|Home | About | Journals | Submit | Contact Us | Français|
The Bcl-2-related protein Bax is toxic when expressed either in yeast or in mammalian cells. Although the mechanism of this toxicity is unknown, it appears to be similar in both cell types and dependent on the localization of Bax to the outer mitochondrial membrane. To investigate the role of mitochondrial respiration in Bax-mediated toxicity, a series of yeast mutant strains was created, each carrying a disruption in either a component of the mitochondrial electron transport chain, a component of the mitochondrial ATP synthesis machinery, or a protein involved in mitochondrial adenine nucleotide exchange. Bax toxicity was reduced in strains lacking the ability to perform oxidative phosphorylation. In contrast, a respiratory-competent strain that lacked the outer mitochondrial membrane Por1 protein showed increased sensitivity to Bax expression. Deficiencies in other mitochondrial proteins did not affect Bax toxicity as long as the ability to perform oxidative phosphorylation was maintained. Characterization of Bax-induced toxicity in wild-type yeast demonstrated a growth inhibition that preceded cell death. This growth inhibition was associated with a decreased ability to carry out oxidative phosphorylation following Bax induction. Furthermore, cells recovered following Bax-induced growth arrest were enriched for a petite phenotype and were no longer able to grow on a nonfermentable carbon source. These results suggest that Bax expression leads to an impairment of mitochondrial respiration, inducing toxicity in cells dependent on oxidative phosphorylation for survival. Furthermore, Bax toxicity is enhanced in yeast deficient in the ability to exchange metabolites across the outer mitochondrial membrane.
Apoptosis, or programmed cell death, is an evolutionarily conserved mechanism by which multicellular organisms regulate cell numbers. The proper regulation of this process is critical during cell accumulation in development as well as during tissue homeostasis in adult organisms. The onset of apoptosis in an individual cell can be triggered by environmental cues, including specific signaling cascades or an insufficiency of survival factors (28).
The Bcl-2 family of proteins includes important regulators of the cellular decision to undergo apoptosis, yet the mechanisms by which these proteins act remain controversial. The Bcl-2 family includes both pro- and antiapoptotic members that appear to be capable of controlling the initiation of apoptosis via actions at the mitochondria. The antiapoptotic family members have been shown to form pores in lipid bilayers and may work by regulating the integrity of the outer mitochondrial membrane, possibly through regulation of ion homeostasis or through regulation of the permeability transition pore (9). Proapoptotic Bcl-2 proteins promote cell death in at least two different ways, depending on their structural motifs. Bcl-2 family proteins are defined by their Bcl-2 homology domains: BH1, BH2, BH3, and BH4. Some proapoptotic proteins contain only a BH3 domain (13). Structural studies have shown that the BH3 domain of BH3-only proteins can interact with the hydrophobic cleft formed by the BH1, BH2, and BH3 domains of antiapoptotic Bcl-2 proteins (24). In vivo, this interaction appears to inhibit the prosurvival activity of Bcl-2 and Bcl-xL. Other proapoptotic family members, such as Bax and Bak, contain BH1 and BH2 domains in addition to a BH3 domain, are more homologous to Bcl-2, and may also be able to form channels (25). These proteins appear to be capable of promoting death in the absence of interactions with other Bcl-2 family members. The mechanism of this proapoptotic activity remains unknown, although theories include the physical destabilization of the outer mitochondrial membrane (1), the formation of ion-conducting channels that dissipate electrochemical gradients necessary to maintain mitochondrial homeostasis (23), participation in the permeability transition pore (16, 20, 30), and elevation of reactive oxygen species levels generated by mitochondria (14).
Some of the evidence that proapoptotic Bcl-2 proteins can act independently comes from experiments done with yeast. It has been shown that Bax expression is toxic to both Saccharomyces cerevisiae and Schizosaccharomyces pombe and that this toxicity can be rescued by coexpression of Bcl-2 or Bcl-xL, even when mutations prevent physical interactions between the pro- and antiapoptotic proteins (8, 10, 12, 18, 31). These observations have led to the proposal that Bax functions similarly in yeast and mammalian cells. Further support for this idea comes from experiments showing that Bax affects mitochondrial physiology in yeast in the same ways it affects mitochondrial physiology in mammalian cells, including alterations in mitochondrial membrane potential (18) and the release of cytochrome c from the intermembrane space (15, 17). The genome of S. cerevisiae has been completely sequenced, and no homologues to Bcl-2 have been identified in this yeast, nor are there any proteins with homology to caspases or Ced-4. Similarly, it has been demonstrated that cytochrome c release is not required for Bax-induced lethality in yeast, probably reflecting the lack of an apoptotic cascade (22). Therefore, Bax toxicity in yeast appears to be due to an activity other than the inhibition of prosurvival Bcl-2 proteins or the activation of the metazoan apoptotic machinery.
