfield et al. 1992; Manoil and Beckwith, 1985; Michaelis et al. 1983; Hoffman and Wright, 1985). Information on the tranporter orientation in membranes might be obtained by the placement of alkaline phosphatase coding sequences within sequences encoding various domains in the test protein (if their behavior in bacteria reflects their natural in vivo orientation). Site-directed mutagenesis has been profitably and extensively used to identify key residues in substrate recognition and/or transport functions (Kaback, 1994; Wright et al. 1994). Chimeric proteins can identify regions of substrate interaction, regulatory domains and binding sites. Protease treatment of membrane proteins has revealed sites protected by the substrate GABA (γ-aminobutyric acid) on the GABA₁ transporter (Kanner, 1994). The available information on the topology of GLUT transporters has been reviewed by Mueckler (1994).

Less is known about the structure of vertebrate amino acid facilitated transporters. Domain-swapping experiments among mCAT transporters located the mCAT-1 retrovirus binding site (Albritton et al. 1993). Hydrophobicity profiles and computer algorithms predict the mCAT proteins to contain 12–14 membrane-spanning domains (Closs et al. 1993; Reizer et al. 1993). Their nearly identical hydrophobicity profiles suggest they have similar structures.

The predicted single transmembrane structure of the type II glycoprotein rBAT family and 4F2hc suggests that they may be modulators rather than direct mediators of amino acid transport (see Palacín, 1994). Physical evidence regarding their structure is being sought, but strong evidence supports their important role in the regulation of amino acid transport. What is known of the unexpected structure of these proteins was recently reviewed (Hediger et al. 1993; Bertran et al. 1994; Palacín, 1994). Their features are compared with those of the mCAT transporters below.

The mCAT cDNAs were cloned serendipitously and their natural function was initially unknown (for a review, see MacLeod et al. 1994). Two known genes (mCAT-1 and mCAT-2) constitute this family and encode three proteins which mediate the transport of cationic and dipolar amino acids when expressed in Xenopus oocytes (MacLeod et al. 1990; Wang et al. 1991; Kakuda et al. 1993; Closs et al. 1993). The mCAT-2 gene encodes both the high-affinity mCAT-2 protein and the low-affinity high-capacity mCAT-2a isoform, as mentioned previously. The different K_m values of substrate exhibited by the isoforms are largely conferred by the alternately spliced exons, as shown by domain-swapping experiments (Closs et al. 1993). Some of their properties are summarized in Table 2 (see legend to Table 2, describing the confusion in the designations for these isoforms).

The second type of protein is represented by rBAT and 4F2hc. Here, the term rBAT is used to refer to the sequence and protein product variously named NAA-Tr (NBAT), D2 (Bertran et al. 1992a,b; 1994; Wells and Hediger, 1992; Tate et al. 1992; Hediger et al. 1993; Markovich et al. 1993; Mosckovitz et al. 1993). A related cDNA clone, 4F2hc, initially isolated as a T-cell marker (Parmacek et al. 1989), was later found also to mediate amino acid transport. Both rBAT and 4F2hc genes encode type II membrane glycoproteins related to α-amylases and α-glucosidases (for reviews, see Hediger et al. 1993; Palacín, 1994). The rBAT protein, when expressed in Xenopus oocytes, exhibits properties similar to transport system b^o,+ (Table 1; Van Winkle et al. 1988). This transport system mediates the Na⁺-independent movement of both cationic and dipolar amino acids. The rBAT protein also mediates the influx of cystine, an amino acid not associated with b^o,+ transport (Table 1). The 4F2hc protein expressed in oocytes mediates transport properties similar to those of system y⁺L (Devéz et al. 1992; Table 1).

A possible association of mCAT and 4F2hc function may occur in activated lymphocytes. Quiescent lymphocytes express mCAT-1 but not mCAT-2 or 4F2hc and exhibit minimal lysine transport via systems y⁺ and y⁺L (Boyd and Crawford, 1992; Devés et al. 1992). After T-cells have been activated, the expression of mCAT-2/2a and 4F2hc are induced (Parmacek et al. 1989; MacLeod et al. 1990; Kakuda et al. 1993; Finley, 1993) in a time frame corresponding to an increased y⁺ transport activity in activated T-cells (Boyd and Crawford, 1992). However, co-expression of mCAT proteins with 4F2hc in Xenopus oocytes has not yet revealed any transport synergy (M. Palacín, personal communication).

If 4F2hc and rBAT span the membrane only once, it is unlikely that they function independently in the transport of substrates. Perhaps these genes encode a family of proteins that function to modulate the predominant direction of transport, regulate cell surface localization and modify their substrate specificity in a tissue-specific or physiologically responsive manner; or protein oligomers may alter substrate specificity and/or affinity. Sequence and structural evidence of possible protein–protein interaction exists. The rBAT protein appears to be disulfide-linked to a smaller 125 kDa subunit (Palacín, 1994); it also has a leucine zipper motif which could mediate dimerization. The 4F2hc protein is complexed with a light chain, which has not been isolated.

