Associations of microbial and indoleamine-2,3-dioxygenase-derived tryptophan metabolites with immune activation in healthy adults

Healthy adults were recruited into the Nutritional Phenotyping Study conducted at the USDA Western Human Nutrition Research Center (ClinicalTrials.gov: NCT02367287↗) as previously described (19, 20). Recruitment targets in this observational study were balanced by sex, three age categories (18-34, 35-49, 50-66 y) and three BMI categories (18-24, 25-29, 30-44 kg/m²). Data on dietary intake and physical fitness were collected during study visits and fasting blood was collected to measure Trp metabolites and immune function. For the current analysis, data were available from 362 study volunteers. Ethical approval for this study was received from the Institutional Review Board of the University of California, Davis, as previously described (19, 20).

Habitual dietary intake was assessed using the Block 2014 Food Frequency Questionnaire (FFQ) by NutritionQuest (21). Proximal dietary intake of individuals was also assessed from 24-hour recalls collected by the Automated Self-Administered 24-Hour (ASA24) Dietary Assessment tool (22).

Physical fitness was assessed using the YMCA Step Test (23) after excluding volunteers with risk factors that would preclude this sort of physical activity (e.g., back or knee problems) using a physical activity readiness questionnaire from the American College of Sports Medicine (24).

Blood was drawn in the morning after a 12 h overnight fast (water being allowed to maintain hydration) following consumption of a standard meal the evening before (19).

PBMC were isolated from blood (using sodium heparin as an anticoagulant) using a single-step density gradient prepared with Ficoll-Histopaque 1077 (Sigma Aldrich, St. Louis MO, USA).

Previous studies reported that cytomegalovirus (CMV) seropositive and seronegative individuals differ in levels of CD4 and CD8 T-cell subsets, possibly due to repeated reactivation of CMV infection expanding memory and effector/memory T-cell subsets (25, 26). To adjust for this potential difference in our analysis of the correlation of immune biomarkers with Trp metabolites, we assessed infection with CMV by measuring antibody status using fasting sodium heparin plasma analyzed with the CMV IgG Reagent Pack with analysis run on the Roche Diagnostics (Indianapolis, IN) Cobas e411 autoanalyzer according to the manufacturer’s instructions. Results are categorized as positive (“reactive”), borderline, and negative (“not-reactive”). One sample in this study was borderline and was considered negative during statistical analysis.

Freshly isolated PBMC were washed using Hank’s Balanced Salt Solution (Thermo Fisher scientific, Pleasanton Ca, USA) and resuspended in Russ-10 medium. Russ-10 was prepared using a 500 mL bottle of RPMI 1640 (Thermo Fisher Scientific) with the addition of 5 mL

L-Glutamine (Thermo Fisher Scientific, “L-Gluta MAX”), 5 mL of antibiotic/antimycotic (Thermo Fisher Scientific), 5 mL non-essential amino acid mixture (Thermo Fisher Scientific), 5 mL Sodium Pyruvate (100 nM; Thermo Fisher Scientific), 5 mL HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) buffer (Thermo Fisher Scientific), and 0.25 ml of 100 mM beta-Mercaptoethanol (Sigma Aldrich). Russ-10 was used in cell culture containing 10% heat-inactivated fetal bovine serum (Thermo Fisher Scientific). PBMC were cultured with 5% CO2 at 37° C in a humidified incubator at a final concentration of 1 x 10⁶ cells/mL for 24 h with LPS stimulation, or for 48 h with T-cell-specific stimulation in 96-well flat-bottom polystyrene plates (Corning, Corning PA). Supernatants were collected by centrifugation at 300 x g for 10 min at 4°C and immediately frozen at -80°C until analysis of cytokine concentrations. LPS (from E. coli 0111: B4; List Biologicals, Campbell Ca USA) was included at a final concentration of 5 ng/mL, a dose that was determined to be sub-maximal in preliminary experiments with PBMC from healthy adults (data not shown). An equal volume of endotoxin-free water (Sigma Aldrich) was used as a negative control for LPS in parallel cultures. T-cell stimulation was provided by anti-CD3 plus anti-CD28 monoclonal antibodies (Thermo Fisher Scientific, anti-human CD3 mouse isotype IgG2a, and anti-human CD28 isotype IgG1). Isotype control antibodies (Thermo Fisher Scientific; mouse IgG1 and mouse IgG2a) were used as the negative control. Antibodies were diluted to 3 μg/mL (individual concentration) in Phosphate Buffered Saline (PBS; Thermo Fisher Scientific) and used to pre-coat plates (50 μL/well) at 4°C overnight using anti-CD3 plus anti-CD28 monoclonal antibodies for T-cell stimulation and the two isotype controls together as the negative control.

