Environmental and genetic factors define the susceptibility of an individual to autoimmune disease. Although common genetic pathways affect general immunological tolerance mechanisms in autoimmunity, the effects of such genes could vary under distinct immune challenges within different tissues. In this study, we demonstrate this by observing that autoimmune type 1 diabetes-protective haplotypes at the insulin-dependent diabetes susceptibility region 10 (Idd10) introgressed from chromosome 3 of C57BL/6 (B6) and A/J mice onto the NOD background increase the severity of autoimmune primary biliary cirrhosis induced by infection with Novosphingobium aromaticivorans, a ubiquitous alphaproteobacterium, when compared with mice having the NOD and NOD.CAST Idd10 type 1 diabetes-susceptible haplotypes. Substantially increased liver pathology in mice having the B6 and A/J Idd10 haplotypes correlates with reduced expression of CD101 on dendritic cells, macrophages, and granulocytes following infection, delayed clearance of N. aromaticivorans, and the promotion of overzealous IFN-γ– and IL-17–dominated T cell responses essential for the adoptive transfer of liver lesions. CD101-knockout mice generated on the B6 background also exhibit substantially more severe N. aromaticivorans-induced liver disease correlating with increased IFN-γ and IL-17 responses compared with wild-type mice. These data strongly support the hypothesis that allelic variation of the Cd101 gene, located in the Idd10 region, alters the severity of liver autoimmunity induced by N. aromaticivorans.
Complex interactions between environmental factors and genetic traits (1–4) underpin the etiology of autoimmune diseases and the functioning of the immune system. Although certain genetic regions are associated with multiple autoimmune diseases, in some cases, the protective allele for one autoimmune disease is the susceptibility allele for another (5–9). The nature of environmental triggers, effector mechanisms mediating various autoimmune pathologies, or tissue-specific events that enable an allele to be protective for one or more autoimmune diseases on the one hand but increase susceptibility to different autoimmune targets on the other hand requires definition. The IL2RA locus, for example, is associated with both type 1 diabetes (T1D) and multiple sclerosis; however, an allele that is protective in one disease is a susceptibility allele for the other (10). NOD congenic mice mimic this situation in humans: when the diabetogenic H2g7 of the NOD mouse is replaced with the H2h4 haplotype, the congenic mice are protected from T1D but now develop autoimmunity directed against the thyroid gland, which is exacerbated by exposure to dietary iodine (11–13). It is believed that the underlying single nucleotide polymorphisms that redirect the organ-specific autoimmune manifestation likely evolved due to microbial pressure and reveal a consequence of natural selection for altered susceptibility to certain pathogens (14–16); selective pressure to promote functional variation in immune-related genes to resist infectious challenge could add to the pool of variants that alter autoimmune disease susceptibility. In this study, we have investigated the effect of alleles that protect from T1D in our infection-induced mouse model of primary biliary cirrhosis (PBC).
The diagnostic hallmark of PBC is the development of autoantibodies to mitochondrial Ags that bind the E2 subunit of the pyruvate dehydrogenase complex (PDC-E2). Histopathological lesions in PBC patients are characterized by mixed lymphoid/mononuclear portal infiltrates and immune-mediated destruction of intrahepatic small bile ducts (17, 18). Whereas inflammatory cytokines including IFN-γ and IL-17 are enhanced, immunomodulatory ones such as IL-10 are reduced (19–26).
Although the stimuli that trigger autoreactivity are often uncharacterized, recent studies indicate that undetected bacterial or viral infections may underlie various forms of autoimmunity (27–36). We have previously shown that patients with PBC exhibit chronic immune reactions against lipoylated proteins of Novosphingobium aromaticivorans (37–41), and we have established a model in which infection of mice with N. aromaticivorans induces anti–PDC-E2 IgG responses and liver lesions resembling PBC in humans (42). Studies in this model suggest that the specific development of lesions in the liver during the more acute phase result from: 1) a greater persistence of N. aromaticivorans in the liver than in other organs; and 2) the activation of NKT cells, which are particularly abundant in the liver of mice, by N. aromaticivorans glycosphingolipids (GSLs) presented by dendritic cells (DCs). Once established, liver disease can be adoptively transferred by CD4+ and CD8+ T cells from wild-type, but not NKT-deficient, mice in the chronic phase. This illustrates the importance of early microbial activation of NKT cells in initiating autonomous, organ-specific autoimmunity and in propagating autoreactive T cells, at least on the C57BL/6 (B6) genetic background. Although we have begun to identify the environmental agent and cellular mechanisms that induce liver-specific immunopathology, the factors underlying genetic susceptibility to this model of PBC remain unknown. Our previous studies suggested that NOD.B6 insulin-dependent diabetes susceptibility region 10 (Idd10)/Idd18 mice, which are partially protected from T1D as compared with the NOD parental strain (43, 44), develop more severe liver lesions following N. aromaticivorans infection than NOD mice (42). Therefore, we reasoned that the Idd10/18 region might harbor susceptibility alleles that determine the severity of bile duct disease upon infection with N. aromaticivorans.
In this study, we show that the B6-derived Idd10 region present in NOD.B6 Idd10 mice primarily accounts for the increased PBC risk of NOD.B6 Idd10/18 congenic mice upon infection and correlates with genotype-dependent protein expression of Cd101, the primary T1D candidate gene of the Idd10 region (44–47). Analyses of additional Idd10 NOD congenic strains, NOD.A/J Idd10 and NOD.CAST Idd10, confirm the correlation of differential control of T1D (44, 47) and PBC with Cd101 gene expression. A newly developed CD101-deficient B6 knockout strain develops PBC with similar severity to that observed in NOD.B6 Idd10 mice and significantly more severe liver disease than parental NOD and B6 mice. Enhanced suppression of CD101 protein expression on DCs, macrophages, and granulocytes from mice carrying the B6 Idd10 allele correlates with prolonged N. aromaticivorans persistence, hyperreactive T cell responses, and extensive liver autoimmunity, indicating that the allelic variations within the NOD and B6 alleles of Cd101 are likely the causal genetic variants for susceptibility and resistance to PBC, defining Cd101 as the first susceptibility gene for infection-induced PBC. The differential regulation of CD101 in various cell types within tissues undergoing distinct immune challenges could explain how Idd10 alleles that protect from T1D increase PBC susceptibility.