The study of Bax toxicity in S. cerevisiae is attractive because of the ease with which the genome of this yeast can be manipulated. Using genetic screens and yeast knockout technology, previous studies of S. cerevisiae have suggested that several genes encoding proteins localized to mitochondria may be required for Bax toxicity. These genes include ATP4, encoding a component of the F1F0 ATPase (17); the adenine nucleotide translocator (ANT) genes (16); and POR1 (26), which codes for the predominant form of the yeast voltage-dependent anion channel (VDAC). The proteins encoded by these genes share a common involvement in cellular respiration. Strains lacking ATP4 or the ANT are deficient in the ability to perform oxidative phosphorylation and therefore derive their ATP from fermentation (6, 21). Strains lacking POR1 are able to perform oxidative phosphorylation, but they do so more slowly than wild-type yeast, potentially as a result of impaired exchange of metabolites and adenine nucleotides across the outer mitochondrial membrane (2, 5). While these data support the notion that mitochondria are critically involved in Bax toxicity, conflicting data have been published regarding the effect of Bax on yeast deficient in the ability to carry out oxidative phosphorylation, thus casting doubt on the role of cellular respiration in Bax toxicity (8, 17, 22).
In order to address the role of cellular respiration in Bax-mediated yeast toxicity, we made a panel of strains with single-gene deletions, targeting each complex of the electron transport chain as well as components of the mitochondrial ATP synthesis machinery. We show here that the ability to respire is an important determinant of Bax toxicity in yeast. Strains that were able to respire, particularly VDAC-deficient strains, showed a much greater growth defect in response to Bax expression than did respiration-incompetent strains. In addition, yeast cells that survived Bax exposure frequently displayed a permanent loss of respiration competence. Populations of yeast expressing Bax showed increased ethanol accumulation, increased sensitivity to ethanol, and diminished oxygen consumption, as would be predicted if Bax were interfering with cellular respiration. These data suggest that Bax expression can impair the ability of yeast to carry out oxidative phosphorylation and that VDAC acts either to limit the ability of Bax to impair oxidative phosphorylation or to promote the ability of cells to survive under conditions where oxidative phosphorylation is limited.
Yeast studies were performed with S. cerevisiae W303 (ade2-1 can1-100 his3-11,15 leu2-3,112 trp1-1 ura3-1). The yeast cells were maintained on YPD medium (2% yeast extract, 1% Bacto Peptone, 2% glucose) (U.S. Biological, Swampscott, Mass.) or on selective synthetic complete medium (1.5 g of U.S. Biological yeast nitrogen base, 5 g of ammonium sulfate, and 2 g of amino acid powder mix lacking leucine per liter with 2% glucose) as appropriate. Induction was done using selective synthetic complete medium with 2% galactose and 2% raffinose replacing the 2% glucose. Gene deletion strains were created by replacing the coding sequence in question with the kanMX4 module, which confers resistance to G418 (29). The yeast cells were transformed with PCR products consisting of the kanMX4 module flanked by 50 bp of upstream and downstream sequence (7). Knockouts were selected for growth on G418-containing plates (200 mg per liter) and confirmed by PCR. The por1Δ por2Δ double-deletion strain was created by mating the strains carrying a single deletion followed by sporulation and dissection to recover the double knockout. The human Bax cDNA was cloned into the multicopy expression plasmid p425 GALL (leucine selection) under the control of a galactose-inducible promoter (19). Yeast cells were transformed with the Bax-p425 GALL expression plasmid or the empty plasmid by the lithium acetate method and selected on leucine-deficient plates (7).