PBMC culture supernatants were analyzed for cytokine concentrations using the MSD U-plex Platform, U-plex Biomarker group 1 (human) Multiplex Assay using a U-plex 5-assay 96 well sector plate (MESO Scale Discovery, Rockville MD USA) with the MSD sector imager 2400 (MESO Scale Discovery) or the Mesoscale SQ 120. PBMC cultures stimulated with anti-CD3 plus anti-CD28 antibodies, or with the corresponding isotype control were diluted 1:10 and 1:2, respectively, for determination of interferon-γ and interleukin (IL)-10, IL-13 and IL-17A concentrations. PBMC cultures stimulated with LPS, or its negative control, were diluted 1:40 and 1:2, respectively, for determination of tumor necrosis-factor (TNF)-α, IL-1β, IL-6 and IL-10 concentrations. Mean concentrations (μg/L) of duplicate wells were used for analysis.

During this four-year recruitment period (June 2015 through July 2019), the CBC analyses were performed using whole blood (treated with EDTA as an anticoagulant) in the UC Davis Health, Department of Pathology and Laboratory Medicine Clinical Laboratory using a Beckman Coulter LH750/780 (prior to October 2016) or a Beckman Coulter DxH800 automated hematology analyzer, with the exception that twelve samples early in the study (prior to August 14, 2015) were analyzed on an Abbott Cell-Dyn 322 analyzer at the WHNRC.

Neopterin concentration (nmol/L) was measured using undiluted sodium heparin plasma using a commercial, competitive enzyme immunoassay (Alpco, BRAHMS GmbH, Salem, NH, USA) according to the manufacturer’s instructions. Myeloperoxidase concentrations (μg/L) were measured using sodium heparin plasma (1:10 dilution) using a commercial ELISA kit (Alpco Immunodiagnostic, Salem, NH, USA) according to the manufacturer’s instructions. Soluble CD14 (sCD14) concentrations (μg/L) were measured in duplicate using sodium heparin plasma (1:400 or 1:600 dilution) using a commercial ELISA kit (Bio-Techne R&D Duoset Systems, Minneapolis, MN USA) according to the manufacturer’s instructions. Plates for these three assays were read on an Agilent BioTek Synergy reader (Santa Clara, CA USA) and data analyzed using the BioTek Gen5 software.

The concentrations of 16 plasma proteins (μg/L) were measured in plasma using MSD assay kits and the MSD sector imager 2400 (MESO Scale Discovery). EDTA plasma was used for 13 proteins (CRP, SAA, ICAM-1, VCAM-1 using the Vplex Vascular Injury Panel 1 with samples diluted 1:1,000; eotaxin, IP-10, MCP-1 and MDC using the Vplex Custom Human Biomarker Chemokine Panel 1 with samples diluted 1:4; TNF-α, IL-1β, IL-6, IL-8 and IL-10 using the Vplex Custom Human Biomarker Proinflammatory Panel 1 with samples diluted 1:2) and sodium heparin plasma for 3 proteins (MMP1, MMP-3 and MMP-9 using the MMP 3-plex Ultrasensitive Kit with samples diluted 1:10). Three levels of lyophilized controls were used on each plate to assess plate-to-plate variation. Mean concentrations (μg/L) of duplicate wells were used for analysis.

Four flow cytometry panels were used for analysis of leukocytes using both isolated PBMC (panels A-C) and whole blood (panel D). PBMC were resuspended in a final volume of 50 μL of Brilliant Stain Buffer (BD Biosciences, San Jose, CA, USA) with one million PBMC for each panel. Then PBMC were stained with the Fixable Viability stain 510 (BD Biosciences) on ice prior to staining with the rest of the antibodies. For panel D, after staining 100 μL whole blood with antibodies, erythrocytes were lysed with BD FACS Lysing solution, and the cell pellet was washed and resuspended in staining Buffer (BD Biosciences, San Jose, Ca, USA) before analyzing sample on BD LSR Fortessa flow cytometry. Antibodies () and gating strategies () for these four panels are shown in supplemental materials: Panel A - naïve and central/effector memory T-cells, with activation markers CD38 and HLA-DR; Panel B - Th1, Th2, Th17 Cells, NK Cells and B Cells; Panel C - total and memory Treg cells with activation markers CD38 and HLA-DR; and Panel D - total T-cells plus monocytes (classic, intermediate and alternative), neutrophils, and eosinophils with activation marker CD11b. Cells were analyzed using an LSRFortessa flow cytometer (BD Biosciences) configured with blue (488nm), red (640nm), violet (405 nm) and UV lasers (355nm). Data were collected using FACSDiva and analyzed using FlowJo version 10.6.1 software (BD Biosciences). 1 1