Materials and Methods
B6 CD45.1, B6 CD101−/−, B6 CD1d−/−, NOD, NOD.B6 Ptprc CD45.2 (line 6908) (48), NOD.B6 Idd10/18 (lines 1101 and 7754, both having the same Idd10/18 segment) (49, 50), NOD.B6 Idd10 (line 3538), NOD.B6 Idd18 (line 3539), NOD.A/J Idd10, and NOD.CAST Idd10 mice were maintained in our laboratories. Additional information about the NOD congenic strains can be obtained at http://www.t1dbase.org/page/DrawStrains. NOD congenic strains having line numbers were obtained from the Emerging Models program at Taconic Farms (Germantown, NY). B6, B6 MHCII−/−, B6β2m−/−, and B6 IFN-γR−/− mice were purchased from The Jackson Laboratory (Bar Harbor, ME); B6 IL-17R−/− mice were obtained from Amgen (Thousand Oaks, CA). The generation of the CD101-knockout B6, NOD.B6 Idd10 (line 3538), NOD.A/J Idd10, and NOD.CAST Idd10 strains is detailed in the companion article (47). NOD.B6 Idd18 (line 3539) N16 mice were developed by selectively breeding a mouse having a recombination event in the distal region of the Idd10/18 congenic segment present in the NOD.B6 Idd3 Idd10 Idd18 (line 1538) strain (49) that was discovered when line 1538 was being backcrossed to the NOD strain. The T1D phenotype of line 3539 will be detailed in a separate publication.
All mice were raised in a specific pathogen-free environment according to the Institutional Animal Care and Use Committee guidelines.
Bacterial strains and live infection experiments
N. aromaticivorans (ATCC 700278; American Type Culture Collection) was grown overnight in Luria-Bertani broth, inoculated in fresh medium, grown for 8 h at 37°C to an OD of 0.5 at 600 nm, washed, and resuspended in sterile PBS. A total of 100 μl bacterial suspension containing 5 × 107 N. aromaticivorans or PBS only (control) was injected i.v. into 4–7-wk-old mice on days 0 and 14. At indicated time points postinfection, bacterial counts were performed after tissue homogenization in 0.5% Triton X-100 and cultured for colony formation.
Cell preparation and coculture assays
Bone marrow-derived mouse macrophages and DCs were collected after 7 d of culture in RPMI 1640 medium supplemented with glutamine, antibiotics, and 10% FCS with 2 ng/ml recombinant mouse M-CSF (macrophages) and 2 ng/ml GM-CSF (DCs) (all from R&D Systems, Minneapolis, MN). Splenocytes and liver lymphocytes were prepared as described (42, 51), and 250,000 of the specified APCs were cocultured with 250,000 lymphocytes. Cell-culture supernatants were assayed after exposure to the indicated stimuli for cytokine release by ELISA (R&D Systems).
Gentamicin protection assay
Bone marrow-derived DCs from wild-type B6 and CD101−/− mice were exposed to N. aromaticivorans at a ratio of 1:50 for 12 h. After rinsing the cultures three times in PBS (Life Technologies-BRL) to remove nonadherent bacteria, cells were incubated for 1 h in fresh prewarmed medium containing 100 μg/ml gentamicin (Life Technologies-BRL) to kill extracellular bacteria. Gentamicin was removed, and the cells were gently rinsed three times in PBS. DCs were lifted and disrupted by incubating them for 3 min in PBS containing 2 mM EDTA and 0.5% saponin (Sigma-Aldrich), followed by transfer to Eppendorf tubes and high-speed vortexing for 30 s. Viable intracellular bacteria were quantitated by gentle lysis of the DCs and subsequent plating on Luria-Bertani agar.
Real-time quantitative PCR analysis for N. aromaticivorans copy number
Total DNA was extracted from different organs using the DNeasy tissue extraction kit (Qiagen, Valencia, CA). The following primers were used for detection of N. aromaticivorans: forward: 5′-TCCGAGTGTAGAGGTGAAAT-3′ and reverse: 5′-CGTCAATACTTGTCCAGTCA-3′. Real-time quantitative PCR (qPCR) was performed with RT2 qPCR SYBR Green Master Mix (SA Biosciences, Frederick, MD) on a Bio-Rad iCycler (Bio-Rad). A 25 μl reaction volume was used for each sample analyzed, using 12.5 μl RT2 SYBR Green mix, 1 μl each 10 μM primer (400 nM final concentration), 4 μl DNA template (100 ng/μl), and 6.5 μl nuclease-free water. N. aromaticivorans copy number per 400 ng total DNA isolated from each tissue was calculated by reference to a standard of pure N. aromaticivorans genomic DNA (700278D-5; American Type Culture Collection). PCR products were analyzed by melting curve analysis and agarose gel electrophoresis to confirm that the amplicons were N. aromaticivorans specific.
Tissue collection and RNA preparation
Spleen and liver samples from uninfected and infected mice were harvested at the indicated time points and immediately immersed in RNAlater solution (Ambion/Applied Biosystems, Austin, TX) to prevent RNA degradation. Total RNA was isolated using the RNAqueous 4-PCR kit (Ambion/Applied Biosystems) following the manufacturer’s instructions. Isolated RNA was treated with DNase I (Ambion/Applied Biosystems) to eliminate DNA contamination during the subsequent PCR steps. The concentration and purity of the isolated RNA was determined using a Nanodrop-1000 spectrophotometer (Thermo Scientific, Waltham, MA).
cDNA synthesis and real-time qPCR for mRNA expression
cDNA was synthesized using a first-strand cDNA synthesis kit for real-time PCR (Roche, Indianapolis, IN) following the manufacturer’s instructions, with slight modifications where necessary. Briefly, 1 μg total RNA and 2 μl random hexamers were incubated at 65°C for 5 min to remove secondary structures in the RNA template and enhance primer annealing, then added to the reverse-transcription reaction mix. The reaction was carried out at 25°C for 10 min (annealing) followed by a 1-h reverse transcription step at 42°C. Finally, the reverse transcriptase enzyme was heat-inactivated at 99°C for 10 min. Quantitative real-time PCR was performed as mentioned before. The primers used for the analysis were as follows: CCL-21c (forward: 5′-ACCCAAGGCAGTGATGGA-3′; reverse: 5′-TCCGGGGTAAGAACAGGATT-3′); IL-17A (forward: 5′-CTCCAGAAGGCCCTCAGACTAC-3′; reverse: 5′-AGCTTTCCCTCCGCATTGACACAG-3′), and IFN-γ (forward: 5′-TCAAGTGGCATAGATGTGGAAGAA-3′; reverse: 5′-TGGCTCTGCAGGATTTTCATG-3′). The PCR products were analyzed by melting curve analysis and also by agarose gel electrophoresis for a single target-specific amplicon (results not shown). mRNA of hypoxanthine guanine phosphoribosyl transferase (HGPRT) were used for normalization (forward: 5′-ACCTCTCGAAGTGTTGGATA-3′, reverse: 5′-CAACAACAAACTTGTCTGGA-3′).
