T Cell Costimulation through CD28 Depends on Induction of the Bcl-xγ Isoform : Analysis of Bcl-xγ–deficient Mice

The T cell response to antigen requires costimulatory signals that promote efficient T cell expansion and cytokine secretion. Transduction of costimulatory signals by the CD28 receptor is a pivotal element in the full stimulation of T cells and prevention of anergy in vitro (1). CD28-dependent signals are also required for effective in vivo T cell responses to allografts (2) and for development of experimental autoimmune encephalomyelitis (EAE;^* reference 3). Recent studies have suggested that CD28-dependent effects on T cell survival and proliferation may represent biochemically distinct pathways that originate from different sites on the CD28 molecule's cytoplasmic domain. One pathway, marked by Bcl-xL expression, may contribute to T cell survival upon growth factor withdrawal in vitro (4) and may enhance survival of mature thymocytes (5, 6) and peripheral T cells (7, 8). A second pathway, proposed to account for the biologic signature of CD28-dependent costimulation-enhanced proliferation and cytokine production, has not been well characterized (9, 10).

Bcl-xγ, a recently identified isoform of the Bcl-x gene, is expressed after TCR and CD28 ligation and may contribute to successful proliferation and IL-2 secretion after stimulation by antigen (11). Importantly, the expression of Bcl-xγ is mainly confined to the T cell compartment (11), while Bcl-xL has a much wider tissue distribution (4). These findings opened the possibility that Bcl-xγ function may be more specifically tailored to T cell activation. In contrast, Bcl-xL, which exerts antiapoptotic activity in a wide variety of different cell types (4), can enhance the survival, but not proliferation of activated T cells (8). Therefore, it was of interest to determine whether Bcl-xγ function might overlap with Bcl-xL and potentiate cell survival or exert a distinctive and complementary role in T cell activation. To this end, we generated mice that either lack or constitutively express Bcl-xγ in T cells. We find that Bcl-xγ–deficient T cells fail to respond to CD28-dependent costimulatory signals according to cell proliferation and cytokine expression, while enforced expression of Bcl-xγ largely replaces the requirement for CD28-dependent costimulatory signals; manipulation of Bcl-xγ expression had no significant effect on T cell survival.

Bcl-xγ Tg mice containing full-length Bcl-xγ cDNA under control of the Lck^Dis (p56^lck) promoter (12) were generated and backcrossed to C57BL/6 for eight generations. Enforced expression of Bcl-xγ (in a T cell–specific manner) in these transgenic mice was confirmed by Western blot analysis as well as by RT-PCR, using oligos that distinguish transgenic and endogenous RNA transcripts. C57BL/6 Bcl-xγ Tg mice were also crossed with Balb/c DO11.10 TCR Tg mice to generate C57BL/6 × Balb/c Bcl-xγ Tg DO11.10 TCR Tg mice and Bcl-xγ wild-type (wt) DO11.10 TCR Tg control littermates.

TC1 embryonic stem (ES) cells were electroporated with 15 μg of PvuI-linearized Bcl-xγ^−/− vector DNA and selected in media containing G418 and Gancyclovir as described previously (14). Bcl-xγ^+/− ES cells were identified by Southern blot analysis using the 5′ knockout probe on HindIII-digested DNA and confirmed by using the 3′ knockout probe on BglII-digested DNA. Bcl-xγ^+/− ES cells were transiently transfected by electroporation with 30 μg/ml of pMC-CreN, and clones were isolated and screened by Southern blot analysis for Cre-loxP–mediated deletion of the neo^r gene. Targeted TC1 ES cells were used to generated chimeric mice. Bcl-xγ^+/− ES cells were injected into C57BL/6 blastocysts, and chimeric mice capable of transmitting the Bcl-xγ^+/− allele to offspring were obtained from two independent clones. Offspring from these mice were bred to 129sv to generate Bcl-xγ^+/+, Bcl-xγ^+/−, and Bcl-xγ^−/− mice.

Single-cell suspensions were prepared from thymus, spleen, and LNs, where indicated. Erythrocytes were lysed with ammonium chloride lysis buffer, and cells were washed in 2% FBS/PBS. Cell surface molecules were detected by staining 10⁶ viable cells with indicated conjugates for 20 min at 4°C in 2% FBS/PBS. Cells were stained with FITC or R-phycoerythrin or Cy-Chrome–conjugated antibodies. Data from ∼50,000 size-gated cells were acquired on a Coulter Epics XL flow cytometer (Beckman Coulter).