Plasma Trp metabolites were measured using MxP^® Quant 500 kits (Biocrates Life Sciences AG, Innsbruck, Austria) in the presence of deuterated internal standards as per manufacturer’s instructions. Fasting plasma from 362 participants were randomized onto five plates. A single subaliquot of NIST Standard Reference Material 1950: Metabolites in Human Plasma (Sigma-Aldrich, St Louis, MO) was processed on each plate along with method blanks, and manufacturer provided quality control plasma dilutions and 7pt calibration curves for a subset of metabolites including Trp and Kyn. Trp and Kyn results were corrected by recoveries of deuterated analogs of the respective compounds, while indole, IAA, and IPA were quantified against single point calibrations relative to various deuterated acylcarnitines. All reported metabolites were detected with negative mode ionization on an API 6500 QTRAP (Sciex, Framingham, MA) after chromatographic separations. Data was processed with manual curation of peak integrations using MultiQuant v 3.2 (Sciex). Peak areas were imported into MetIDQ Oxygen (Biocrates Life Sciences AG). Results were corrected for plate specific effects by normalization to the median of four mid-range quality control solutions analyzed on each plate. For the replicate NIST plasma, the coefficients of variation for Trp, Kyn, IAA and IPA across the 5 analyzed plates were 9%, 13%, 40% and 20%, respectively. Indole was not detected in this sample.

Amplification and sequencing of the 16S rRNA V4 – V5 region from bacterial DNA extracted from stool was performed by the Dalhousie University Integrated Microbiome Resource using primers 515F, GTGYCAGCMGCCGCGGTAA, and 926R, CCGYCAATTYMTTTRAGTTT (27–29) as previously described (30). Sequences were analyzed using Qiime2 version 2019.10 (31) also as previously described (30).

Statistical analyses were conducted using SAS 9.4 (SAS Institute, Cary, North Carolina, United States). Correlation analysis between dietary Trp/protein intake and Trp metabolites was calculated using the Spearman test, and a p-value<0.05 was considered statistically significant. Group comparisons of Trp metabolites across age and BMI categories and sex were performed by the nonparametric Kruskal-Wallis test and differences between groups were compared using the Dwass–Steel–Critchlow–Fligner (DSCF) multiple comparison adjustments.

Association of Trp metabolites with the 88 immune biomarkers were characterized using a multiple linear regression model with Trp metabolites as the independent variables and individual groups of immune biomarkers as the dependent variables. Variables were all normalized using rank-based normal transformation prior to analysis. The analysis was controlled for age and BMI groups, and sex. It was also controlled for CMV infection status as CMV infection history confirmed by the reactive IgG level was associated with some immune markers in our study, primarily but not exclusively differences in T-cell measures (data not shown). Spearman correlation analysis between Trp metabolites and YMCA step scores showed that IPA had a significant positive association with the YMCA step scores (data not shown). For this reason, correlation analysis between IPA and the immune variables was also controlled for the YMCA step scores. To account for multiple comparisons, Benjamini-Hochberg adjusted p-values (32) were computed within six groups of immune biomarkers including A. effector/memory T-cells and activation levels (n=24), B. other lymphocytes including Th cells, NK, NK T-cells and B cells (n=11), C. PBMC cytokines(n=8), D. complete blood count (n=15), E. innate cell activation (n=11), and F. plasma markers (n=19). Statistical significance was set at p< 0.05. Heatmaps were plotted to observe the correlation between Trp metabolites and immune markers. Beta values of the standardized regression coefficients estimated by the linear regression, and the corresponding raw and adjusted p-values are documented in Supplemental Table S3. The standardized regression coefficients were used to represent effect size (33). The size of the standardized effects used conventionally are 0.1-0.3 for small effect, 0.3-0.5 for moderate effect, and 0.5-1 for large effect (34, 35).

For the association of Trp metabolites and microbial taxa, the likelihood ratio test in the DESeq2 package (v 1.26.0) (36) from R (v 3.6.3) (37) was used to identify differentially abundant taxa at the family and genus levels between three groups of rank-normalized Trp metabolites. Significant Trp metabolites-microbiota associations were determined based on Benjamini-Hochberg adjusted p-values of<0.05.