Flow cytometry and intracellular cytokine staining
CD1d-lipid tetramers were prepared as described (52). Anti-Nkp46, -CD4, -CD8, -TCRβ, -CD25, -γ/δ TCR, -CD69, -CD44, -CD45.1, -CD45.2, -CD11b, -CD11c, -F4/80, and -MHC class II (MHC II) Abs were purchased from eBioscience (San Diego, CA). Regulatory T cells (Tregs) and intracellular cytokines were analyzed by using a staining kit from eBioscience following the manufacturer’s instructions. Anti-CD101 was obtained from R&D Systems. Cells were analyzed on an LSR II (BD Biosciences, San Diego, CA) with FlowJo software (Tree Star).
Bone marrow radiation chimeras
For the preparation of mixed bone marrow radiation chimeras, a mixture of 5 × 106 CD45.2+ NOD.B6 Ptprc and 5 × 106 CD45.1+ NOD.B6 Idd10 bone marrow cells was injected i.v. into 7–12-wk-old CD45.2+ NOD.B6 Ptprc that had been irradiated twice with a cesium source (Gammacell 40; Nordion International, Ontario, Canada) 1 d before within a 4-h interval (first dose 700 rad, second dose 435 rad). For the preparation of reciprocal bone marrow radiation chimeras,either 5 × 106 CD45.2+ NOD.B6 Ptprc or 5 × 106 CD45.1+ NOD.B6 Idd10 bone marrow cells were injected i.v. into 7–12-wk-old (NOD.B6 Ptprc × NOD.B6 Idd10) F1 mice that had been irradiated twice with a cesium source (Gammacell 40; Nordion International) 1 d before within a 4-h interval (first dose 700 rad, second dose 435 rad). Similarly, B6 CD45.1+/B6 CD101−/− mixed and reciprocal bone marrow chimeras have been prepared. The reconstitution of the mixed bone marrow chimeras with the different cell populations was determined 6–8 wk after bone marrow injection by FACS analysis of blood samples using CD45.1 and CD45.2 Abs and the respective surface markers. The chimeras were i.v. injected with 5 × 107 N. aromaticivorans at days 0 and 14 and their livers and spleens analyzed at the indicated time points by flow cytometry.
Cell transfer experiments
A total of 1 × 107 TCRβ+, 1 × 107 NKp46+, or 1 × 107 γ/δ TCR+ cells mixed from spleen and liver of diseased (week 10 postinfection) B6 or NOD.B6 Idd10 mice were injected i.v. into syngeneic adult irradiated NOD.B6 Idd10, B6, B6 CD1d−/−, B6 MHCII−/−, B6β2m−/−, B6 IFN-γR−/−, and B6 IL-17R−/− recipients, respectively. Livers were harvested 6 wk after transfer for the histopathological analysis of inflammatory bile duct lesions.
Diabetes frequency studies
Diabetes frequency studies were conducted using female mice. The presence of T1D was tested every 10 to 14 d beginning at ∼80 d of age by the detection of urinary glucose >500 mg/dl using Diastix (Miles, Elkhart, IN). Kaplan-Meier survival curves were plotted for each mouse strain, and these were compared using the log rank test (Prism4 software; GraphPad).
Liver tissue was fixed in 10% buffered formalin, embedded in paraffin, and cut into 5-μm sections. Liver sections were deparaffinized, stained with H&E by the Cincinnati Children’s Hospital Medical Center’s Histology Research laboratories, and evaluated microscopically in double-blind studies for leukocytic and lymphocytic infiltration as described before (42).
Statistical significance of the data was analyzed by one-way ANOVA, Student t test, or Mann–Whitney U test as indicated in the respective experiments. A sample size of at least three (n = 3) was used for each sample group in a given experiment, and a p value <5% (*p < 0.05) or <1% (**p < 0.01) was considered significant to accept the alternate hypothesis.
The B6 Cd101 allele is associated with severe autoimmune liver disease and prolonged bacterial persistence on the NOD background upon infection with N. aromaticivorans
Our previous study characterized NOD.B6 Idd10/18 congenic mice as particularly susceptible to autoimmune liver disease upon N. aromaticivorans infection (42). To identify the PBC-susceptibility genes within the Idd10/18 congenic genetic region, in this paper, we study NOD.B6 congenic mice having only the Idd10 or the Idd18 T1D-protective regions (Supplemental Fig. 1).
Infection with 5 × 107 N. aromaticivorans CFUs triggered similar massive liver pathology in NOD.B6 Idd10 as in NOD.B6 Idd10/18 mice (Fig. 1A). Although considerably more prominent than in parental NOD mice, liver lesions remained significantly less severe in NOD.B6 Idd18 compared with NOD.B6 Idd10 and NOD.B6 Idd10/18 mice (Fig. 1A, Supplemental Fig. 2A), suggesting a more critical role for Idd10 in liver disease progression. Infection of NOD × NOD.B6 Idd10 F1 mice also induced significantly more severe liver disease compared with parental NOD mice, suggesting that allelic variations within the Idd10 region drive susceptibility to PBC in a gene dose-dependent manner (not shown). As introgression of the B6 Idd10 allele protects from spontaneous type 1 diabetes (T1D) (43, 44, 47), we tested the effect of N. aromaticivorans infection on T1D in NOD.B6 Idd10 mice (Fig. 1B). N. aromaticivorans infection did not alter the B6 Idd10 allele-mediated protection from T1D.