Genomic DNA was isolated and Southern blotting was performed as described previously (13), using Zetaprobe membranes (Bio-Rad Laboratories) and probes generated by random hexamer priming (Boehringer Mannheim) using (γ-³²P)dCTP (NEN Life Science Products). The 5′ knockout probe is a 1-kb NotI-EcoRI genomic fragment, and 3′ knockout probe is a 600-bp MscI-XbaI genomic DNA fragment.

Total RNA was prepared from cultured T cells and mouse thymus using RNAzole (Tel-Test, Inc.). RT-PCR and PCR cloning were performed as described previously (4). In brief, total RNA from mouse thymus was reverse transcribed in a GeneAmp PCR 9600 (Perkin-Elmer) using a GeneAmp RNA-PCR kit (Perkin-Elmer) at 45°C for 60 min, followed by 95°C for 3 min, and then placed on ice. For Bcl-xγ detection, primers (e4f: 5′-GGA ATT CAG TGG AGG TAC ACC CCT CAG ATCT-3′ and e4b: 5′-CCA CTC CTG GCA CTT GAA CAA GCCT-3′) specific for the 5′ and 3′ ends of exon 4 were used (see Fig. 1 A). To detect Bcl-xL, primers to exon 2 (3e: 5′-TCG CTC GCC CAC ATC CCA GCT TCA CAT AAC CCC-3′) and exon 3 (15e: 5′-CCA CCA ACA AGA CAG GCT-3′) were used (see Fig. 1 A). Each three-step thermal cycle for routine PCR analysis consisted of 30 s at 95°C, 30 s at 60°C, and 60 s at 72°C. Additional PCR reactions were performed for 35–40 cycles using Advantage cDNA Polymerase Mix (CLONTECH). PCR products were visualized by agarose gel electrophoresis and ethidium bromide staining, and were positively identified by size markers. A negative control containing all reagents except cDNA was included in each PCR analysis.

For immunoblot detection of Bcl-xγ and Bcl-xL protein, 100 μg of primary T cell PNS (per lane) were resolved on SDS-polyacrylamide gels containing 10% acrylamide and 0.27% N,N′-methylene bis-acrylamide under nonreducing conditions. To obtain better separation of Bcl-xγ and Bcl-xL proteins, a full-length gel apparatus of 20 × 20 cm was used. After electrophoresis, proteins were transferred to PVDF membranes (Millipore). Membranes were blocked, washed, and probed with an affinity-purified rabbit anti–Bcl-x antibody (R&D Systems) at a final concentration of 0.3 μg/ml. To confirm the identity of the (top) Bcl-xγ band, membranes were stripped and reprobed with an affinity-purified rabbit antibody (1.0 μg/ml) specific to the unique COOH terminus of Bcl-xγ. HRP-conjugated secondary antibodies and chemiluminescence detection reagents were from Amersham Pharmacia Biotech.

CD4⁺ T cells were isolated from LNs of Bcl-xγ^+/+, Bcl-xγ^−/−, or Bcl-xγ Tg mice by negative section. In brief, LN cells were incubated with rat anti–mouse B220, Mac-1, Gr-1, CD8, and NK1.1 antibodies (BD PharMingen) for 30 min, washed, and incubated with magnetic beads coated with sheep anti–rat antibody (Dynal) for 45 min. Finally, antibody-coated cells were removed by magnetic separation and the isolated T lymphocytes were washed again. LN T cells (5 × 10⁵ cells per well) were cultured in triplicates in round-bottomed 96-well plates (Costar) in 200 μl of freshly prepared Iscove's DMEM (10% FCS, 50 μM 2-mercaptoethanol). T cells were activated by culturing in the presence of plate-bound anti-CD3 (145.2C11 clone) and anti-CD28 antibodies (BD PharMingen) at final concentrations of 0.2 and 1.0 μg/ml, respectively. Alternatively, soluble anti-CD3 (1 μg/ml) and IL-2 (BD PharMingen) were used at the indicated concentrations. T cells were cultured for 48 h, and 1 μCi ³[H]thymidine was added to the wells for the last 18 h of culture. T cell culture supernatants were assayed for IL-2 in triplicates at indicated time points by ELISA. Where indicated, supernatants were harvested after 24, 48, and 72 h of culture. IL-2 levels were determined by sandwich ELISA, according to the manufacturer's instructions (BD PharMingen).

