Metabolomic screening analysis of serum samples obtained from DILD and recovered patients

To identify new biomarkers for DILD, we performed HILIC/TOF–MS-based non-targeted metabolomic analysis focusing on hydrophilic molecules. Serum samples obtained from patients with DILD in the acute phase (n = 60), including DAD/DAD-mixed (n = 20), OP (n = 31), and NSIP (n = 9), and those from patients with DILD in the recovery phase (n = 34) were extracted and analyzed (Additional file 1: Table S1 and Fig. 1A). Volcano plot analysis revealed that five peaks and three peaks among 943 peaks were significantly increased (p-value < 5.30 × 10−5 = 0.05/943; two-tailed Welch’s t-test with Bonferroni correction, fold change > 2) and decreased (p-value < 5.30 × 10−5, fold change < 0.5) in the serum of DILD patients in the acute phase compared to those of recovered patients (Fig. 1B). The DILD biomarker candidates were further narrowed down using the g-value (> 0.8, or < -0.8) and area under the ROC curves (> 0.8), yielding five peaks. All were annotated as KYN-derived peaks (Fig. 1B).

Fig. 1

Identification of potential serum DILD biomarkers using metabolomic analyses for hydrophilic molecules. A Scheme showing unbiased metabolomic analysis (screening cohort) and quantitative analysis (validation cohort) for hydrophilic molecules in serum obtained from patients with DILD and recovered patients. Subsequently, the potential of the identified DILD biomarker candidates was evaluated using the combined cohort of patients with DILD, along with DILD-tolerant patients, patients with other lung diseases, and HC. Patient characteristics are summarized in Table 1, Additional file 1: Tables S1, and S2. B Serum levels of 943 hydrophilic molecules detected in metabolomic analysis were compared between patients with DILD and recovered patients (n = 60 and 34, respectively) in the volcano plot. In the analysis, a very stringent criterion was applied (p-value < 5.30 × 10−5 = 0.05/943; two-tailed Welch’s t-test with Bonferroni correction). Significantly increased or decreased molecules (fold change > 2 or < 0.5, and p-value < 5.30 × 10−5) are colored with red. Molecules not-significantly increased or decreased (fold change > 2 or < 0.5, p-value > 5.30 × 10−5) are colored with green. Only five detected molecules showed significantly increased (p-value < 5.30 × 10−5, fold change > 2) serum levels with high g-value (> 0.8) and high diagnostic potential (area under the ROC curve > 0.8) in acute DILD patients. Those molecules are labeled with tentative metabolite names

Validation of the elevation of serum KYN concentrations in patients with acute DILD

We collected serum from patients with DILD in the acute phase (n = 22) and recovery phase (n = 17) as a validation set and measured the serum KYN concentration by quantitative analysis. The median KYN concentration in patients with acute DILD was significantly higher than that in recovered patients (3.1 µM vs. 2.3 µM, p-value < 0.05, Mann–Whitney U-test, Additional file 3: Fig. S1). This is consistent with the result of the metabolomic study and supports the potential of KYN as a DILD biomarker.

Evaluation of serum concentrations of KYN pathway metabolites as new DILD biomarkers

KYN is generated from TRP and metabolized into QUNA (Fig. 2A) via the KYN pathway [29]. According to the previous publication, KYN [1.61 µM], QUNA [0.267 µM], and TRP [64.3 µM] are abundant in the bloodstream of healthy adults, unlike other metabolites of the KYN pathway (0.0258–0.0444 µM, or not detected) [30]. Based on the results of metabolomic screening and quantitative validation tests, we speculated that decreased TRP levels and elevated QUNA levels might also be DILD biomarker candidates. Therefore, in addition to KYN, we also focused on the concentrations of TRP and QUNA in the serum of patients with DILD in a combined cohort, where the combined sample set from screening and validation studies was used (Additional file 1: Table S1). The serum concentrations of KYN and TRP were quantified by RP-LC/MS. The serum concentrations of QUNA were quantified by IC/MS, since QUNA did not give clean peaks in our RP-LC/MS system.