As Cd101 is a prime candidate T1D gene in the Idd10 region (44–46) and haplotype analysis comparing NOD, NOD.B6 Idd10, NOD.A/J Idd10, and NOD.CAST Idd10 mice demonstrated a correlation between protection from spontaneous T1D and variation in CD101 expression (47), we tested the severity of liver lesions in NOD.A/J Idd10 and NOD.CAST Idd10 mice, which have T1D-protective and T1D-susceptible Idd10 haplotypes, respectively. NOD.A/J Idd10 mice developed severe liver disease indistinguishable from that of NOD.B6 Idd10 mice, whereas liver disease remained comparably mild in NOD and NOD.CAST Idd10 mice (Fig. 1C). As the Cd101 allele in NOD.A/J Idd10 mice is identical to the Cd101 allele in NOD.B6 Idd10 mice, whereas other Idd10 genes have sequence differences between the A/J and B6 Idd10 haplotypes (47), we concluded that genetic variations within Cd101 were likely the causal genetic event in triggering severe PBC during infection. If Cd101 is the Idd10 gene responsible for altering susceptibility to PBC, we would predict that a mouse having a disrupted CD101 gene would have altered disease susceptibility. Therefore, we compared the severity of liver lesions postinfection in heterozygous and homozygous B6 CD101-knockout and wild-type mice. Wild-type, heterozygous, and homozygous CD101−/− offspring were born in the expected Mendelian ratios and uninfected, homozygous CD101−/− mice appear healthy for >18 mo without developing any spontaneous infiltration at any organ site. Upon infection, however, B6 CD101−/− mice developed more severe liver disease compared with wild-type B6 mice (Fig. 1D). B6 CD101+/− littermates exhibited also more severe PBC than B6 wild-type mice indicating that a 50% reduction of CD101 expression already impacts disease (Fig. 1D, Supplemental Fig. 2B). Severity of liver disease in infected CD101+/− and CD101−/− mice in an F2 cross was comparable to F1 and F0 mice (not shown) excluding any influence of unknown, randomly segregating genes in the B6 background of the CD101−/− strain on PBC pathogenesis. Collectively, these results strongly support allelic variations within Cd101 as the major causal genetic events for infection-induced PBC within the Idd10 region, which contains six other protein coding genes (44, 47).
As persistence of N. aromaticivorans is absolutely required for the induction of liver lesions and the long-term evolution to autoimmunity (42), we tested if introgression of the B6 Cd101 allele influences the course of N. aromaticivorans infection and subsequently impacts the outcome of disease. As observed previously (42), we detected the preferential accumulation of N. aromaticivorans in the livers of infected mice compared with other organs. N. aromaticivorans was observed at much lower levels in the spleen and kidneys, but was completely absent from the pancreas and the heart (Fig. 1E). Although bacterial persistence was prolonged in the livers of NOD.B6 Idd10 and NOD.B6 Idd10/18 congenic compared with parental NOD mice (Fig. 1F, left panel), bacterial infection was cleared 2 wk later (Fig. 1F, right panel). Similarly, bacterial infection was more slowly cleared from the livers of B6 CD101−/− compared with B6 wild-type mice (Supplemental Fig. 2C). As clearance of bacteria was delayed in NOD.B6 Idd10 mice compared with B6 mice correlating with substantially enhanced liver pathology, we concluded that the increased susceptibility to autoimmune liver disease in NOD.B6 Idd10 mice requires both non-Idd10 NOD alleles and the B6 Idd10 haplotype.
Overall, these results point to the B6 allele at Idd10 as being the primary liver disease-promoting subregion within the larger Idd10/18 region upon infection; increased liver pathology may be associated with the prolonged bacterial persistence in the livers of NOD mice with the B6 and A/J Idd10 haplotypes compared with NOD mice with the NOD and CAST haplotypes.
N. aromaticivorans infection does not alter CD101 expression on T cells, but enhances their activation status in the liver
Expression of CD101 has been described in 25–30% of CD4+Foxp3+ Tregs, ∼10–20% of conventional CD4+ and CD8+ memory T cells, and few naive CD4+ and CD8+ T cells from splenocytes of B6 mice (53). In contrast to the spleen, where most of the CD101+ T cells express Foxp3, most of the CD101+ T cells in the liver do not express Foxp3 (Fig. 2A). As the presence of the B6 CD101 allele as part of the B6 haplotype spanning the Idd10 region is associated not only with protection from T1D, but also with more severe PBC upon exposure of mice to N. aromaticivorans, we evaluated the distribution of the different T cell populations and the expression of CD101 expression on these different T cell populations in the liver by flow cytometry (Fig. 2B–D) upon infection, the organ site where N. aromaticivorans preferentially persists (42) (Fig. 1E, 1F). T cell numbers, especially in NOD.B6 Idd10 mice, significantly increased at day 18 postinfection (Fig. 2B, left panel). Foxp3+ Tregs were only found in the livers of infected mice, preferentially NOD mice, whereas in uninfected mice they were rarely detected (Fig. 2B, right panel). B6 CD101−/− mice recruited significantly fewer hepatic Tregs than B6 mice upon infection, but significantly more T cells (Fig. 2C); thus, changes on the B6 CD101−/− background closely resemble the phenotype observed in NOD.B6 Idd10 mice. NKT cells from neither naive nor infected mice expressed CD101 at any time in the liver (Fig. 2D). Infection did not change the percentage of CD101-expressing cells in any of these T cell populations (Fig. 2D) or the mean of CD101 expression on these T cells (Supplemental Fig. 3A). Most of the T cells that expressed CD69, IFN-γ, or IL-17 were CD101 negative (Fig. 2D, Supplemental Fig. 3C). Few CD44+ T cells in the livers of NOD.B6 Idd10 mice expressed CD101 (Supplemental Fig. 3D); however, these T cells expressed CCR7 (Fig. 2E). CCR7 binds to CCL21 (see also Fig. 3E), a chemokine expressed during lymphoneogenesis and in inflamed portal tracts (54) that has been implicated in the pathogenesis of PBC in humans (55). CXCR3, which recruits T cells with regulatory functions upon NKT cell activation to the liver (56), was preferentially expressed on T cells from NOD compared with NOD.B6 Idd10 mice (Fig. 2E). Livers of B6 CD101−/− mice revealed higher numbers of CCR7+ and lower percentages of CXCR3+ T cells compared with wild-type mice (not shown), suggesting that the genetic deletion of CD101 or the introgression of the B6 Cd101 allele on the NOD background affects the upregulation of different chemokines within the liver environment and subsequently the recruitment of different T cell populations. Albeit these data indicate that expression of the B6 Cd101 allele or genetic deletion of CD101 affects T cell activation, this is an indirect process regulated by DCs as outlined later (Figs. 3, 4) in this study.