EAE was induced in mice by immunization with a 12mer peptide representing amino acids 172–183 (PVYIYFNTWTTC) of proteolipid protein peptide (PLP)172–183. Per mouse, 150 μg of PLP peptide and 300 μg of killed Mycobacterium tuberculosis in CFA were injected (subcutaneously), by means of three injections over the flanks. In addition, 400 ng of pertussis toxin was injected, intraperitoneally on days 0 and 2. Disease was scored as follows: score 0, normal mouse with no sign of disease; score 1, limp tail; score 2, limp tail and partial hind limb weakness; score 3, partial hind limb paralysis; score 4, complete hind limb paralysis; and score 5, moribund state per death by EAE.

24- or 96-well plates containing plate-bound anti-CD3 antibody (0.1–0.2 μg/ml, clone 145.2C11; BD PharMingen) were washed three times with PBS before addition of cells. Cells (either 5 × 10⁶ or 5 × 10⁵ cells per well) were washed twice with PBS at the indicated times, and resuspended in 1× Annexin V binding buffer. Cells in the early stages of apoptosis measured positive for fluorescein-Annexin V (Annexin V-FITC; BD PharMingen) as determined in an Epics XL FACScan™ using standard settings in semilogarithmic mode, and these cells contained higher forward scattering levels (for gating) than dead cells in the samples. Fully activated T cells from Bcl-xγ^−/− mice exhibited a 10–20% increase in apoptosis versus control (wt) cells while the levels of apoptosis in nonactivated Bcl-xγ Tg⁺ T cells were only 10–30% lower compared with nonactivated control T cells. In certain experiments, cells were stained with FITC-conjugated anti-CD3 antibody or propidium iodide (PI) (BD PharMingen) before measurement of cell survival.

The percentage of cells in G₀/G₁, S, and G₂/M phases was determined by cytofluorometric analysis after PI staining.

For PI staining, ∼10⁶ cells were suspended in 100 μl PBS and 1 ml of ice-cold 80% ethanol was added drop-wise, while vortexing. Samples were then incubated for 1 h on ice. Cells were then washed gently twice with PBS, and resuspended in 250 μl of PBS containing 12.5 μg of RNase. Cells were incubated at 37°C for 30 min before staining cellular DNA with 250 μl of a PI solution (50 μg/ml) in PBS for 1 h at 37°C. Cells were then immediately place on ice and analyzed using FACScan™. The 1n chromosome-containing lymphocyte peak (main peak) was set to identical values on the semilogarithmic scale at the beginning of each analysis. Cell yields from cultures of Bcl-xγ wt, −/−, or Bcl-xγ Tg T cells were not significantly different at the time of analysis (36 h after stimulation).

Peripheral and mesenteric LN lymphocytes from Bcl-xγ wt and −/− mice were isolated and washed in 2% FBS/PBS. To monitor the proliferation of these donor cells in recipient (syngenic Bcl-xγ wt) mice, CFSE (Molecular Probes) labeling was performed by incubating 10⁷ cells per milliliter in PBS/0.1% BSA with CFSE (final concentration 10 μM) for 10 min at 37°C. Labeled cells were washed twice with PBS/0.1% BSA before they were resuspended in sterile PBS for tail vein injection. Approximately 5 × 10⁶ of CFSE-labeled cells were injected, intravenously, per mouse.