Recent immunohistochemical analyses of fibroblasts, endothelial cells and hepatoma cells have revealed mCAT-1 protein clustering associated with the membrane cytoskeleton (Woodard et al. 1994). The clustering of other membrane-associated proteins has been reported, including that of the N-methyl-D-aspartate receptor (Benke et al. 1993). The physiological significance of this clustering remains to be determined, but the association of several proteins into a complex might alter the substrate specificity and/or affinity and could thereby regulate substrate transport, depending on the availability of the subunits.

The rBAT and 4F2hc proteins contain conserved residues similar to the α-amylase and α-glycosidase sequences proposed as sites critical for calcium binding and catalytic activity. The rBAT and 4F2hc proteins do not exhibit detectable enzymatic activity (Wells and Hediger, 1992). Recently, an unrelated single membrane-spanning protein, aminopeptidase-N, was found to be associated with the Na⁺-dependent transport of zwitterionic amino acids alanine, glutamine, leucine and phenlyalanine (Plakidou-Dymock et al. 1993). Aminopeptidase-N is both an enzyme and a mediator of amino acid transport. It has been postulated that, like rBAT and 4F2hc, the peptidase is associated with an unidentified transport protein. It is possible that the transport and enzymatic activities are not coincidental, but rather that the enzyme digestion products are the transported substrates, since aminopeptidase-N functions preferentially as a dipolar peptidase, which cleaves alanine, leucine and other bulky hydrophobic substrates (Kenny et al. 1987). The protein sequence similarity among rBAT, 4F2hc, α.-amylase and α.-glycosidase is intriguing in the light of the recent information that aminopeptidase-N is an active enzyme that mediates amino acid transport. Perhaps multiple membrane-spanning proteins form pathways for substrate flux whose substrate specificity and/or kinetic properties can be modulated in a tissue-specific manner by other less hydrophobic membrane glycoproteins responsive to cell signaling systems. This is an area of active investigation (see, for example, Palacín, 1994).

The recent advances in mammalian amino acid transporter analysis have relied on their forced expression in Xenopus oocytes, since they have low levels of endogenous amino acid and glucose transport. Expression cloning in Xenopus oocytes was used to identify mRNA fractions responsible for specific transport functions and for isolating first the glucose transporters (for a review, see Wright et al. 1994) and later for those encoding amino acid transporters (Palacín, 1994). The Xenopus system was also used to identify the function of the mCAT genes for which the natural function of the isolated cDNAs was being sought (Kim et al. 1991; Wang et al. 1991; Kakuda et al. 1993).

Although the Xenopus oocyte expression has been invaluable for the identification and characterization of transporters, there are problems in using the system. Two groups have reported the detection of endogenous amino acid transport (Bertran et al. 1992a; Van Winkle, 1993). A second problem is illustrated by studies indicating that the Xenopus oocytes may be limited in their capacity to translate selected mRNAs or to modify their protein products correctly (Snutch, 1988). For example, translational selectivity was noted when only a subset of 5-hydroxytryptamine (5-HT) receptors and voltage-gated Ca²⁺ channels were detected after mRNA from cells expressing a wider set of proteins had been injected into oocytes (Leonard et al. 1987; Lübbert et al. 1987). Failure to modify transporters properly was evident when microinjected eel electroplax mRNA resulted in the production of only precursor forms of Na⁺ channel protein (Thornhill and Levinson, 1987). A potential concern involves the frequent use of mRNA from tissues consisting of several cell types. When such mixtures of RNA are expressed in oocytes, the co-expression of their proteins could reveal non-physiological activities (Snutch, 1988).

Expression cloning or expression analysis may fail in situations were the functional protein unit is a hetero-oligomer and not all of their mRNA precursors are injected or expressed as proteins. Conversely, the protein expressed from injected mRNA might interact with an endogenous Xenopus oocyte protein and modify transport, resulting in transport activity that may not reflect the normal physiological function of this protein (Bertran et al. 1992a,b; Markovich et al. 1993; Tate et al. 1992; Wells and Hediger, 1992). To be confident that the function of the cloned gene product measured in Xenopus oocytes reflects its normal activity, other experimental verification is desirable, although sometimes quite difficult to obtain. Antisense or other tests of specificity have been employed in Xenopus and might be profitably used in somatic cells.

The uniporters discussed here are important for nutrient homeostasis and further research will yield exciting new information on their structure, the mechanisms of their transport and regulation, and their role in health and disease.

We are grateful for helpful advice from Dr H. N. Christensen and thank Drs M. Palacín and M. Pastor-Anglada for stimulating discussions. C.L.M. is a Clayton Foundation Investigator. D.K. is a predoctoral fellow in the Biomedical Sciences Graduate program supported in part by USPH GM 07752-16 and the American Cancer Society IM653 (C.L.M.).