MHC class I- and MHC II-restricted T cells that produce IFN-γ and IL-17 are required for the adoptive transfer of liver lesions
As we previously published, chronic liver inflammation in the absence of long-term microbial persistence and the adoptive transfer of disease by TCRα/β+ CD4+ and CD8+ T cells suggest an autoimmune etiology of the liver disease (42). To identify critical Ag-presenting molecules and cytokines necessary for the transfer of liver lesions by T cells, we transferred 1 × 107 TCRβ+ liver and spleen cells from diseased B6 mice into irradiated recipients that cannot present Ags via CD1d, MHC class I, or MHC II or do not respond to cytokines such as IFN-γ or IL-17 (Fig. 5A) 10 wk postinfection when all persisting bacteria have been cleared (Fig. 1F, right panel, Supplemental Fig. 2C). Notably, IFN-γ and IL-17 were critical for the adoptive transfer of liver disease (Fig. 5A). The requirement for Ag presentation by MHC class I and II, but not CD1d, confirmed the need for CD4+ as well as CD8+ T cells for the transfer of liver disease (Fig. 5A) (42). Although γ/δ+ T cells and NK cells accumulated in the livers of diseased mice (not shown), these cells did not adoptively transfer liver lesions (Fig. 5B).
Collectively, our data suggest that the dominant IFN-γ and IL-17 cytokine profile correlates with the induction of severe liver disease and promotes autoreactive CD4+ and CD8+ T cells that are responsible for the adoptive transfer of liver lesions.
N. aromaticivorans infection suppresses CD101 expression on granulocytes, DCs, and macrophages in the liver
CD101 is not only expressed on subsets of T cells, but also on macrophages, DCs, and granulocytes and has been implicated in the maintenance of peripheral tolerance in humans (53, 57–59). As N. aromaticivorans localizes to the liver in our infection-induced mouse model (Fig. 1E, 1F) (42) and as CD101 expression is not changed on hepatic T cells postinfection (Fig. 2D, Supplemental Fig. 3A), we focused on these other immune cell subsets in the hepatic compartment that constitute the majority of CD101-expressing cells in the liver (Fig. 2A). To evaluate if N. aromaticivorans infection affects CD101 expression on these cell populations, we analyzed mononuclear cells from the liver of NOD and NOD.B6 Idd10 mice. Around 40–70% of Gr1+, CD11c+, and F4/80+ cells expressed CD101 in naive, uninfected NOD and NOD.B6 Idd10 mice to similar levels and frequencies (Fig. 3A–C). Absolute cell numbers and the relative distribution of cell populations were comparable between the analyzed mouse strains (not shown). All DC subpopulations, including CD8+, B220+, and CD11b+, increased upon infection without a significant change in the proportion of the subpopulations (Supplemental Fig. 4A). CD101 was preferentially expressed on CD11b+CD11c+ myeloid DCs in the spleen, whereas CD101-positive DCs also coexpressed B220 in the liver (Supplemental Fig. 4B); very few CD8+ lymphoid DCs (CD8+CD11c+) expressed CD101 in either tissue (Supplemental Fig. 4B).
Surface expression of CD101 (Supplemental Fig. 3B) and the number of CD101-positive cells were reduced on Gr1+ (Fig. 3A), CD11c+ (Fig. 3B), and F4/80+ cells upon infection (Fig. 3C). Although the mean fluorescence intensity (MFI) on Gr1highCD101high in the livers of uninfected NOD.B6 Idd10 was higher than in uninfected NOD mice, as observed in the spleen (47), Gr1+ cells from NOD.B6 Idd10 mice had reduced CD101 levels compared with cells from parental NOD mice postinfection (Fig. 3A); a similar enhanced CD101 suppression was also observed on DCs and macrophages from NOD.B6 Idd10 mice (Fig. 3B, 3C). CD101 levels on DCs did not return to baseline levels up to 8 wk postinfection in the liver (Supplemental Fig. 5A). However, infection did not influence the distribution of CD101 expression (Supplemental Fig. 5B) on DCs in pancreatic lymph nodes, whereas CD101 expression was only transiently reduced in the spleen returning to baseline levels ∼3 to 4 wk postinfection (Supplemental Fig. 5C and data not shown). These data suggest that suppression of CD101 expression correlates with the presence/persistence of N. aromaticivorans at these different organ sites (Fig. 1E, 1F). Accordingly, increased copy numbers of N. aromaticivorans were preferentially detected by a 16S rRNA-specific qPCR in cell-sorted, purified, CD101-negative DCs, whereas CD101-positive DCs contained, in general, less N. aromaticivorans rRNA (Supplemental Fig. 5D). As CD101 deficiency did not alter the rate of infection (Supplemental Fig. 5E), these data suggest that either infection suppresses CD101 expression on APCs or APCs downregulate the negative costimulatory molecule CD101 upon bacterial contact. Whereas greater numbers of granulocytes accumulated in the livers of NOD compared with NOD.B6 Idd10 mice, a reduced influx of DCs into the livers of NOD mice was observed (Fig. 3D). Similarly, reduced numbers of granulocytes accumulated in the livers of infected B6 CD101−/− mice as compared with B6 mice, whereas DCs were increased (Fig. 3D). CD101 may not only have a negative costimulatory role (53), but also may be involved in the recruitment or containment of certain cell populations within a particular tissue. DCs and neutrophils represent resident cell populations in the liver (Fig. 3D); DC and granulocyte numbers increase upon exposure to N. aromaticivorans, suggesting that both are recruited to the liver upon infection (Fig. 3D). Infection of neutrophils occurs during the systemic response to N. aromaticivorans and subsequent downregulation of CD101 may impair their recruitment to the liver, which is exacerbated in NOD.B6 Idd10 and B6 CD101−/− mice. Alternatively, CD101-negative granulocytes may be recruited preferentially to the livers of NOD.B6 Idd10 and B6 CD101−/− mice responding to the increased expression of chemotactic factors and/or inflammatory cytokines in these mice. Because the survival time of DCs is longer than that of granulocytes, the magnitude of CD101 expression on DCs is lower than on granulocytes, and DC numbers are not reduced in the livers of CD101-deficient mice without infection, subsequent upregulation of chemotactic factors by infection may be contributing to the accumulation of DCs at the site of inflammation despite CD101 deficiency. We found that CCL21 mRNA expression was most prominently enhanced in the livers, but not in the spleens, of NOD.B6 Idd10 mice (Fig. 3E, 3F) compared with parental NOD mice. The increased expression of CCL21, the ligand for CCR7, may be responsible for the preferential recruitment of CCR7-expressing T cells to the infected and inflamed livers of NOD.B6 Idd10 and B6 CD101−/− mice (Fig. 2E). In summary, the suppression of CD101 on APCs from NOD.B6 Idd10 mice upon infection is correlated with an inflammatory profile in the liver environment that resembles the phenotype in B6 CD101−/− mice.
