Abstract
Organ transplants are rapidly rejected because T cells in the recipient attack the foreign MHC molecules on the graft. The robustness of the T cell response to histoincompatible tissue is not understood. We found that mice have many small T cell populations with Ag receptors specific for a foreign MHC class II molecule type loaded with peptides from leukocytes from the graft. These T cells proliferated modestly after skin transplantation and underwent relatively weak functional differentiation compared with T cells stimulated by a vaccine. Thus, the potency of the T cell response to histoincompatible tissue is likely due to many small T cell populations responding weakly to hundreds of MHC-bound peptides from graft-derived leukocytes.
Introduction
Allogeneic organ transplants are rejected by T lymphocytes in the recipient that recognize MHC class I (MHCI) and class II (MHCII) molecules on the transplant (1). MHCI and MHCII molecules bind short peptides for display on the cell surface where they can be identified by Ag receptors (TCRs) on CD8+ or CD4+ T cells, respectively (2). An individual’s T cells are selected such that clones with TCRs with strong affinity for the MHC-peptide complexes displayed in their thymus are deleted to avert autoimmunity, whereas the clones with TCRs with weak affinity are preserved (3). During infection, a few of these T cells will by chance have TCRs capable of avid binding to MHC-bound peptides derived from the microbe (4). These T cells will proliferate, differentiate into effector cells, and eliminate microbe-infected host cells. This system is detrimental to survival of allografts because the grafted tissue displays MHC molecules that are foreign to the recipient’s T cell repertoire and for which tolerance has not been established (1). Recipient T cells are thought to initially encounter donor MHC molecules on leukocytes or their exosomes that passage from the graft to the recipient’s lymph nodes via lymphatic vessels (5, 6). The rapidity of allograft rejection has been attributed to a high frequency of T cells that directly recognize allogeneic MHC molecules alone or with bound peptides (7) or the abnormally potent activation of individual T cells by graft passenger leukocytes (8).
These possibilities have been difficult to resolve because of an inability to directly detect T cells that express graft-reactive TCRs. Although peptide-MHC tetramers and flow cytometry are well suited to this purpose (9), a lack of knowledge of relevant peptides has limited the use of this powerful technology for studies of graft rejection. Recently, however, Fugmann and colleagues (10, 11) identified over 1000 different mouse peptides that are naturally bound to MHCII molecules of the I-Ab type expressed by C57BL/6 (B6) mice. Many of these peptides were derived from proteins that are abundantly expressed by dendritic cells, which are thought to be important graft passenger leukocytes (12). Thus, it was possible that dendritic cell peptide–I-Ab complexes are targets for T cells in mice lacking I-Ab that receive an organ transplant. We tested this hypothesis by using fluorochrome-labeled I-Ab tetramers containing peptides from CD74 (CLIP), IL-4 receptor α-chain (IL-4Rαp), IL-6 receptor α-chain (IL-6Rαp), lymphocyte cytosolic protein 1 (LCP1p), or CD11b and CD11c (ITGAM/Xp), which are all abundantly expressed by dendritic cells. Our results show that small and distinct T cell populations respond to each of these peptide–I-Ab complexes in recipients of I-Ab–expressing skin allografts, but relatively weakly compared with T cells stimulated by a vaccine.
Materials and Methods
Mice
BALB/c and B6 mice were from the Jackson Laboratory; NOD and I-Ab–deficient mice (13) were from Taconic. All mice were housed in specific pathogen-free conditions in accordance with University of Minnesota Institutional Animal Care and Use Committee regulations. Tail skin from H2-Ma−/− H2-DM–deficient mice (14) and CD74-deficient mice (15) was obtained from L. Eisenlohr (University of Pennsylvania). All mice were 6–10 wk old at the initiation of the experiments.
Transplant and immunization
Tail skin was grafted onto the flanks of recipient NOD or BALB/c mice (16). B6 mice were immunized by s.c. injection of 0.1 ml emulsion of CFA with 100 μg of 2W peptide. Secondary lymphoid organs (SLO) or blood samples were obtained for analysis 7–14 d after transplantation or immunization.