Fig. 2

Serum KYN and QUNA concentrations and KYN/TRP ratio in DILD, other lung diseases, and HC. A The KYN pathway. Major metabolic enzymes are written in italics. IDO, indoleamine 2,3-dioxygenase; KATs, kynurenine aminotransferases; KMO, kynurenine 3-monooxygenase; KYNU kynureninase, 3-HAO; 3-hydroxyanthranilate oxidase; QPRT, quinolinate phosphoribosyl transferase. B Serum concentrations of KYN and QUNA, and KYN/TRP ratio among sample groups are shown as box-and-whisker plots. The middle lines represent the median values and the whiskers represent the highest and lowest values. The results of the statistical comparisons are summarized in Table 2. DAD/DAD-mixed: DILD patients in acute phase with CT pattern of diffuse alveolar damage, OP: DILD patients in acute phase with CT pattern of organizing pneumonia, NISP: DILD patients in acute phase with CT pattern of nonspecific interstitial pneumonia: Other: DILD patients in acute phase with CT pattern other than DAD, OP and NSIP, DILD recovery: patients recovered from DILD, DILD-tolerant: the patient group taking similar medications to DILD group but without DILD onset, BP: bacterial pneumonia, NTM: nontuberculous mycobacteriosis, IIPs: idiopathic interstitial pneumonias, CTD: lung disease associated with connective tissue disease, COPD: chronic obstructive pulmonary disease, BA: bronchial asthma, HC: healthy control

In the combined cohort, KYN, QUNA, and TRP were quantified in serum samples of DILD patients in acute phase (n = 81, including DAD/DAD-mixed [n = 22], OP [n = 32], NSIP [n = 26] and others [n = 1]), DILD patients in recovery phase (n = 53), patients with tolerance (n = 20) and other lung diseases, such as lung cancer (n = 45), BP (n = 14), NTM (n = 15), IIPs (n = 23), CTD (n = 20), COPD (n = 15), BA (n = 12), and healthy controls (HC, n = 30) (Fig. 2B and Additional file 3: Fig. S2). The median values of serum KYN and QUNA levels of all-DILD patients in the acute phase (4.0 µM and 1193.5 nM) were significantly higher than those of the DILD recovery patients (2.1 µM and 325.9 nM, Mann–Whitney U-test with Bonferroni correction) (Fig. 2B and Table 2). Compared with recovered patients, increased KYN and QUNA levels were observed in all types of DILD subgroups (Fig. 2B and Table 2). The median values of serum KYN and QUNA levels tended to be higher in the DAD/DAD-mixed group than other DILD subgroups (KYN, 5.0 µM vs. 3.4–3.5 µM; QUNA, 1557.9 nM vs. 821.0–1246.1 nM), although the only significant differences were observed between the DAD/DAD-mixed and OP groups (Table 2, Mann–Whitney U-test with Bonferroni correction). Statistically significant differences in KYN and QUNA concentrations between acute DILD patients and DILD-tolerant patients (who were administered similar drug(s) but without DILD onset) (KYN, 2.1 µM; QUNA, 286.9 nM) or HC (KYN, 1.4 µM; QUNA, 210.0 nM) were also confirmed (Fig. 2B and Table 2). Serum concentrations of KYN and QUNA in acute DILD patients were higher than those of patients with other lung diseases, including lung cancer (KYN, 2.1 µM; QUNA, 252.3 nM), BP (KYN, 2.3 µM; QUNA, 486.8 nM), NTM (KYN, 2.2 µM; QUNA, 472.2 nM), IIPs (KYN, 2.4 µM; QUNA, 496.4 nM), CTD (KYN, 2.4 µM; QUNA, 480.9 nM), COPD (KYN, 2.2 µM; QUNA, 365.8 nM) and BA (KYN, 2.1 µM; QUNA, 263.5 nM) (Table 2).

Table 2 Statistical comparison of serum concentrations of TRP metabolites and TRP/KYN ratio among patient groups

Consistent with the increased serum levels of KYN and QUNA, levels of TRP in all-DILD patients in acute phase (41.3 µM) were significantly lower than those of DILD recovery patients (52.1 µM), DILD-tolerant patients (55.3 µM), patients with lung cancer/NTM/BA (50.4–61.0 µM), and HC (51.4 µM) (Additional file 3: Fig. S2 and Table 2). However, significant differences in TRP levels among all patients with DILD and BP/IIPs/CTD/COPD were not observed (Table 2). The extent of differences in the median values of serum concentrations of TRP between all-DILD patients and other groups were less than those of KYN and QUNA (Table 2), implying that KYN and QUNA might be better DILD biomarkers than TRP.