Associations of microbial taxa with immune variables were determined using a linear regression analysis performed in SAS. Each model included sex, age and BMI categories, and CMV infection status as covariates, the significant Trp metabolites-associated taxa as the independent variables, and individual groups of immune biomarkers as the dependent variables. Significant associations were determined based on Benjamini-Hochberg adjusted p-values of<0.05. All p-values are documented in the corresponding. 1

Study participants were from the WHNRC Nutritional Phenotyping Study which is a cross-sectional observational trial and included healthy adults, men and women, aged 18–66 y, with BMI (kg/m²) 18–44, and living near Davis, California. The recruitment strategy and basic demographic characteristics of the participants have been described previously (20). Participant characteristics by sex, age, and BMI categories for those included in the present analysis are summarized in Table 1.

Sex, age, and BMI can affect immune function. We thus examined plasma Trp metabolite concentrations using these three categorical variables that were used to balance recruitment by sex, age, and BMI categories prior to examining associations with immune variables. Our study population includes 172 men and 190 women. Compared to women, men had a significantly higher concentration of Trp (p<0.0001), Kyn (p=0.0002) and IAA (p= 0.015) (Figures 1A, B, E; Table S1C). The three age groups were A1:18–34 y (n=129), A2: 35–49 y (n=119), and A3:

50–66 y (n=114). The youngest age group (group 1) had a significantly higher Trp concentration compared to each of the two older groups (group 1 vs 2: p=0.0006, group 1 vs 3: p=0.0001) (Figure 1A; Table S1A). The Kyn/Trp ratio was significantly different across the three age groups, with the ratio increasing with age (group 1 vs 2: p =0.004, group 1 vs 3: p= 0.0001, group 2 vs 3: p=0.034) (Figure 1C; Table S1A). The three BMI groups were B1:18–24.99 kg/m² (n=139), B2: 25.00–29.99 kg/m²(n=131), and B3: 30.00–44 kg/m² (n=92). The group with the lowest BMI, group 1, had a significantly higher IPA concentration compared to the group with the highest BMI, group 3 (p=0.0002) (Figure 1F; Table S1B). Kyn concentration was significantly higher in group 3 vs group 1 (p=0.005) and group 2 (p=0.024) (Figure 1B; Table S1B). Similarly, Kyn/Trp ratio was significantly higher in group 3 vs group 1 (p=0.005) and group 2 (p=0.009) (Figure 1C; Table S1B). Indole concentrations did not differ by sex, age or BMI (Figure 1D; Table S1).

To determine if habitual dietary Trp intake was associated with Trp metabolite concentrations in plasma, we examined the relationship between Trp and total protein intake from FFQs, estimating cohort usual intake over the preceding year, and protein intake from 24-hour recalls collected by the ASA24 by Spearman analysis. These analyses showed that FFQ-based dietary Trp and protein intake, as well as ASA24-based protein intake, were positively associated with plasma Trp (FFQ Trp intake: p=0.003; FFQ protein intake: p=0.002, ASA24 protein intake: p=3.3E-07*) and Kyn (FFQ Trp and protein intake: p=0.02, ASA24 protein intake: p=0.04) concentrations. FFQ- and ASA24-based protein intake were also positively associated with IAA (FFQ protein intake: p=0.04, ASA24 protein intake: p=0.01) (). 1

We used multiple linear regression analysis to examine the relationships between plasma Trp metabolites and 88 immune variables in six categories (): (A) effector and memory T-cell subsets and their level of activation; (B) other lymphocytes including Th1, Th2, Th17, natural killer T (NKT), NK, and B cells; (C) cytokines produced by PBMC stimulated with LPS or with anti-CD3 plus anti-CD28 antibodies to activate T-cells; (D) standard complete blood count with differential; (E) innate cells types and activation level of these types, and; and (F) plasma cytokines, chemokines, acute phase proteins, matrix metalloproteinases and other markers of immune activation. 1

The microbial metabolites of Trp included in this analysis were indole, IAA and IPA. Plasma indole was significantly associated with three, IAA with six, and IPA with two immune variables though none of these associations remained significant after adjustment for multiple comparisons (Figure 2; Table S3C).