Tetramer production
pRMHa-3 vectors contained antigenic peptide sequences and a flexible polyglycine linker in the I-Ab β-chain as described (17). The sequences were YEVHNPVPLIV (ITGAM/Xp), SQMRMATPLLMR (CLIP), DRYVASLAARNK (IL-4Rαp), KPFQNPVPNQSP (IL-6Rαp), or ASFKDPKISTSL (LCP1p). Insect cells were transfected with the peptide-linker–I-Ab β-chain vectors and a vector encoding the I-Ab α-chain with a BirA ligase site. Biotinylated p:I-Ab monomers were purified from insect cell cultures on a Pierce Monomeric Avidin UltraLink Resin (Thermo Fisher Scientific) and used to make tetramers with streptavidin-PE, streptavidin-allophycocyanin (Prozyme), streptavidin-BV421, or streptavidin-PE-Cy7 (eBioscience).
Cell enrichment and flow cytometry
Spleen and lymph node cells or blood cells were harvested and stained with p:I-Ab tetramers, and in a subset of experiments with CXCR5 (L138D7; BioLegend) Ab for 1 h at room temperature, then incubated with anti-PE and anti-APC magnetic beads. Samples were then passed over magnetized columns (Miltenyi). Bound cells were eluted from the columns and stained for 30 min on ice with Abs from eBioscience to CD3e (145-2C11), CD4 (GK1.5; BD), CD8 (53-6.7), CD90.1 (HIS51), CD90.2 (53-2.1), CD11b (M1/70), CD11c (N418), F4/80 (BM8), CD44 (IM7), CD45.2 (104), B220 (RA3-6B2), or CD45.1 (A20) and with GhostDye Red 780 (Tonbo). Stained cells were fixed and permeabilized with the Foxp3/Transcription Factor Staining Buffer Set (eBioscience) for 1 h at room temperature and stained overnight at 4°C with Abs to T-bet (4B10; BioLegend), Foxp3 (FJK-16s; eBioscience), RORγt (Q31-378; BD), and/or Bcl-6 (K112-91; BD). Cells were analyzed on a Fortessa X-20 (BD) using FlowJo software v10 (Tree Star). The total number of tetramer-positive cells in the enriched fraction from each mouse was determined using AccuCheck Counting Beads (Thermo Fisher Scientific).
Statistical analysis
Differences in T cell expansion and differentiation following transplantation between and across groups was established by one-way or two-way ANOVA corrected for multiple hypothesis testing using either Sidak or Dunnett multiple comparisons test as indicated in each figure legend using Prism version 7.02 (GraphPad), as indicated in each figure legend.
Results and Discussion
We explored the specificity of directly alloreactive T cells using a peptide-MHCII tetramer-based approach. Cells from the SLO of I-Ag7–expressing NOD mice, to which I-Ab molecules are foreign, were stained with pairs of I-Ab tetramers containing the same dendritic cell–derived peptide but labeled with different fluorochromes to maximize the specificity of the assay (18). Tetramer-bound cells were then labeled with magnetic beads, enriched on magnetized columns, stained with Abs for lineage-specific surface markers, and analyzed by flow cytometry (17). CD4+ T cells were identified as viable lymphocyte-sized single cells that expressed CD3 or CD90.2, but not B cell, macrophage, or dendritic cell markers, and CD4, but not CD8 (Fig. 1A). Tetramer-bound cells were identified within the CD4+ T cell population as cells that bound both tetramers.
Identification of tetramer-binding alloreactive T cells. (A) Gates used to identify single viable lymphocytes that expressed CD3 and CD4 but not CD8 or non-T cell lineage markers from tetramer-enriched samples of SLO. (B) CD4+ T cells identified as in (A) from naive NOD mice or NOD mice that received B6 skin grafts 12–14 d earlier. The first and third columns show quadrant gates on the indicated tetramers labeled with different fluorochromes, whereas the second and fourth columns show quadrant gates on one of the tetramers versus CD44. Cells that bound the tetramers via their TCRs are in the upper right quadrants of plots in the first and third columns. The percentages of cells in the indicated quadrants are shown. (C) Numbers of the indicated dual tetramer-binding cells in the SLO of individual naive or B6 skin-grafted NOD or BALB/c mice. Fold expansion values were calculated by dividing the average number of dual tetramer-binding cells in skin-grafted mice by the average number in naive mice. A total of 4–15 mice from two to four independent experiments are shown for each tetramer. (D) Percentages of CD44high cells in the indicated dual tetramer-binding populations from individual naive or B6 skin-grafted NOD or BALB/c mice. (E) Numbers of the indicated dual tetramer-binding CD4+ T cells per million total CD4+ T cells in the blood of individual B6 skin-grafted NOD mice at the indicated times from two independent experiments. Horizontal lines indicate the geometric (C) and (E) or arithmetic mean values (D). Sidak multiple comparisons test was used to correct for multiple hypothesis testing. *p < 0.05, ****p < 0.001. ns, not significant by two-way ANOVA of log10 transformed data (C) and (E) or linear data (D) relative to naive NOD mice.