Additionally, as the increases in KYN and QUNA levels and the reduction of TRP levels were observed synchronously in patients with acute DILD, we compared the KYN/TRP ratio among the groups to investigate the activation of metabolic conversion of TRP via the KYN pathway (Fig. 2B). As with KNY and QUNA levels, the median value of the KYN/TRP ratio in all-DILD patients in the acute phase (0.108) was higher than in those in the DILD recovery group (0.038), tolerant patients (0.036), patients with other lung diseases (0.039–0.051), and HC (0.028) (Fig. 2B, Table 2).

Collectively, our findings demonstrated that serum concentrations of KYN and QUNA were significantly increased in acute DILD patients (especially in acute patients with DAD/DAD-mixed patterns), and that the serum concentration of TRP was decreased in acute DILD patients. Along with the data on raw concentrations of the three metabolites, the elevated KYN/TRP ratio in acute DILD patients suggested that the value of each metabolite and the ratio might be DILD biomarker candidates.

To examine sampling bias, we performed sensitivity analysis for the serum concentrations of KYN pathway metabolites (KYN, QUNA, and TRP) and KYN/TRP ratio between patients with DILD in the acute and recovery phases by dividing the samples into two sub-cohorts by hospital location (Chiba University and Nippon Medical School; Tokyo metropolitan area [sub-cohort A], Shinshu University, and Hiroshima University; the other area [sub-cohort B]). Compared with recovered patients, significant increases in KYN and QUNA levels and the KYN/TRP ratio and a reduction in TRP levels were observed in both subgroups (Additional file 3: Fig. S3, Mann–Whitney U-test). These results indicate that the current findings were not affected by sampling bias.

Next, we asked if the extent of elevated serum KYN and QUNA levels and the KYN/TRP ratio is associated with patient backgrounds, such as specific underlying disease and types of medications in patients with acute DILD. We first tested the effect of the existence of cancers on serum KYN and QUNA levels as well as the KYN/TRP ratio in patients with DILD. The metabolite levels and KYN/TRP ratio between DILD patients with and without cancer were comparable (adjusted p-value > 0.05, Mann–Whitney U-test with Bonferroni correction, Additional file 3: Fig. S4A). Next, the association of serum KYN and QUNA levels, as well as the KYN/TRP ratio, with underlying lifestyle-related diseases was investigated. No statistically significant differences among groups were detected between the patients with DILD with or without lifestyle-related diseases (adjusted p-value > 0.05, Mann–Whitney U-test with Bonferroni correction, Additional file 3: Fig. S4B). The effect of medication type on the DILD biomarker candidates was also examined. No significant correlation was observed (adjusted p-value > 0.05, Mann–Whitney U-test with Bonferroni correction, Additional file 3: Fig. S4C). These results suggest that serum concentrations of KYN, QUNA, and the KYN/TRP ratio are not related to at least the patient characteristics analyzed, indicating that elevated serum KYN and QUNA concentrations as well as the KYN/TRP ratio may be biomarkers for acute DILD patients with a wide range of patient characteristics.

Furthermore, it is intriguing to explore whether the degree of the activation of KYN pathway correlates with the severity or mortality of DILD. Initially, we evaluated the correlation between KYN/TRP ratio and SpO2/FiO2 ratio, an indicator of the degree of hypoxemia, in DILD in acute phase DILD patients. Our data shown that no statistically significant correlation was observed (r = − 0.1882, p = 0.22, Additional file 3: Fig S5A). Additionally, we compared the KYN/TRP ratio between acute DILD patients who survived and those who died due to DILD exacerbation. The result indicated no statistically significant difference in the KYN/TRP ratio between the two groups (Additional file 3: Fig S5B). Collectively, although further investigations are warranted, our findings demonstrate that the activation level of KYN pathway in acute DILD patients may not be associated with its severity and mortality.

Potential of KYN, QUNA, and KYN/TRP as biomarkers for the diagnosis of the onset of DILD and its recovery

The diagnostic potential of serum KYN, QUNA, and TRP levels and the KYN/TRP ratio for DILD was analyzed using the AUROC values and compared with those of conventional serum markers for interstitial lung diseases (ILD) (KL-6 and SP-D) and inflammation (C-reactive protein [CRP]). The concentration ranges of these markers and results of statistical comparisons are shown in Additional file 3: Fig. S6 and Additional file 1: Table S5, respectively, for all samples. Representative ROC curves for all-DILD patients and DAD/DAD-mixed patients are shown in Fig. 3 and Additional file 3: Fig. S7, respectively. The results of AUROC values are summarized in Table 3.