We selectively disrupted this isoform within the Bcl-x locus to generate Bcl-xγ–deficient mice. Analysis of Bcl-xγ and Bcl-x genomic structure (14) has indicated that mature Bcl-xγ RNA transcripts are comprised of coding exons 1, 2, and 4 (Fig. 1 A) joined with alternatively-spliced combinations of exons 5–8 that comprise several distinct noncoding 3′-UTR (14). We generated Bcl-xγ–deficient mice (Materials and Methods) using a Cre/LoxP-mediated gene targeting vector to delete exon 4 in ES cells as outlined in Fig. 1 B. Southern blot analysis of tail cut DNA from ES cell–injected mice showed that exon 4 had been successfully disrupted (Fig. 1 C, left panel). Additional RT-PCR analysis with primers specific for exon 4 (for Bcl-xγ) or exons 2 and 3 (for Bcl-xL) confirmed that Bcl-xγ mRNA was no longer expressed while Bcl-xL mRNA expression was unchanged in the same activated T cells in three independent experiments (Fig. 1 C). To document the effect of this mutation on Bcl-xγ and Bcl-xL protein expression, Western blotting of Bcl-xγ and Bcl-xL after coligation of CD3 and CD28 was performed. These experiments confirmed other studies (14) that CD28 coligation is essential for full and early expression of Bcl-xγ in activated T cells (Fig. 1 D, left panel and legend). They also showed that CD3/CD28 coligation failed to induce detectable Bcl-xγ protein in Bcl-xγ^−/− T cells after overexposure of Western blots (Fig. 1 D, right panel). At the same time, expression of Bcl-xL protein after CD3 or CD3/CD28 ligation was indistinguishable in Bcl-xγ^−/− and Bcl-xγ wt T cells according to normally exposed Western blots (Fig. 1 E).

We measured the effects of CD3 and CD28 coligation on the proliferative and IL-2 responses of Bcl-xγ^−/− T cells (15, 16). As expected, CD3/CD28 coligation of wt T cells resulted in a 4–6-fold increase in ³[H]thymidine incorporation and IL-2 production compared with anti-CD3 alone; however, CD28 coligation failed to augment the response of T cells from Bcl-xγ^−/− mice (Fig. 2, A and B), suggesting that the Bcl-xγ mutation nullifies CD28-dependent costimulation.

Then, we asked whether enforced expression of Bcl-xγ in T cells might replace the need for CD28-dependent costimulation. Expression of a Bcl-xγ transgene under the control of the Lck distal promoter resulted in constitutive overexpression of Bcl-xγ in peripheral T cells independent of overt activation (Fig. 3 A). Enforced Bcl-xγ expression did not enhance survival of unstimulated T cells, in contrast to the reported effects of Bcl-xL (Fig. 3 B, legend). However, proliferative and IL-2 responses of T cells from Bcl-xγ Tg⁺ mice after CD3 ligation were 3–4-fold greater than T cells from nontransgenic littermates and were not significantly increased by CD28 coligation (Fig. 3, B and C). Moreover, enforced expression of Bcl-xγ restored the response of CD4 DO11.10⁺ T cells to OVA peptide presented by B7.1/B7.2^−/− APC to levels provoked by B7.1/B7.2⁺ APC (Fig. 3 D).

CD28 costimulation induces cell cycle progression through IL-2–dependent and IL-2–independent mechanisms (17–19). Although earlier studies of T cell lines that contained Bcl-xγ antisense constructs indicated a critical role in successful T cell activation, it was not possible to distinguish between a primary effect on cellular proliferation versus antigen-induced apoptosis (11) since the effects of Bcl-xγ expression could not be evaluated in nontransformed T cells. We find that the proliferative response of Bcl-xγ^−/− T cells to CD3/CD28 coligation was reduced by >80% according to ³[H]thymidine incorporation and IL-2 expression (e.g., Fig. 2), while levels of apoptosis changed <15%. Levels of apoptosis and cell death of nonactivated wt. Bcl-xγ Tg or Bcl-xγ^−/− T cells were indistinguishable over a 96-h incubation period (not shown).

The proliferative response of Bcl-xγ Tg, −/−, and wt T cells after TCR ligation was analyzed for alterations in progress through the cell cycle. Coligation of CD3/CD28 that resulted in a doubling of progression of wt T cells into G₂/M did not enhance progression of Bcl-xγ^−/− T cells (Fig. 4 A). This observation was complemented by the finding that enforced Bcl-xγ expression largely replaced the need for CD28 ligation for efficient entry into G₂/M (Fig. 4 A). A role for Bcl-xγ in the IL-2–independent component of CD28-dependent regulation of T cell proliferation was underlined by the inability of IL-2 to fully restore the Bcl-xγ^−/− T cell response to CD3 or CD3/CD28 coligation (Fig. 4, C and D). Finally, we asked whether the reduced ability of Bcl-xγ^−/− T cells to proliferate after antibody-dependent receptor ligation in vitro was also apparent in vivo. Analysis of CFSE-labeled Bcl-xγ^−/− T cells in mice stimulated with the superantigen staphylococcal enterotoxin A (SEA; references 20 and 21) revealed that SEA-dependent replication was reduced by ∼60% (Fig. 4 B).