Cell-intrinsic differences in CD101 specific for the genetic background determine the recruitment of granulocytes and DCs to the infected liver
To determine if Idd10/Cd101 gene haplotype-dependent expression differences are mediated in a cell-intrinsic manner or rather secondary events, we measured variations in the accumulation of different cell populations in N. aromaticivorans-infected NOD.B6 Ptprc CD45.2+/NOD.B6 Idd10 as well as B6 CD45.1+/B6 CD101-deficient mixed bone marrow chimeras. These mixed bone marrow chimeras expressed similar numbers of CD45.1- and CD45.2-positive cells as determined by FACS analysis (55.3 ± 4.9 versus 44.7 ± 6.1 and 49.7 ± 8.8 versus 50.3 ± 9.1, respectively) in the blood. We found that the altered accumulation of granulocytes by genotype (Fig. 3A) is cell intrinsic, as CD45.2-positive granulocytes originating from NOD.B6 Ptprc mice accumulated to greater numbers in the livers of infected mixed bone marrow chimeras than granulocytes from NOD.B6 Idd10 mice (Fig. 6A). Similar results were obtained with reciprocal bone marrow chimeras (not shown). The increased influx of Gr1+ cells expressing the NOD Idd10 allele correlated with increased CD101 expression on these Gr1+ cells as compared with CD101 levels on Gr1+ cells from NOD.B6 Idd10 mice (Supplemental Fig. 6A). Accumulation of granulocytes derived from CD101-deficient B6 bone marrow was also reduced as compared with those from the wild-type B6 bone marrow (Fig. 6B); these data strongly suggest that the intrinsic reduction of CD101 on NOD.B6 Idd10 granulocytes upon infection leads to the reduced accumulation of granulocytes in the liver. The differential accumulation of DCs in the liver following infection (Fig. 6C, 6D) was also cell intrinsic: introgression of the B6 Idd10 region on the NOD background and CD101 deficiency on the B6 background both enhanced the accumulation of DCs in the respective mixed bone marrow chimeras (Fig. 6C, 6D). As an intrinsic consequence of reduced CD101 expression, DCs of NOD.B6 Idd10 and CD101-deficient B6 origin expressed higher levels of MHC class II than parental NOD and B6 mice (Fig. 6E, 6F).
Overall, CD101 expression influences in a cell-intrinsic manner the accumulation of granulocytes and DCs in the infected liver and the upregulation of MHC class II on the accumulated DCs.
Suppression of CD101 on DCs triggers an IFN-γ– and IL-17–dominated cytokine profile
Interactions with APCs are required for T cell activation. Even though conventional T cells express small amounts of CD1d in the periphery, they cannot present GSL Ag directly to NKT cells (60–63) in contrast to APCs. Therefore, conventional T cell activation by NKT cells responding to GSL Ags is an indirect process that requires NKT cell stimulation via APCs. The expression of costimulatory molecules by APCs that are induced by productive interactions with NKT and T cells and the local cytokine milieu are critical for the subsequent induction/suppression of T cell responses (60, 62). Because CD101 has been described as a negative costimulatory molecule (58, 64, 65), it is possible that CD101 contributes to the ability of APCs to activate the T cell response. To evaluate if infection-induced suppression of CD101 on APCs contributes to the overzealous T cell response in infected mice, bone marrow-derived DCs from NOD and NOD.B6 Idd10 mice were left untreated or exposed to N. aromaticivorans for 48 h (Fig. 4A). Untreated cells showed no major differences in the distribution of CD101 expression. However, as observed in vivo for liver CD11c+ cells (Fig. 3B), CD101 expression was suppressed on bone marrow-derived DCs (Fig. 4A), but not T cells (not shown) when exposed to N. aromaticivorans. This suppression of CD101 expression was more prominent in NOD.B6 Idd10 and NOD.A/J Idd10 mice than in NOD and NOD.CAST Idd10 mice. Similarly, CD101 expression was suppressed on CD11c+ and Gr1+ cells in liver cell preparations 1 d after exposure to N. aromaticivorans in vitro, whereas CD101 expression on T cells in the same cell preparations remained unaffected (Fig. 4B, Supplemental Fig. 6B). As observed in vivo (Fig. 6E, 6F), MHC II expression was increased preferentially on DCs of NOD.B6 Idd10 or NOD.A/J Idd10 origin compared with DCs from NOD or NOD.CAST Idd10 mice (Fig. 4C). A comparable increase in MHC II expression was detected on DCs from CD101-deficient mice compared with DCs from B6 mice (81.1 ± 7.8 versus 44.3 ± 6.6). Cocultures of N. aromaticivorans-pulsed DCs with liver lymphocytes (containing NKT and T cells) from infected mice demonstrated that the secretion of IFN-γ and IL-17 was enhanced in a NOD.B6 Idd10 DC-dependent manner (Fig. 4D). Similarly, cocultures of CD101-deficient DCs pulsed with GSL-1, an NKT cell ligand expressed in the cell wall of N. aromaticivorans, and T/NKT cells from uninfected mice induced enhanced IFN-γ and IL-17 responses compared with cocultures with wild-type DCs (Fig. 4E). These results suggest that infection-mediated reduced CD101 expression is not required for NKT cell-dependent DC activation; however, the lack of CD101 or the suppression of CD101 expression on APCs is sufficient to drive the overzealous Th1- and Th17-dominated conventional T cell response. mRNA copy numbers of IFN-γ and IL-17 in livers from infected CD101-deficient B6 (Fig. 4F) and NOD.B6 Idd10 mice (Fig. 4G) were increased compared with parental B6 and NOD mice, respectively, suggesting a role of the B6 Cd101 allele in the induction of these Th1/Th17-dominated immune responses on the NOD background (Fig. 7).
Collectively, our data suggest that enhanced IFN-γ and IL-17 responses in NOD.B6 Idd10 and NOD A/J Idd10 mice promote severe liver disease. Reduced CD101 expression on DCs following N. aromaticivorans infection is therefore critical for the induction of these overzealous T cell responses.