Cultures were grown for 24 h in liquid, selective, noninducing medium (2% glucose). At 24 h, the cultures were spun down and resuspended in liquid, selective, inducing (2% galactose, 2% raffinose) medium. After 12 h, the cells were harvested and the RNA was extracted using TRIzol reagent (Life Technologies, Grand Island, N.Y.). The extracted RNA was treated with DNase I amplification grade (Life Technologies) and quantitated by spectrophotometry, and 2 μg of RNA from each strain of yeast was reverse transcribed with Moloney murine leukemia virus reverse transcriptase (Life Technologies). One microliter of each reverse transcription (RT) reaction mixture was used for a PCR of 27 cycles.
Yeast cells grown for 24 to 48 h on solid, selective, noninducing medium were resuspended in liquid, selective, inducing medium and normalized to the same optical density at 650 nm (OD650). Five 10-fold serial dilutions were made of each strain in 200 μl of liquid, selective, inducing medium in 96-well plates. The plates were cultured at 30°C under continuous agitation for 60 h. Every 12 h, the cells were resuspended and the OD650 was measured. For the spot tests, 4 μl of the serial dilutions was spotted onto solid, selective, inducing or solid, selective, noninducing medium and incubated for 3 to 4 days at 30°C.
Yeasts were grown overnight in liquid, selective, noninducing medium; washed three times in phosphate-buffered saline (PBS), resuspended in liquid, selective, inducing medium; and cultured for 4 to 8 h at 30°C. The cultures were then normalized to an OD600 of 0.06 in 400 μl of fresh liquid, selective, inducing medium containing the specified percentage of ethanol. The cultures were incubated overnight at 30°C, and the OD600 was measured after 24 h.
Cultures were grown overnight in liquid, selective, noninducing medium and washed three times in PBS; the density was normalized to an OD600 of 0.006 in 6 ml of fresh liquid, selective, inducing medium, and the cultures were incubated at 30°C for 48 h. The cells were then pelleted at 4°C, and the ethanol concentration of the supernatant was tested using a commercially available enzymatic assay (Sigma, St. Louis, Mo.). Briefly, exogenous alcohol dehydrogenase was used to catalyze the reaction between ethanol and exogenous NAD to form acetaldehyde and NADH. The amount of NADH generated was measured by an increase in the OD340; this increase is directly proportional to the concentration of ethanol in the supernatant. The results were normalized to the OD600 of the sample.
Cultures were grown overnight in liquid, selective, noninducing medium and washed three times in PBS; the density was normalized to an OD600 of 0.006 in 6 ml of fresh liquid, selective, inducing medium, and the cultures were incubated at 30°C for 48 h. The density of yeast was normalized to an OD600 of 0.06. Three hundred microliters of a 1:10,000 dilution was plated onto YPD plates using 4-mm-diameter sterile glass beads. The colonies were counted after 4 to 5 days of growth at 30°C.
The ability of each knockout strain to perform oxidative phosphorylation was checked by plating it onto medium containing the respiratory carbon sources ethanol and glycerol (YPEG; 2% yeast extract, 1% Bacto Peptone, 3% ethanol, 3% glycerol, 1.7% agar) instead of dextrose. The ability of the strains to respire after Bax induction was assessed after 48 h by plating cells from the clonogenicity assays onto YPD medium at a dilution that gave between 100 and 350 colonies per plate, streaking out these colonies onto YPD plates, and then replica plating the colonies to YPD and YPEG media. Colonies that exhibited growth on YPEG medium were considered respiration competent.
Cultures were grown overnight in liquid, selective, noninducing medium and washed three times in PBS; their densities were normalized to an OD600 of 0.006 in liquid, selective, inducing medium, and they were allowed to grow for 48 h. The OD600 was measured, and the rate of oxygen consumption was measured in a respirometer equipped with a calibrated polarographic O2 electrode as described previously (4).