Identification of tetramer-binding alloreactive T cells. (A) Gates used to identify single viable lymphocytes that expressed CD3 and CD4 but not CD8 or non-T cell lineage markers from tetramer-enriched samples of SLO. (B) CD4+ T cells identified as in (A) from naive NOD mice or NOD mice that received B6 skin grafts 12–14 d earlier. The first and third columns show quadrant gates on the indicated tetramers labeled with different fluorochromes, whereas the second and fourth columns show quadrant gates on one of the tetramers versus CD44. Cells that bound the tetramers via their TCRs are in the upper right quadrants of plots in the first and third columns. The percentages of cells in the indicated quadrants are shown. (C) Numbers of the indicated dual tetramer-binding cells in the SLO of individual naive or B6 skin-grafted NOD or BALB/c mice. Fold expansion values were calculated by dividing the average number of dual tetramer-binding cells in skin-grafted mice by the average number in naive mice. A total of 4–15 mice from two to four independent experiments are shown for each tetramer. (D) Percentages of CD44high cells in the indicated dual tetramer-binding populations from individual naive or B6 skin-grafted NOD or BALB/c mice. (E) Numbers of the indicated dual tetramer-binding CD4+ T cells per million total CD4+ T cells in the blood of individual B6 skin-grafted NOD mice at the indicated times from two independent experiments. Horizontal lines indicate the geometric (C) and (E) or arithmetic mean values (D). Sidak multiple comparisons test was used to correct for multiple hypothesis testing. *p < 0.05, ****p < 0.001. ns, not significant by two-way ANOVA of log10 transformed data (C) and (E) or linear data (D) relative to naive NOD mice.
Small T cell populations were detected with each of the dendritic cell peptide–I-Ab tetramers in naive NOD mice (Fig. 1B). The number of tetramer-binding CD4+ T cells ranged from an average of 11 LCP1p:I-Ab–specific cells per mouse to 41 CLIP:I-Ab–specific cells per mouse (Fig. 1C). In each case, almost all the tetramer-binding cells had the CD44low phenotype of naive cells, as expected because nontransplanted NOD mice had no prior exposure to cells from B6 mice (Fig. 1B, 1D).
NOD mice were then transplanted with B6 skin grafts, which were rejected in 10–12 d. All five of the tetramer-binding populations converted to a CD44high activated phenotype and increased in the SLO of NOD mice 12–14 d after transplant, but to varying degrees (Fig. 1B–D). CLIP:I-Ab and ITGAM/Xp:I-Ab tetramer-binding NOD T cells were also detected in blood samples from NOD recipients of the B6 skin grafts by day 14 posttransplant (Fig. 1E). The average expansion for the five populations in the SLO was 22-fold over baseline with CLIP:I-Ab–specific T cells undergoing the largest expansion (71-fold) and IL4Rαp:I-Ab–specific T cells undergoing the smallest (4-fold) (Fig. 1C). Notably, the graft-responsive CD4 T cells expanded an average of 28-fold following transplantation. This represents a ∼10-fold decrease in expansion compared to I-Ab–restricted CD4 T cells following peptide vaccination, which expanded an average of 248-fold (19). CLIP:I-Ab–specific T cells in I-Ad/I-Ed–expressing BALB/c mice transplanted with B6 skin were also activated and underwent clonal expansion, indicating this phenomenon was not limited to autoimmunity-prone NOD mice (Fig. 1C, 1D).