Fig. 3

Diagnostic potentials of KYN, QUNA, KYN/TRP ratio, and conventional DILD biomarkers. ROC curve analyses of serum levels of KYN and QUNA, KYN/TRP ratio, and levels of conventional ILD biomarkers (SP-D and KL-6) were performed between the groups using the quantitative data of all samples in the combined cohort. The ROC curves of all-DILD patients compared with DILD-tolerant A, DILD recovery B, IIPs C, or CTD D are shown. The values of AUROC are described in the parentheses of the labels for each tested biomarker. The AUROC values for other comparisons are summarized in Table 3

Table 3 Summary of DILD biomarker potentials of TRP metabolites in comparison with conventional serum biomarkers

The potential for diagnosing the onset of DILD was evaluated for the identified biomarker candidates via ROC curves analysis, comparing acute DILD to the DILD-tolerant control patients or HC (Fig. 3A and Table 3). The diagnostic potentials (AUROC values) of serum KYN (0.85) and QUNA (0.90) levels and KYN/TRP ratio (0.91) were comparable or superior to those of KL-6 (0.74), SP-D (0.89), and CRP (0.88). Moreover, the AUROC value of TRP between the all-DILD and tolerant groups was 0.75, which was lower than that of KYN, QUNA, and the KYN/TRP ratio (Table 3), indicating that the diagnostic potential of TRP as a DILD biomarker is insufficient. KYN, QUNA, and KYN/TRP also showed high AUROC values (≥ 0.97, Table 3) compared with HC. These data demonstrated the usefulness of the KYN, QUNA, and KYN/TRP in the diagnosis of DILD onset.

As a next step, we assessed their biomarker potentials for diagnosing DILD recovery. ROC analyses showed that the AUROC values between the all-DILD and DILD recovery groups of KYN (0.82) and QUNA (0.81) were higher than those of conventional serum biomarkers (KL-6 [0.59], SP-D [0.77], and CRP [0.75], Fig. 3B and Table 3). In addition, compared with that of KYN, the KYN/TRP ratio showed a slightly improved diagnostic potential (0.86) between the all-DILD and DILD recovery groups (Fig. 3B). Collectively, our results indicate that serum KYN and QUNA levels and the KYN/TRP ratio are feasible biomarkers for monitoring recovery from DILD.

The positivity rate of the biomarker candidates and conventional ILD biomarkers was analyzed (Additional file 1: Table S6). Using Youden’s index of the ROC curve built between patients with acute DILD and those with DILD recovery, the optimal cutoff values of the identified biomarker candidates were determined. The positivity rate of biomarker candidates in patients with acute DILD (78.8–82.5%) was comparable to that of conventional biomarkers (67.9–81.5%). Notably, KYN, QUNA, and KYN/TRP (new biomarkers) exhibited lower positivity rates than conventional biomarkers, not only in DILD-tolerant patients (5.0–30.0% [new biomarkers] vs. 25.0–35.0% [conventional biomarkers] but also in recovered patients (19.6–29.4% [new biomarkers] vs. 45.3%–52.8% [conventional biomarkers]), demonstrating the higher specificity of the new biomarkers.

In addition to the evaluation of biomarker potential in all-DILD patients, our focus extended to DAD/DAD-mixed patients who suffered from more severe DILD than patients with other imaging patterns. In the ROC analyses between the DAD/DAD-mixed and tolerant groups, a strikingly high diagnostic potential of KYN, QUNA, and KYN/TRP was observed (AUROC ≥ 0.96, Additional file 3: Fig. S7A and Table 3). Furthermore, the AUROC values of these biomarker candidates for diagnosing DILD recovery ranged from 0.89 to 0.97 (Additional file 3: Fig. S7B and Table 3). These data suggest that the performance of the novel DILD biomarker candidates tends to be higher in patients with DAD/DAD-mixed DILD than in those with all-DILD.

Meanwhile, the diagnostic potential of the KYN, QUNA, and KYN/TRP for detecting of the onset of DILD and its recovery was also assessed in OP and NSIP patients (Additional file 1: Table S7). While the AUROC values were lower than those observed in DAD/DAD-mixed patients, the AUROC values still exhibited significant diagnostic capability in OP and NSIP patients. These findings suggest the usefulness of these biomarker candidates in OP and NSIP patients.