In order to investigate the role of respiration in Bax toxicity in S. cerevisiae, a series of yeast strains carrying single-gene deletions was created by targeting each complex of the electron transport chain, as well as cytochrome c, the F1F0 ATPase, VDAC, and the ANT (Fig. (Fig.1).1). These disruptions used homologous recombination to replace the gene of interest with the heterologous kanMX marker, which confers resistance to G418. The disruptions were confirmed by PCR (data not shown). In addition, [rho0] strains, which lack the components of the electron transport chain and F1F0 ATPase encoded by the mitochondrial genome, were generated by exposure to ethidium bromide. Each strain was tested for the ability to respire by assaying its growth on a nonfermentable carbon source. Yeast cells that cannot respire must derive all of their ATP from fermentation and therefore cannot grow on nonfermentable carbon sources such as ethanol and glycerol. Wild-type yeast and the ndi1Δ (encoding NADH dehydrogenase; complex I), sdh3Δ (encoding succinate dehydrogenase cytochrome b; part of complex II), and por1Δ (encoding the major species of yeast VDAC) knockouts all retained the ability to grow on nonfermentable carbon sources and were therefore respiration competent. The cyc3Δ (encoding cytochrome c heme lyase), qcr7Δ (encoding ubiquinol-cytochrome c oxidoreductase subunit 7; part of complex III), cox4Δ (encoding subunit IV of cytochrome c oxidase; part of complex IV), cox7Δ (encoding subunit VII of cytochrome c oxidase; part of complex IV), atp4Δ (encoding subunit 4 of the F0 sector of the F1F0 ATPase), and pet9Δ (also known as aac2Δ; encoding one of three isoforms of the yeast adenine nucleotide translocator) knockouts, as well as the [rho0] yeast, were unable to grow on nonfermentable carbon sources and were therefore respiration incompetent.
These strains were transformed with either an empty plasmid or a plasmid that expressed Bax using a galactose-inducible promoter. The GALL promoter is repressed in the presence of glucose and is induced by galactose. When inducing Bax, a combination of galactose and raffinose, a sugar that does not induce transcription from the GALL promoter, was used, since the petite, or respiration-incompetent, strains did not grow well on galactose alone. The transformants were induced with galactose and tested for growth on solid medium. Bax induction resulted in significant toxicity in wild-type yeast (Fig. (Fig.2A).2A). Other respiration-competent yeast strains showed Bax-dependent growth impairment that was comparable to that of the wild type. In contrast, the respiration-deficient Bax-transformed strains showed less significant growth defects when compared to the empty-vector control (Fig. (Fig.2A2A and B). The VDAC-deficient strain (por1Δ) was the most sensitive to Bax expression (Fig. (Fig.2C).2C). To confirm that Bax was induced under these conditions, each strain was tested for Bax expression by RT-PCR (Fig. (Fig.2D).2D). As expected, each strain expressed Bax when grown in inducing medium. The minor differences in Bax expression between strains did not correlate with differences in phenotype, suggesting that differences in expression levels between strains do not explain the differences in sensitivity to Bax.
In order to better quantify these results, Bax-mediated toxicity was also assayed in liquid medium. Yeast growth was followed by measuring the OD650 every 12 h for a period of 60 h. While the growth impairment observed in the respiration-competent strains was comparable to that of the wild type (Fig. (Fig.3A),3A), this toxicity was reduced compared to that of the wild type in the strains that could not perform oxidative phosphorylation (Fig. (Fig.3A3A and B). Again, the only strain in which Bax toxicity was more severe than in the wild type was the VDAC-deficient strain (Fig. (Fig.3C).3C). Bax toxicity could be observed in wild-type yeast and other respiration-competent strains as early as 24 h, and differences persisted throughout the subsequent growth phase. All the strains tested were capable of some growth in the presence of Bax; in none of the strains was Bax expression completely lethal.
In S. cerevisiae, there are two isoforms of VDAC: Por1 and Por2. POR2 is 49% identical to POR1 and is a multicopy suppressor of the temperature-sensitive respiratory growth defect of por1Δ yeast (2). To more completely assess the role of VDAC in Bax toxicity, a strain lacking both POR1 and POR2 was generated. As previously described (2), this strain grew slowly on glucose at 30°C and grew much more slowly than either single knockout on ethanol and glycerol at 30°C, demonstrating a reduced but still measurable ability to respire (data not shown). (Bax toxicity in por2Δ yeast was comparable to that in the wild type.) Bax induction in por1Δ por2Δ yeast results in a clear impairment in growth (Fig. (Fig.4).4). These results suggest that interaction with VDAC is not required for Bax toxicity in yeast. The data also suggest that, in wild-type yeast, VDAC functions may interfere with Bax activity or mitigate its harmful effects.