Autoimmune diseases often cluster in the same family; this indicates that there are common genetic pathways affecting immunological tolerance mechanisms, although the autoimmune assault may occur in diverse tissue types. Genome-wide association studies have now revealed very significant sharing of mapped susceptibility loci (66–68); in this study, we have shown that a gene region previously shown to influence susceptibility to T1D in the NOD mouse model (44–46, 69), Idd10, also influences an autoimmune disease involving a different tissue, the liver, which follows an infection with N. aromaticivorans (42). Using a CD101-deficient B6 mouse combined with haplotype analysis studies in NOD.CAST Idd10 and NOD.A/J Idd10 mice, our data suggest that allelic variations within Cd101 drive susceptibility for infection-induced PBC. Intravenous infection of mice with N. aromaticivorans leads to the allele-dependent downregulation of CD101 protein expression on DCs, macrophages, and granulocytes in the liver, the preferential site of bacterial persistence. Because downregulation of CD101 is more pronounced in NOD congenic strains that express the B6 Cd101 allele than in parental NOD mice, and the clinical and histopathological phenotypes in these mice resemble those of CD101-deficient B6, we conclude that the allele-specific downregulation of CD101 protein underlies the induction of severe liver autoimmunity in NOD.B6 Idd10 mice (Fig. 7). The substantially enhanced liver disease in these mice is associated with overzealous Th1 and Th17 responses propagated by NKT cells responding with an enhanced release of inflammatory cytokines to GSL Ags of N. aromaticivorans in the absence of CD101 expression on APCs. Whereas genetic deletion of CD101 in the B6 background mimics the phenotype in the NOD.B6 Idd10 congenic strain, the B6 haplotype of CD101 on the NOD background alone is not sufficient for triggering severe liver disease. As we observed greater downregulation of CD101 expression in NOD.B6 Idd10 compared with B6 mice after N. aromaticivorans infection (Fig. 3A–C), it is likely that the combination of multiple loci on the NOD background together with the B6 Idd10 haplotype confers increased susceptibility to infection induced PBC, which likely accounts for the more profound downregulation of CD101 in NOD.B6 Idd10 mice. That downregulation of CD101 is particularly striking in Gr1+ cells: Gr1+ cells in NOD.B6 Idd10 mice start with higher CD101 expression than Gr1+ cells from naive NOD mice, but downregulate CD101 more significantly upon infection (Supplemental Fig. 6A). In contrast, B6 mice exhibit compensatory mechanisms that better control bacterial infection and T cell activation, two major mechanisms that are lost once the CD101 allele is deleted. Because B6 CD101−/− and NOD.B6 Idd10 congenic mice develop overzealous, Th1- and Th17-dominated T cell responses and clear bacterial infection more slowly than parental NOD and B6 mice, and genome-wide association studies in PBC patients identified genes affecting the Th1 response and T cell signaling as PBC risk factors including IL-12 (70), Spi-B, and IFN regulatory factor 5 (71, 72), it is tempting to speculate that bacterial infection triggers inappropriate T cell responses in genetically predisposed individuals. The murine orthologs of IL-12, Spi-B, and IFN regulatory factor 5 may be also the non-Idd10 NOD alleles in NOD.B6 Idd10 mice that interfere with the B6 Cd101 haplotype.
Downregulation of CD101 on granulocytes correlates with their reduced accumulation in the liver, an intrinsic property to this subset, likely contributing to the prolonged persistence of N. aromaticivorans; downregulation of CD101 on DCs is closely associated with the evolution of overzealous, IFN-γ/IL-17–dominated T cell responses (Fig. 7). These latter data confirm a critical role of CD101 as a negative costimulatory molecule in T cell activation (73, 74) and define its action on T cells and T cell homeostasis as APC intrinsic. The reduced accumulation of granulocytes upon infection in the liver and the intrinsic reduction of granulocytes in the bone marrow of CD101-deficient mice (47) suggest an additional role of CD101 in the containment and/or trafficking of granulocytes in/to defined organ sites.
DCs, however, still accumulate in the livers of infected mice despite CD101 downregulation or deficiency. As DCs express lower levels of CD101, survive longer upon infection and are preferentially infected in the liver compared with granulocytes that respond to systemic infection, secondary upregulation of inflammatory molecules during prolonged infection may overcome the lack of CD101 expression as suggested by the liver-specific upregulation of the chemokine CCL21. Alternatively, DCs may not require CD101 expression for trafficking in contrast to granulocytes, and preferentially, CD101-negative DCs may be recruited to the site of inflammation to induce an effector instead of a Treg response. As CCL21 is known to recruit DCs to the site of infection (75, 76) and to be upregulated in portal tracts of chronically inflamed livers (55), induction of CCL21 may be also responsible for the recruitment of CCR7-expressing T cells to the infected and inflamed liver, an effect that may be augmented by CD101 suppression. Downregulation of CD101 on DCs and macrophages occurred upon bacterial contact: intracellular bacteria either suppress CD101 expression or APCs downregulate CD101 upon bacterial encounter. However, because CD101-deficient DCs and macrophages are infected similar to wild-type cells, we exclude a role of CD101 as a receptor involved in bacterial uptake. Alternatively, infection may induce the expression of a less functional CD101 splice variant in NOD.B6 Idd10 mice that is not recognized by our Abs. This may be an allele intrinsic effect or reflect the alternative amino acid sequence of the B6 allotype. As one or more of the variations in the Cd101 allele are causal genetic events in the mouse model and may translate into human disease, future studies need to define if single nucleotide polymorphisms in the Cd101 gene or in genes encoding proteins associated with CD101 are associated with PBC and whether there is a correlation with reduced CD101 expression and an inflammatory cytokine profile.