The specificity of tetramer binding was then examined. ITGAM/Xp:I-Ab or CLIP:I-Ab tetramer-binding T cells increased at most 5-fold in NOD mice transplanted with skin from B6 mice lacking I-Ab due to gene targeting, syngeneic NOD mice, or BALB/c mice expressing I-Ad and I-Ed but not I-Ab (Fig. 2A, 2B). Additionally, most of the CLIP:I-Ab tetramer-binding cells did not bind to the ITGAM/Xp:I-Ab tetramer and most of the ITGAM/Xp:I-Ab tetramer-binding cells did not bind to the CLIP:I-Ab tetramer, although a small population of ∼20 cells per mouse bound to both tetramers (Fig. 2C–E). These results indicate that the TCRs on most of these alloreactive T cells derived their binding affinity from surfaces on both the I-Ab molecule and the peptide, whereas only a minority appeared to bind to the I-Ab molecule alone.
MHCII and peptide specificity of tetramer-binding alloreactive T cells. (A and B) Numbers of ITGAM/Xp:I-Ab (A) or CLIP:I-Ab (B) tetramer-binding cells in the SLO of individual NOD mice that received skin grafts from the indicated strains 12–14 d earlier. Results from two independent experiments are shown for each group. Mean fold expansion values are shown below the values for each transplanted group and were calculated by dividing the average number of tetramer-binding cells in skin-grafted mice by the average number in naive mice. (C) Top left panel shows CD4+ T cells identified as in Fig. 1A from an NOD mouse that received a B6 skin graft 14 d earlier with a gate on cells that bound CLIP:I-Ab tetramers labeled with different fluorochromes. Top right panel shows ITGAM/Xp:I-Ab tetramer-binding cells in the CLIP:I-Ab tetramer-binding population from the top left panel. The percentages of cells in the indicated gate or quadrant are shown. Bottom panels are from the same mouse, showing ITGAM/Xp:I-Ab tetramer-binding cells in the CD4+ T cell population in the left panel and CLIP:I-Ab tetramer-binding cells in the ITGAM/Xp:I-Ab tetramer-binding population in the right panel. (D) Percentages of CLIP:I-Ab tetramer-binding cells that also bound ITGAM/Xp:I-Ab tetramers and ITGAM/Xp:I-Ab tetramer-binding cells that also bound CLIP:I-Ab tetramers. (E) Numbers of cells in the SLO of individual B6 skin-grafted NOD mice that bound CLIP:I-Ab tetramer alone, ITGAM/Xp:I-Ab tetramer alone, or both tetramers. The naive NOD and B6 skin transplant recipient data from the first two groups for both ITGAM/Xp:I-Ab and CLIP:I-Ab are also shown in Fig. 1C and are replotted in (A) and (B) for reference. Results from two independent experiments are shown. Horizontal lines indicate the geometric (A), (B), and (E) or arithmetic mean values (D). Dunnett multiple comparisons test was used to correct for multiple hypothesis testing. ***p < 0.005, ****p < 0.001. ns, not significant by one-way ANOVA of log10 transformed data relative to naive NOD mice (A) and (B), bottom, or relative to B6 skin-grafted NOD mice (B), top.
MHCII and peptide specificity of tetramer-binding alloreactive T cells. (A and B) Numbers of ITGAM/Xp:I-Ab (A) or CLIP:I-Ab (B) tetramer-binding cells in the SLO of individual NOD mice that received skin grafts from the indicated strains 12–14 d earlier. Results from two independent experiments are shown for each group. Mean fold expansion values are shown below the values for each transplanted group and were calculated by dividing the average number of tetramer-binding cells in skin-grafted mice by the average number in naive mice. (C) Top left panel shows CD4+ T cells identified as in Fig. 1A from an NOD mouse that received a B6 skin graft 14 d earlier with a gate on cells that bound CLIP:I-Ab tetramers labeled with different fluorochromes. Top right panel shows ITGAM/Xp:I-Ab tetramer-binding cells in the CLIP:I-Ab tetramer-binding population from the top left panel. The percentages of cells in the indicated gate or quadrant are shown. Bottom panels are from the same mouse, showing ITGAM/Xp:I-Ab tetramer-binding cells in the CD4+ T cell population in the left panel and CLIP:I-Ab tetramer-binding cells in the ITGAM/Xp:I-Ab tetramer-binding population in the right panel. (D) Percentages of CLIP:I-Ab tetramer-binding cells that also bound ITGAM/Xp:I-Ab tetramers and ITGAM/Xp:I-Ab tetramer-binding cells that also bound CLIP:I-Ab tetramers. (E) Numbers of cells in the SLO of individual B6 skin-grafted NOD mice that bound CLIP:I-Ab tetramer alone, ITGAM/Xp:I-Ab tetramer alone, or both tetramers. The naive NOD and B6 skin transplant recipient data from the first two groups for both ITGAM/Xp:I-Ab and CLIP:I-Ab are also shown in Fig. 1C and are replotted in (A) and (B) for reference. Results from two independent experiments are shown. Horizontal lines indicate the geometric (A), (B), and (E) or arithmetic mean values (D). Dunnett multiple comparisons test was used to correct for multiple hypothesis testing. ***p < 0.005, ****p < 0.001. ns, not significant by one-way ANOVA of log10 transformed data relative to naive NOD mice (A) and (B), bottom, or relative to B6 skin-grafted NOD mice (B), top.