From the growth curves in Fig. Fig.3,3, it is clear that the strains expressing Bax are still able to grow, albeit more slowly than the strain containing the control plasmid. In order to assess whether Bax was actually killing the yeast, clonogenicity assays were performed to assess the ability of cells to recover from Bax exposure. Following 48 h in galactose-containing medium, wild-type and por1Δ cells were transferred to YPD (nonselective glucose-based) plates. Since the plates contained glucose, Bax expression would be repressed in any yeast that kept the plasmid. Alternatively, viable yeast might recover through loss of the Bax plasmid. The Bax-exposed culture yielded 31% ± 6% fewer colonies than did the wild-type control culture, and the Bax-exposed por1Δ culture yielded 40% ± 4% fewer colonies than did the por1Δ culture. Thus, the majority of yeasts were able to survive Bax expression even though significant impairment of growth was observed (Fig. (Fig.5A).5A). Surprisingly, many of the colonies that grew from the Bax-exposed populations were significantly smaller than other colonies on the same plate. Small colony size in yeast is frequently an indicator that the yeast can no longer perform oxidative phosphorylation (petite phenotype).
To determine whether these cells had lost the ability to respire, the colonies that grew out of a clonogenicity assay were streaked out onto YPD medium and then replica plated onto both YPD and YPEG (nonselective ethanol- and glycerol-based) plates. None of the small colonies were able to grow on YPEG plates, whereas all of the large colonies were able to do so, indicating that the small colonies were in fact respiration incompetent (data not shown). This effect was quantified by plating wild-type and por1Δ cells from 48-h, galactose-induced Bax and control cultures onto YPD plates and comparing the numbers of petite colonies. A total of 31% ± 3% of wild-type and 29% ± 2% of VDAC-deficient cells exposed to Bax had become petite, while only a small number of control cells became petite (Fig. (Fig.55B).
Yeast can derive cellular ATP from two different mechanisms, respiration and fermentation. In fermentation, the pyruvate generated by glycolysis is converted into ethanol instead of being transported into the mitochondria and used to generate NADH in the citric acid cycle. In general, S. cerevisiae growing on glucose will initially ferment the available sugar and will switch to respiration after the sugar is fermented (the diauxic shift) (11). Bax showed less toxicity towards respiration-incompetent strains than it did towards respiration-competent strains. Furthermore, Bax exposure appeared to select for loss of respiration competence. Thus, it seemed likely that Bax induction might be forcing the yeast to become more reliant on fermentation for energy. To investigate this possibility, normalized cultures of Bax-expressing and control cells were grown in liquid culture and their ethanol levels were measured after 48 h of growth. At this time point, both wild-type and por1Δ Bax-expressing cells demonstrated increased ethanol accumulation compared to control cells (Fig. (Fig.6A).6A). This could be due either to an increased production of ethanol (increased glycolysis) or to a decrease in the metabolism of ethanol due to a decrease in respiration (decreased oxidative phosphorylation) or both.
The addition of ethanol to yeast cultures is known to limit the rate of yeast fermentation (3). To determine whether Bax-expressing yeasts are more reliant on fermentation for ATP synthesis than control yeast, ethanol was added to normalized cultures of Bax-expressing and control cells and growth was measured after 24 h. The difference in growth between Bax-expressing and control cells became progressively greater as an increasing amount of ethanol was added to the cultures (Fig. (Fig.6B),6B), suggesting that impairing the ability to derive ATP from fermentation is more detrimental to the growth of Bax-expressing yeast than to the growth of control yeast. At concentrations of ethanol of ≥9%, the growth of both Bax-expressing and control strains was completely inhibited (data not shown). Both the increased accumulation of ethanol in Bax-expressing cultures and the increased sensitivity of Bax-expressing cultures to growth inhibition in the presence of ethanol suggest that Bax may be interfering with the ability of the cell to derive energy through oxidative phosphorylation.
Finally, to determine whether respiration was decreased in Bax-expressing yeast, normalized cultures of Bax-expressing and control cells were grown for 48 h and the oxygen consumption of the cultures was measured. The number of yeast cells was controlled for by normalizing to the OD600. Bax-expressing yeast consumed less oxygen as a population than did control yeast (Fig. (Fig.66C).