Plasma concentration of the host metabolite Kyn was significantly associated with twenty-four immune variables, fourteen of which remained significant after adjustment for multiple comparisons (Figure 3; Table S3A). In PBMC cultures, T-cell cytokine IL-13 and LPS-stimulated cytokines IL-6, IL-1β and IL-10 were positively associated with Kyn concentration (Figure 3C; Table S3A). Kyn was also positively associated with the plasma acute phase proteins CRP and SAA, the markers of vascular inflammation ICAM-1 and VCAM-1, the pro-inflammatory cytokine TNF-α, the regulatory cytokine IL-10, and the chemokines IL-8, eotaxin, and IP-10, and the marker of macrophage activation neopterin (Figure 3F; Table S3A).

Similarly, the Kyn/Trp ratio, included here as an index of IDO activity, was positively associated with twenty-one immune variables, fifteen of which remained significant after adjustment for multiple comparisons (Figure 3; Table S3B). Activated effector memory CD8 T-cells with CD38 activation marker (Figure 3A; Table S3B) and percentage of intermediate monocytes (Figure 3E; Table S3B) were positively associated with Kyn/Trp. Also, CRP, SAA, ICAM-1, VCAM-1, TNF-α, IL-10, the chemokines IL-8, IP-10, MCP-1, eotaxin, and MDC and the markers of macrophage activation neopterin and sCD14 were all positively associated with Kyn/Trp (Figure 3F; Table S3B).

The estimated effect size of the significant correlations adjusted for multiple comparisons are represented in Figures 4A, B. The largest effect size among the significant correlations between plasma immune variables with Kyn falls within the medium range for the immune markers including TNF-α (R² =0.112, p=1.88E-09), IP-10 (R² =0.123, p=9.17E-10), and neopterin (R² =0.154, p=2.22E-13) which respectively account for 11.2%, 12.3%, and 15.4% of variations among the plasma immune variables. The effect size of the significant correlations between Kyn and the rest of the plasma immune markers and the PBMC cytokines is within the small range (Figure 4A).

The standardized effect size for significant correlations between plasma immune markers and Kyn/Trp falls within the medium range for the immune markers including IL-10 (R² =0.098, p=1.20E-07), IP-10 (R² =0.132, p=5.70E-10), and TNF-α (R² =0.168, p=5.40E-14) which account for 9.8%, 13.2%, 16.8% ofplasma Kyn and Kyn/Trp with plasma markers variations among the plasma immune markers while neopterin (R² =0.291, p=8.52E-26) with large effect size explains 29.1% of variation among the plasma immune markers. The effect size of the significant correlations between Kyn/Trp and the rest of the immune markers is within the small range (Figure 4B).

Since gut bacteria may metabolize Trp to produce bioactive metabolites, we conducted a microbiome-wide analysis to identify microbial taxa either positively or negatively associated with plasma Trp metabolite concentrations divided into lower, middle, and upper terciles 1-3 (Figure 5; Table S4). We reasoned that such taxa might themselves be associated with immune variables because their production (or catabolism) of Trp metabolites would help determine the plasma concentration of these metabolites. Our results were as follows: At the family level, there was a significant negative association between the relative abundance of Lachnospiraceae and IAA concentration of the lower and upper terciles 1 (0.002-1.58 μM) and 3 (2.79-11.4 μM) (Figure 5A), and between Erysipelotrichaceae and IPA concentration of the lower and upper terciles 1(0.001-0.653 μM) and 3 (1.37-23.3 μM) (Figure 5B). At the genus level, there was a significant negative association between the relative abundance of Ruminococcus and IPA of the lower and upper terciles (Figure 5C) and Dorea and IPA concentration of the lower and upper terciles (Figure 5D).

Next, we determined if these taxa were associated with markers of immune activation and inflammation using the same regression analysis approach as used for the metabolites themselves (Table 2). At the family level, relative abundance of Lachnospiraceae was significantly (p<0.05) positively associated with total lymphocyte concentration in peripheral blood and negatively associated with activated (CD38 positive) effector memory CD4 T-cells. At the genus level, relative abundance of Dorea was significantly positively associated with T-cell production of IFN-γ in PBMC cultures. Also, at the genus level, relative abundance of Ruminococcus was positively associated with LPS-stimulated IL-6 in PBMC cultures (Table 2).

Dietary total protein and Trp intake were both positively correlated with plasma Trp (Table S2), as would be expected for an essential amino acid (38). Also, they were both positively correlated with plasma Kyn (Table S2), the precursor of which is plasma Trp. Interestingly, total protein intake was also positively associated with plasma concentration of IAA, but not IPA or indole (Table S2) suggesting that dietary Trp intake may determine the plasma concentrations of some Trp metabolites produced by intestinal bacteria and thus potentially affect health via a microbiome-dependent pathway, in addition to the well-defined role of Trp in host metabolism.