Bcl-x_L has been characterized as an essential molecule in the regulation of cell survival and death (4). It was shown that overexpression of Bcl-xL could restore cell survival in CD28-deficient T cells, which normally fail to upregulate this molecule (8). Similarly, when Bcl-xγ was first described, this new isoform of Bcl-x appeared to modulate the sensitivity of CTLL-2 cells to apoptosis after TCR ligation (11). However, levels of apoptosis were found to differ only slightly between Bcl-xγ–deficient and wt T cells after TCR stimulation (unpublished data). To further investigate the possible role of Bcl-xγ in the regulation of cell survival and death, Bcl-xγ–deficient T lymphocytes were compared with Bcl-xγ Tg cells in their response to cytotoxic agents. We found no difference in survival between the two cell types after either irradiation (1,000 rad), treatment with dexamethasone (5 μM) or culture in the absence of stimulation for 96 h (Fig. 5 A). A comparison of the response of T cells from Bcl-xγ Tg/ DO11.10 TCR Tg mice with cells from Bcl-xγ wt/ DO11.10 TCR Tg littermates after successive stimulation with OVA 323–339 peptide followed by TCR religation with anti-CD3 antibody indicated that overexpression of Bcl-xγ did not protect cells from activation-induced cell death (AICD); both groups of T cells displayed increased cell death and dramatic reduction of the proliferative response after TCR religation (Fig. 5 B). Together, these results suggest that the proliferative advantage conferred to T cells by the Bcl-xγ transgene is not mediated through enhanced survival, and that Bcl-xγ does not seem to play an obvious inhibitory role in the process of apoptosis.

CD28-dependent costimulatory signaling is essential for development of several T cell–mediated autoimmune diseases (3, 22, 23) including experimental autoimmune encephalomyelitis (EAE), which depends on efficient expansion of T cells that recognize peptides derived from myelin proteins (for a review, see reference 24). In view of the defective CD28-dependent costimulatory response of Bcl-xγ^−/− T cells noted above, we defined EAE susceptibility of Bcl-xγ^−/− mice after immunization with PLP172–183. Bcl-xγ^−/− mice failed to develop significant EAE compared with substantial disease displayed by Bcl-xγ^+/+ or ^+/− littermates (Fig. 6 A) and displayed defective development of T cell memory to PLP172–183. T cells from PLP-immunized Bcl-xγ^−/− mice exhibited a substantial (50–100-fold) reduction in secondary response as judged by PLP-specific proliferation and IL-2 production after in vitro restimulation by PLP172–183 peptide (Fig. 6 B).

Although development of thymocytes and T cells from Bcl-xγ^−/− and Bcl-xγ Tg mice was not obviously dysregulated according to FACS^® analysis (Table I), levels of memory T cells, as judged by expression of CD62L and CD44, were decreased by ∼50% in Bcl-xγ^−/− mice and increased by this amount in Bcl-xγ Tg mice (Fig. 6 C). These findings suggest a direct correlation between Bcl-xγ expression and successful generation of memory T cells secondary to experimental (Fig. 6, A and B) or environmental (Fig. 6 C) antigenic stimuli.

An efficient T cell response to antigen requires costimulatory signals provided by the CD28 receptor that promote cell expansion and cytokine secretion. Although the importance of CD28 ligation to complete T cell activation is well documented (25), its molecular basis remains poorly understood. Since expression of the Bcl-xγ isoform is tightly coupled to CD28/CD3 coligation, we investigated its potential role in expression of the costimulatory phenotype using mice containing T cells that specifically lack or overexpress this member of the Bcl-x family.

LoxP-dependent excision of exon 4 of the Bcl-x gene eliminated Bcl-xγ expression without a detectable effect on expression of Bcl-xL at the RNA or protein level (Fig. 1). This genetic lesion was associated with defective T cell memory in vivo; Bcl-xγ^−/− mice displayed reduced numbers of CD45RB⁺ CD44^hi T cells, reduced secondary responses after restimulation with PLP172–183 peptide, and failed to develop PLP-induced EAE. Analysis of the response of Bcl-xγ^−/− T cells in vitro mapped this phenotype to a defect in CD28-dependent costimulation, since CD3-dependent activation was unimpaired while CD28 coligation failed to enhance ³[H]thymidine incorporation, IL-2 production, and entry into G₂/S. These findings were complemented by studies showing that enforced expression of Bcl-xγ largely replaced the need for CD28-dependent costimulation and the response of Bcl-xγ Tg T cells to CD3 ligation was similar to the response of nonTg T cells to CD3/CD28 coligation.