Superiority of KYN, QUNA, and KYN/TRP over conventional ILD biomarkers in the differential diagnosis between DILD and other lung diseases, including IIPs and CTD

Discrimination of DILD from IIPs and CTD is important because administration of causative drugs should be stopped in patients with DILD but not in those with IIPs and CTD. The drawbacks of the clinically available ILD biomarkers (KL-6 and SP-D) include their inability to distinguish DILD from IIPs and CTD [10, 13,14,15]. To overcome these drawbacks, new DILD biomarkers must exhibit higher specificity for DILD, enabling them to effectively differentiate DILD from IIPs and CTD. To determine whether the identified biomarker candidates are DILD-specific biomarkers, we calculated AUROC values in combinations of all-DILD and patients with IIPs or CTD (Fig. 3C and D, Table 3). Higher diagnostic potentials between IIPs patients and all-DILD patients were observed for KYN (0.73), QUNA (0.77), and KYN/TRP ratio (0.77) than for KL-6 (0.65) and SP-D (0.64) (Fig. 3C). As for the diagnostic power to differentiate between DILD and CTD, the new biomarkers had much higher AUROC values (0.75–0.80) than KL-6 (0.56) and SP-D (0.62) (Fig. 3D). Among the new biomarkers, KYN/TRP ratio showed the lowest the positivity rates in patients with IIPs (26.1%) and CTD (35.0%); the rates were much lower than those of KL-6 (70.0–82.6%) and SP-D (50.0–82.6%) (Additional file 1: Table S6). Additionally, a trend toward higher diagnostic performance to distinguish DILD from IIPs and CTD (0.87–0.93 [new biomarkers] vs. 0.59–0.69 [conventional biomarkers]) was observed in DAD/DAD-mixed patients (Additional file 3: Fig. S7C and D and Table 3). Similar tendency was also observed in OP and NSIP patients (Additional file 1: Table S7). These data indicate that the new biomarkers have the potential to overcome the weakness of the conventional markers in terms of discriminating DILD from IIPs and CTD.

In the current study, we also compared the diagnostic potential of the new biomarkers for discriminating DILD from other lung diseases, including lung cancer, BP, NTM, COPD, and BA. The AUROC values for the new biomarkers were comparable to those of the ILD biomarkers in the analyses comparing all-DILD with each related-lung disease (0.79–0.93 vs. 0.72–0.94, Table 3). Similar results were observed for DAD/DAD-mixed, OP, and NSIP patients (Additional file 1: Table S7 and Table 3). Taken together, our findings indicate that serum KYN and QUNA levels, as well as the KYN/TRP ratio, are useful biomarkers to support the specific diagnosis of DILD in patients with heterogeneous clinical backgrounds.

IDO1 induces KYN pathway activation upon macrophage differentiation and IFNγ stimulation in monocytic cell lines

Patients with DILD usually have severe lung inflammation and may show immunological activation. A significant correlation between CRP levels and KYN/TRP ratios was observed in patients with DILD (r = 0.47, p < 0.0001, Fig. 4A), suggesting the potential association of inflammation with metabolic activation of the KYN pathway. Next, we compared the percentage of major white blood cell types (monocytes, lymphocytes, and neutrophils) between patients with acute DILD and recovered patients to show the cell types responsible for KYN production. The percentage of monocytes was significantly higher in patients with acute DILD than in recovered patients (7.30% vs. 5.45%, p < 0.05, Mann–Whitney U-test, Fig. 4B). Meanwhile, the percentage of lymphocytes and neutrophils between acute and recovery phases was comparable (Fig. 4B). These data suggest that monocytes or cells differentiated from monocytes are involved in KYN pathway activation.

Fig. 4

Activation of KYN pathway in monocytic cell lines upon macrophage differentiation and IFNγ stimulation. A Correlation analysis of KYN/TRP ratio with serum CRP concentrations. B Comparison of median values of the percentage of major white blood cell types in patients with DILD and recovered patients. C Expression levels of IDO1 mRNA were examined using RT-qPCR in undifferentiated (U937 and THP1) cells and differentiated (dU937 and dTHP1) macrophage-like cell lines. D Fold changes of IDO1 mRNA expression levels upon vehicle (10% FBS-PBS) or IFNγ (10 ng/mL) treatment were examined in differentiated macrophage cell lines (dTHP1 and dU937). E Western blotting analysis of IDO1 expression in dU937 and dTHP1 cells upon treatment with vehicle (10% FBS-PBS) or IFNγ (10 ng/mL) for 24 h. GAPDH expression was analyzed as a loading control. Uncropped images are shown in Additional file 4. F Levels of TRP, KYN, and QUNA were measured in undifferentiated or differentiated THP1 cell supernatant treated with 0.1% (v/v) DMSO or PMA at 10 ng/mL for 24, 48, 72, and 96 h. The metabolite level in undifferentiated THP1 cells after 24 h DMSO exposure was set as 1. The time-dependent changes of each metabolite level in undifferentiated cells and levels between undifferentiated and differentiated THP1 cells were statistically compared. G and H Relative levels of TRP, KYN, and QUNA in supernatant and whole cell lysate of dTHP1 cells treated with 10% FBS-PBS (control) or IFNγ (10 ng/mL) for 24 h are shown. Each bar represents the mean ± standard deviation of three independent experiments. Statistical significance of mean or median values was tested using Student’s t-test or Mann–Whitney U test (ns, not significant; *p-value < 0.05, **p-value < 0.01; ***p-value < 0.001), respectively. Bonferroni correction was conducted for multiple comparisons