The role of the peptide in the stimulation of CLIP:I-Ab tetramer-binding T cells was next analyzed in more detail. CLIP is part of the CD74 invariant chain protein, which binds to newly synthesized MHCII molecules (20). Proteases trim away most of the protein leaving CLIP bound to the MHCII groove. H2-DM molecules then remove CLIP and catalyze the binding of other peptides as evidenced by the fact that nearly all MHCII molecules are loaded with CLIP in H2-DM–deficient mice (14, 21). Skin from H2-DM–deficient, B6, and CD74-deficient B6 mice therefore contains cells with most, some, or none of their I-Ab molecules loaded with CLIP. We found that CLIP:I-Ab tetramer-binding NOD T cells increased 684-fold over baseline in NOD recipients of H2-DM–deficient B6 skin compared with 71-fold in recipients of wild-type B6 skin (Fig. 2B). Interestingly, CLIP:I-Ab–specific NOD T cells still increased 15-fold in NOD recipients of CD74-deficient B6 skin grafts, which could have been due to CLIP from the NOD recipients exchanging into some of the I-Ab molecules on cells from the graft. The fact that the amount of expansion of CLIP:I-Ab–specific NOD T cells scaled with the amount of CLIP:I-Ab complexes on the cells from the graft is strong evidence that these T cells recognize the peptide component of the complex. It is remarkable that CLIP:I-Ab–specific TCRs in an I-Ag7–tolerized T cell repertoire could distinguish CLIP:I-Ab from CLIP:I-Ag7 given that the CLIP sequence in the two complexes was identical.
We then determined the phenotype of alloreactive T cells in NOD mice to gain mechanistic insight into graft rejection. These cells were compared with B6 T cells specific for an I-Ab–bound nonmouse peptide called 2W (22), in B6 mice immunized 14 d earlier with 2W peptide. The 2W:I-Ab–specific population expanded >100-fold over baseline (Fig. 3A) and consisted of a mixture of T-bet+ T helper 1 (Th1) macrophage helpers, Foxp3+ regulatory T cells, RORγt+ Th17 neutrophil helpers, CXCR5+ or Bcl-6+ T follicular B cell helpers (Tfh), and uncommitted effector cells (Fig. 3B, 3C). The CLIP:I-Ab– and ITGAM/Xp:I-Ab–specific T cell populations in NOD recipients of B6 skin grafts expanded less than the 2W:I-Ab–specific T cells in 2W-primed B6 mice and were composed of Th1 and Tfh cells but completely lacked Th17 cells. The latter finding may shed some light on the controversy about the role that Th17 cells play in direct alloreactivity (23).
Activation phenotypes of tetramer-binding alloreactive T cells. (A) Numbers of 2W:I-Ab tetramer-binding T cells in the SLO of individual B6 mice injected with 2W peptide in CFA 14 d earlier, or CLIP:I-Ab or ITGAM/Xp:I-Ab tetramer-binding cells in the SLO of individual NOD mice that received B6 skin grafts 12–14 d earlier. Results from two to six independent experiments are shown. (B) Left column shows CD4+ T cells identified as in Fig. 1A from the indicated mice with quadrant gates on the indicated tetramer staining. Middle column shows dual tetramer-binding cells from the upper right quadrants of plots in the left column with gates on T-bet and Foxp3 staining. Right column shows cells lacking T-bet and Foxp3 from the plots in the middle column with gates on RORγt and CXCR5 staining. The percentages of cells in the indicated gates or quadrant are shown with designations for the relevant Th effector cell types. (C) Percentages of regulatory T cells, Tfh, Th1, Th17, and other effector cell types in the indicated tetramer-binding cell populations. Results from two to six independent experiments are shown. Horizontal lines indicate the geometric (A) or arithmetic (B) mean values. Dunnett multiple comparisons test was used to correct for multiple hypothesis testing. **p < 0.01, ***p < 0.005, ****p < 0.001. ns, not significant by two-way ANOVA of linear data (C) relative to corresponding 2W:I-Ab responses in B6 mice.