Ligation of CD28 enhances T cell proliferation through increased expression of IL-2 as well as IL-2–independent acceleration of entry into cell cycle (17, 25, 26). Bcl-xγ appears to contribute to both elements of costimulation; although Bcl-xγ^−/− T cells display reduced IL-2 production, reduced proliferation is only partially remedied by exogenous IL-2 and IL-2–independent entry of Bcl-xγ–deficient T cells into G₂ is impaired (Fig. 4). Since expression of other Bcl-x isoforms including Bcl-xL is unaltered in these cells (Fig. 1), these experiments underline the essential intracellular role of Bcl-xγ in expression of CD28-dependent costimulation.

Initial studies of Bcl-xγ showed that this Bcl-x isoform is expressed in T cells that successfully proliferate after engagement of the TCR by peptide antigen, whereas expression of Bcl-x antisense constructs in T cell lines reduced proliferation and increased apoptosis (11). However, analysis of the effects of dysregulated Bcl-xγ gene expression in transformed cells did not delineate the potential role of this cytosolic protein in T cell proliferation and CD28-dependent costimulation. The finding that CD28-dependent costimulation depends on expression of the cytosolic Bcl-xγ protein supports an emerging view that CD28-dependent enhancement of T cell survival and T cell proliferation to antigen may reflect separate signaling pathways originating from distinct regions of the CD28 cytoplasmic domain (9, 10). Other studies have indicated that Bcl-xL expression marks a costimulatory signaling pathway that contributes to T cell survival (4, 7) but not proliferation or cytokine expression after antigen stimulation (8–10). Our studies suggest that Bcl-xγ expression is essential for enhanced entry into cell cycle and IL-2 production after TCR ligation, but not for protection from AICD or irradiation or dexamethasone-induced apoptosis. Furthermore, overexpression of Bcl-xγ did not enhance cell survival in the absence of stimuli or after TCR ligation. These findings support a division of labor among the Bcl-x isoforms that relies on Bcl-xγ enhancement of T cell proliferation and Bcl-L to potentiate cell survival after CD28 ligation.

Unlike Bcl-xL, which is tethered to the outer surface of mitochondrial membranes, the Bcl-xγ protein locates in the cytosol, lacks a hydrophobic tail, and may be phosphorylated at serine/threonine (14). A two-hybrid screen using the unique COOH terminus of Bcl-xγ has suggested that the Jun-binding protein QM (27, 28) specifically interacts with Bcl-xγ, and not Bcl-xL (unpublished data), opening the possibility that CD28-dependent induction of Bcl-xγ may facilitate DNA binding of AP-1 or Jun/Jun dimers. Alternatively, Bcl-xγ may interact with caspases and/or caspase-activating molecules in a similar fashion to Bcl-xL (29). According to the latter view, Bcl-xγ might specifically block negatively acting caspases while sparing stimulatory caspases (30, 31). Insufficient Bcl-xγ induction after incomplete T cell activation would thus allow negatively acting caspases to dampen T cell activation. Although further experiments are required to precisely define the mechanism of Bcl-xγ−dependent enhancement of the T cell response, definition of the complementary roles of Bcl-xγ and Bcl-xL in CD28-dependent costimulation represents an important step toward understanding this activation process.

We thank Dr. L. Klein for advice and reagents and Ms. Alison Angel for kind assistance in the preparation of this manuscript.

This work was supported in part by National Institutes of Health research grants A13600, A112184, and AI48125 to H. Cantor, AI10468 to L.A. Trimble, AI07386 to X.-F. Yang, and AI20047 and AI35714 to F.W. Alt, who is also supported by the Howard Hughes Medical Institute; B.P. Sleckman is a recipient of a Burroughs Wellcome Fund Career Development Award; B. Press is a fellow of the Cancer Research Institute; C.H. Bassing is a fellow of the Irvington Institute; and M. Jansson is a fellow of the Swedish Foundation for International Cooperation in Research and Higher Education (STINT).