The correlation of severe pathological phenotypes in NOD.B6 Idd10, NOD.A/J Idd10, and B6 CD101 KO mice, along with the downregulation of CD101 in NOD.B6 Idd10 and NOD.A/J Idd10 mice mimicking the deletion of CD101 in the B6 background, strongly support the hypothesis that Cd101 variation within the Idd10 region is primarily responsible for altering the severity of liver autoimmunity induced by N. aromaticivorans. However, on the NOD background, other genes within the B6-derived Idd10 region may contribute to severe liver disease such as B7h4 (encoded by Vtcn1), known for inhibiting T cell functions and granulocyte responses (77–79). Although a loss of B7h4 function contributed by the B6 Idd10 region may contribute to the overzealous T cell response, it does not explain the reduced recruitment of granulocytes to the livers of NOD.B6 Idd10 mice that is observed in B6 CD101−/− mice. In addition, the development of comparably severe liver disease in B6 CD101−/− and NOD.B6 Idd10 as well as NOD.A/J Idd10 mice that expresses similar, but different, B7h4 alleles strongly support the candidacy of Cd101 within Idd10. Because B7h4 was only very transiently expressed on some macrophages in the spleen on day 1 postinfection (data not shown), the blockade of B7h4 by neutralizing Abs (80) had no impact on the amount of cytokines produced in the coculture system of T cells with DCs independent of the Idd10 alleles expressed (data not shown), and no genotype-dependent differences between NOD and NOD.B6 Idd10 mice in the expression of B7h4 protein were detected, we excluded a major contribution of the B7h4 alleles for influencing the severity of liver disease.
Even though several Idd-encoded molecules might be involved in different autoimmune diseases, the functional consequences of introgression of these alleles into the NOD background may differ from disease to disease and may also change the susceptibility of the host to infectious challenges. Therefore, the mechanisms through which allelic variations cause disease need to be unraveled individually for each disease. This is strikingly reflected by the fact that the B6 Cd101 allele protects from spontaneous T1D (43, 44, 46, 47) but exacerbates the PBC-like autoimmune process induced by N. aromaticivorans infection. However, as introgression of the B6 Idd10 haplotype on the NOD background delays N. aromaticivorans clearance similar to that observed in B6 CD101−/− mice, CD101 should not only be considered as an autoimmune gene, but also an infectious disease gene regulating the persistence of bacterial infection. Not only the distribution of CD101 expression, but also its magnitude may contribute to these tissue-specific effects. Additionally, as cellular composition and total cell numbers differ between organs, even though the percentages of CD101-expressing cells may remain similar, altered cellular interactions may occur due to the absolute increase/decrease in CD101 available for the respective cells. In addition, different cell populations may express CD101 in various tissues; for example, most CD101+ T cells in the livers are CD44+, whereas they express predominantly Foxp3 in the spleen. Based on compelling evidence from various laboratories that development of diabetes in NOD mice is tightly controlled by Tregs (81–85), the interactions of Foxp3+ T cells with DCs likely influence the incidence of spontaneous T1D in our congenic mice. There are also only few Gr1+ myeloid cells or NKT cells found in pancreatic lymph nodes, and there is no clear association of NKT cells with spontaneous T1D (60). In addition, no endogenous GSL Ag influencing the progression of T1D has been described to date. In contrast, we hypothesize that the interplay of DCs with NKT cells that are most abundant in the liver and respond to GSL Ags of N. aromaticivorans (42, 51) propagate an overzealous T cell response. This inflammatory environment is also sustained due to the prolonged persistence of N. aromaticivorans in the liver following the reduced recruitment of Gr1+ cells. These observations and the fact that environmental factors like infection may influence CD101 expression may all contribute to the tissue and/or cell-specific effects of CD101 and determine if a Cd101 allele acts as a protective or susceptibility gene for a particular autoimmune disease. Although the effects of CD101 in promoting severe disease are ultimately T cell-mediated in our PBC model, the role of B cells and the anti–PDC-E2 response remains unclear and will be addressed in a separate study. It also needs to be explored if CD101 has a receptor or if it is involved in homotypic interactions.
Collectively, our data identify CD101 as the likely candidate molecule for N. aromaticivorans-induced PBC and dissect the mechanisms by which N. aromaticivorans infection triggers liver inflammation. The defined diabetes susceptibility locus Idd10 that encodes CD101 as major candidate uncovers a key tool for understanding the circuits of cell–cell interactions in PBC. We confirm in this study not only the negative costimulatory role of CD101 in T cell activation, but also uncover a novel function of CD101 in the containment and/or trafficking of granulocytes to defined organ sites. Although not within the scope of the current study, dissecting the pathways in each autoimmune entity will allow us to understand the regulation of molecules in autoimmunity in general and their complex interplay as diabetes susceptibility loci have been associated with other autoimmune diseases as well. Consequently, our work will help to identify therapeutic targets and approaches that can be used to guide the development of effective therapies for PBC as well as to identify shared targets in autoimmune disease for clinical intervention in the future.
We thank the Digestive Health Center core facility (Cincinnati Children’s Hospital) for the sectioning of histological slides, Jorge Bezzera for critical reading of the manuscript, Kankana Chava for help with the flow cytometry and qPCR analysis, Stacey Burgess and Christina S. Sexton for data collection and help with in vivo experiments, and Patrick S. Leung and Eric M. Gershwin for support with the anti–PDC-E2 ELISA studies. We also thank the National Institutes of Health Tetramer Facility for providing α-galactosylceramide CD1d-tetramers.
This work was supported by National Institutes of Health Grants P01 AI39671 (to L.S.W. and J.A.T.) and R01 DK074768 (to W.M.R. and L.S.W.). L.S.W. and J.A.T. are supported by a joint grant from the Juvenile Diabetes Research Foundation, the Wellcome Trust, and the National Institute for Health Research Biomedical Research Centre. H.I.F. was supported by a Wellcome Trust 4-year studentship. J.M. is supported by the Lupus Research Institute, in part by Public Health Service Grant P30 DK078392, a grant from the University of Cincinnati microbial pathogenesis core center, by Award R01DK084054 from the National Institute of Diabetes and Digestive and Kidney Diseases, the German Research Foundation Deutsche Forschungsgemeinschaft (MA 2621/2-1), and by the Interdisciplinary Center for Clinical Research of the Universitätsklinikum Erlangen (IZKF_JB10_A48). Cambridge Institute for Medical Research received Wellcome Trust Strategic Award 079895. The availability of NOD congenic mice through the Taconic Emerging Models Program was supported by grants from the Merck Genome Research Institute, National Institute of Allergy and Infectious Diseases, and the Juvenile Diabetes Research Foundation.
The online version of this article contains supplemental material.
Abbreviations used in this article:
hypoxanthine guanine phosphoribosyl transferase
mean fluorescence intensity
- MHC II
MHC class II
primary biliary cirrhosis
E2 subunit of the pyruvate dehydrogenase complex
type 1 diabetes
regulatory T cell.
The authors have no financial conflicts of interest.