Although the localization of Bax to mitochondria has been shown to be important for the ability of Bax to induce cell death, the mechanism by which Bax exerts its toxicity is poorly understood. Here we show that Bax is selectively toxic to cells dependent on oxidative phosphorylation and that it can interfere with the ability of cells to perform oxidative phosphorylation. Previous studies have suggested that specific deficiencies in yeast mitochondrial transport, for example, in the F1F0 ATPase or the ANT, abrogate Bax toxicity (16, 17). Our data suggest that any mutation in the electron transport chain or ATP synthesis machinery that yields respiration-incompetent yeast results in resistance to Bax toxicity, whereas strains with mutations in this machinery that permit continued respiration remain sensitive to Bax. When Bax is expressed in respiration-competent yeast, the cells show an increase in ethanol production, increased sensitivity to ethanol, and decreased oxygen consumption, all of which are suggestive of an increased dependence on glycolysis and a reduction in oxidative phosphorylation. These metabolic changes are accompanied by growth arrest prior to the onset of cell death. Consistent with the hypothesis that Bax toxicity is selective for cells dependent on oxidative phosphorylation, cells that recover from growth arrest following Bax expression frequently exhibit a permanent loss of respiration competence.
The above-mentioned data suggest that Bax toxicity may be a result of a Bax-induced impairment in the ability to perform oxidative phosphorylation. Consistent with this hypothesis, it has been reported that Bax-induced lethality is diminished under conditions that favor fermentation (22). Oxidative phosphorylation takes place across the inner mitochondrial membrane, and the mitochondrial proteins known to influence the rate of oxidative phosphorylation localize to the mitochondrial matrix or the inner membrane. However, since Bax has been shown to localize to the outer mitochondrial membrane, it is not readily apparent how it might exert an effect on oxidative phosphorylation. Although respiratory metabolites and adenine nucleotides must cross the outer mitochondrial membrane, current evidence suggests that this is a diffusion-regulated process. The major channel through which metabolites such as ATP and ADP cross the outer membrane is VDAC (also referred to as Por1 in yeast). Deletion of VDAC leads to a limitation in the rate at which yeast can carry out oxidative phosphorylation. However, other transport mechanisms must exist, as yeasts deficient in both POR1 and its homologue POR2 are viable and capable of carrying out oxidative phosphorylation (2).
In light of these observations, it is interesting that VDAC-deficient mutants display enhanced sensitivity to Bax toxicity. Like wild-type yeast, VDAC-deficient cells accumulate increased quantities of ethanol in response to Bax induction, and Bax induction selects for respiration-deficient mutants. This suggests that Bax toxicity results in further impairment of the ability of VDAC-deficient cells to carry out oxidative phosphorylation. As in wild-type cells, Bax-induced toxicity in por1Δ por2Δ yeast can be prevented by coexpression of Bcl-xL (M. Vander Heiden and C. Thompson, unpublished observation). Recently, both Bax and Bcl-xL have been reported to interact with VDAC (26). Our results, however, suggest that neither the ability of Bax to induce cell death nor the ability of Bcl-xL to prevent Bax toxicity is dependent on an interaction between these proteins and VDAC. Instead, our data are more consistent with a model in which Bax expression limits the diffusion of one or more compounds required for mitochondria to carry out oxidative phosphorylation, and this limitation has greater consequences in the absence of the major outer membrane channel VDAC. Based on our data, if Bax acts at the outer mitochondrial membrane, it is more likely to impair the diffusion of substrates required for coupled respiration across the outer mitochondrial membrane via a mechanism that can be VDAC independent.
Together, the data suggest that Bax toxicity results from a perturbation in the physiology of the organelle to which it localizes. The ability to couple mitochondrial respiration to the energy requirements of the cell involves shuttling of the substrates and products of oxidative phosphorylation across the mitochondrial membranes. Recent evidence from mammalian cells has suggested that Bcl-2 proteins play a role in regulating adenine nucleotide transport across the mitochondrial membranes and that this activity can act to regulate ADP-coupled respiration (27). The observation that Bax expression in yeast leads to alterations in the regulation of oxidative phosphorylation provides further evidence that Bcl-2 proteins play a role in regulating mitochondrial physiology even in the absence of other components of the metazoan apoptotic pathway, including caspases and Ced-4 homologues.
We thank members of the Thompson laboratory and members of the Kron Laboratory, in particular John Choy and Jonathan FitzGerald, as well as Navdeep Chandel, for helpful comments and stimulating discussions.
In addition, we thank the University of Chicago MSTP and the University of Pennsylvania Combined Degree Program for their support. S.J.K. thanks the James S. McDonnell and Edward Mallinckrodt Jr. Foundations for their generous support.