Activation phenotypes of tetramer-binding alloreactive T cells. (A) Numbers of 2W:I-Ab tetramer-binding T cells in the SLO of individual B6 mice injected with 2W peptide in CFA 14 d earlier, or CLIP:I-Ab or ITGAM/Xp:I-Ab tetramer-binding cells in the SLO of individual NOD mice that received B6 skin grafts 12–14 d earlier. Results from two to six independent experiments are shown. (B) Left column shows CD4+ T cells identified as in Fig. 1A from the indicated mice with quadrant gates on the indicated tetramer staining. Middle column shows dual tetramer-binding cells from the upper right quadrants of plots in the left column with gates on T-bet and Foxp3 staining. Right column shows cells lacking T-bet and Foxp3 from the plots in the middle column with gates on RORγt and CXCR5 staining. The percentages of cells in the indicated gates or quadrant are shown with designations for the relevant Th effector cell types. (C) Percentages of regulatory T cells, Tfh, Th1, Th17, and other effector cell types in the indicated tetramer-binding cell populations. Results from two to six independent experiments are shown. Horizontal lines indicate the geometric (A) or arithmetic (B) mean values. Dunnett multiple comparisons test was used to correct for multiple hypothesis testing. **p < 0.01, ***p < 0.005, ****p < 0.001. ns, not significant by two-way ANOVA of linear data (C) relative to corresponding 2W:I-Ab responses in B6 mice.
Our results demonstrate that directly alloreactive T cells recognize an allogeneic MHCII molecule and a graft leukocyte peptide in the same way that microbe-reactive T cells recognize a syngeneic MHCII molecule and a microbial peptide. This finding argues against the potency of direct alloreactivity being due to a high frequency of T cells that recognize MHC molecules alone. In addition, our finding that most of the individual graft peptide–MHCII-specific T cell populations underwent relatively weak clonal expansion and did not form Th17 cells is evidence against the possibility that the ferocity of allograft rejection is related to abnormally potent activation of individual T cells. Rather, our results indicate that allografts are rapidly rejected because the weak activation of small numbers of T cells specific for hundreds of different graft leukocyte peptide–allogeneic MHC complexes results in a large total sum of graft-attacking cells. The relatively poor activation of alloreactive T cells could stem from grafts lacking pathogen-associated molecular pattern molecules that strongly enhance the costimulatory functions of APCs (24). This form of activation, however, was sufficient to generate a large component of B cell–helping Tfh cells, which may be relevant to Ab-mediated rejection.
The unique function of CLIP in MHCII maturation makes it a potentially universal alloantigen. Because the identical CLIP fragment binds to all allelic forms of MHCII and some CLIP-MHCII complexes make it to the cell surface (25), these complexes will be displayed on passenger leukocytes from all allografts. The availability of over a dozen allelic forms of human MHCII tetramers containing CLIP creates a situation where at least one graft-reactive T cell population could be tracked following transplantation of MHCII-disparate tissues. The number and activation status of CLIP–donor MHCII-specific T cells in the blood or graft of a transplant recipient could be a useful biomarker for graft rejection.
Acknowledgements
We thank J. Walter, N. Sahli, J. Wilson, and E. Finger for technical assistance. We thank L. Eisenlohr for providing tail skin from H2-DM–deficient and CD74-deficient B6 mice.
Footnotes
This work was supported by grants awarded by the National Institutes of Health (P01 AI035296, R01 AI039614, and R37 AI027998 to M.K.J.; R01 AI106791, P01 AI035296, and U24 AI118635 to B.T.F.; F32 AI114050 to D.M.; and T32 DK007203 to A.L.B.), a grant awarded by Regenerative Medicine Minnesota (11215 TR002 to B.T.F.), and a grant awarded by the Juvenile Diabetes Research Foundation (2-2011-662 to B.T.F.).
References
Disclosures
The authors have no financial conflicts of interest.