Plasma Trp concentrations were higher in men than women, as previously reported (39) and higher in the youngest category of study volunteers compared to the older two groups (Figure 1A). The reason for this age difference is uncertain but is not due to differences in protein or Trp intake as we saw no differences in dietary intake of total protein or specifically of tryptophan among these groups (data not shown). The positive association of Kyn/Trp with age (Figure 1C) is likely due to increasing IDO activity with age related to the phenomenon of inflammaging (2). Similarly, the higher Kyn and Kyn/Trp ratio in obese individuals compared to volunteers with a healthy BMI (Figures 1B, C) is presumably due to elevated IDO activity and inflammation in obesity (18). The lower IPA concentration in obese study subjects relative to normal-weight individuals (Figure 1F) is consistent with previous studies which also found a decreased plasma IPA level in obese individuals and reflects changes in the production of the gut microbial metabolites (40, 41).

Trp can be converted into indole, and its derivatives such as IAA and IPA, by intestinal bacteria such as Bacteroides thetaiotaomicroin, Citrobacter sp., E. coli, Peptostreptococcus russellii, Clostridium sporogenes, and Lactobacillus spp. (53–55). However, we did not find any positive correlation between gut bacteria and plasma levels of Trp metabolites derived from microbial metabolism (indole, IAA, IPA). One potential explanation for this lack of a positive association is that further metabolism of these compounds by host enzymes results in their lower concentration in plasma and obscures associations with intestinal bacteria. A study which reports that genetic factors which are responsible for further metabolism of IPA partially affect its circulating level supports this hypothesis (56).

Our study shows that the genera Dorea and L-Ruminococcus, both members of the family Lachnospiraceae were negatively associated with plasma IPA, and the family Lachnospiraceae itself was negatively associated with plasma IAA (Figure 5). Negative associations of Lachnospiraceae and Ruminococcus with plasma IPA have been previously reported (56) but the other associations, to our knowledge, have not been reported and their mechanism is not known. Further experiments would be needed to demonstrate a cause-effect association.

The family Lachnospiraceae, as well as the genera Dorea and L-Ruminococcus from the same family, were significantly associated with markers of immune activation. Members of family Lachnospiraceae have been shown to degrade mucins, potentially increasing gut permeability. Also, genera Dorea and L-Ruminococcus colonizing mucin regions of the GI tract can degrade mucins and metabolize the released sialic acids (57–59). Negative correlation of plasma IPA and IAA with this family could be illustrative of the relationship between mucin-associated bacteria that increase gut permeability and thus potentially induce immune activity. For example, Dorea was positively associated with T-cell production of IFN-γ and L-Ruminococcus with LPS-stimulated production of IL-6 (Table 2). This finding for Dorea is confirmed by a previous study which also showed that higher abundance of Dorea is associated with increased production of IFN-γ (60), though the association of L-Ruminococcus with IL-6 production is apparently novel. The negative association of Lachnospiraceae with effector-memory CD4 T-cells expressing the activation marker CD38 is not obviously consistent with the immune activation associated with this group of bacteria at the family level. It is likely that these effects are genus-specific. In summary, these data tend to support the suggestion that some Lachnospiraceae family member may cause immune stimulation, but the picture is complicated and further work involving experimental approaches in model systems is needed to confirm the association of L-Ruminococcus with IL-6 production.

In summary, our examination of the association of Trp metabolites with a broad spectrum of 88 immune markers showed that the Kyn/Trp ratio, a marker of IDO activity, was strongly associated with several markers of systemic inflammation, showing ongoing immune activation in a group of healthy adults without active acute or chronic disease. The Kyn/Trp ratio was positively associated with age and was higher in obese compared to non-obese individuals, emphasizing the association of these two factors with immune activation. In addition, commensal bacteria negatively associated with plasma indole metabolite concentrations were also associated with markers of systemic immune activation, suggesting that these taxa, perhaps by decreasing the availability of Trp metabolites such as indole, IAA and IPA to host tissues, increase the risk of systemic inflammation triggered by disruption of the mucosal barrier by these bacteria. Another possibility is that the relative abundance of these microbial taxa could be higher in a slightly more inflamed environment associated with lower levels of Trp metabolites for a yet to be identified another reason.