To clarify detailed molecular mechanisms underlying activation of the KYN pathway in patients with DILD, we explored immune-related biological molecules that induce the expression of the rate-limiting enzyme, IDO1, which mediates the conversion from TRP to KYN in the KYN pathway (Fig. 2A) [29, 31]. To date, IDO1 expression and activity in alveolar macrophages have been reported in a mouse model of pneumonia caused by allogeneic hematopoietic stem cell transplantation or viral infection [32, 33]. As monocytes can differentiate into macrophages under inflammatory conditions, we performed in vitro analyses to investigate the effects of macrophage differentiation and immune-stimulations on the induction of IDO1 expression and metabolic changes of KYN, QUNA, and TRP.

First, the effect of monocytes on macrophage differentiation was studied using THP1 and U937 monocytic leukemia cell lines as models. These cell lines differentiate into macrophage-like cells upon stimulation with PMA. To examine IDO1 expression in differentiated macrophages, we measured IDO1 mRNA expression in THP1 and U937 cell lines with or without PMA treatment (Fig. 4C). The results of RT-qPCR showed more than tenfold induction of IDO1 mRNA expression can be observed in differentiated THP1 and U937 cells (dTHP1 and dU937) compared with undifferentiated cells, implying that differentiation of monocytes into macrophages contributes to IDO1 induction.

Given that the expression of IDO1 is affected by immunological stimulation [34,35,36], we examined the effect of various inflammatory stimuli (IFNα-2a, IFNα-2b, IFNγ, IFNβ, TNFα, IL-1β, IL-6) and anti-inflammatory stimuli (IL-4 and IL-10) on IDO1 mRNA expression in dTHP1 and dU937 cells (Additional file 3: Fig. S8A and B). In both dTHP1 and dU937 cells, levels of IDO1 mRNA were significantly increased by treatment with IFNα-2a, IFNα-2b, IFNβ, IFNγ, and TNFα (adjusted p-value < 0.01, Student’s t-test with Bonferroni correction), whereas its levels were decreased by treatment with anti-inflammatory IL-4 (adjusted p-value < 0.01). IL-1β and IL-10 treatment did not change IDO1 expression in dTHP1 cells, but significantly induced or reduced IDO1 expression in dU937 cells (adjusted p-value < 0.01). IL-6 treatment moderately decreased IDO1 expression in dTHP1 cells but not in dU937 cells. As IFNγ treatment produced the highest induction of IDO1 mRNA in differentiated macrophages, we used Western blotting to examine the protein levels in cells with and without IFNγ treatment. The expression of IDO1 was higher in dTHP1 cells treated with IFNγ than in untreated cells (Fig. 4E). The protein expression of IDO1 in dU937 cells was mildly induced by IFNγ treatment (Fig. 4E). Collectively, these results demonstrate that differentiated macrophages can respond to stimulation by various cytokines to produce IDO1 in vitro.

To examine the effects of IDO1 induction on the extracellular secretion of KYN and QUNA upon the differentiation of THP1 cells, we measured TRP, KYN, and QUNA levels in the supernatant of THP1 cells over the time course of PMA treatment (Fig. 4F). TRP levels in the THP1 cell supernatant decreased in a time-dependent manner from 24 to 96 h, even without PMA treatment (adjusted p-value < 0.01, Student’s t-test with Bonferroni correction), indicating basal consumption of TRP for cell proliferation. TRP levels in the supernatant of dTHP1 cells were significantly lower than those in undifferentiated THP1 cells after 72 h of incubation (adjusted p-value < 0.01, Fig. 4F), implying that the differentiation of THP1 cells enhanced TRP consumption. KYN levels in dTHP1 cells increased in the supernatant of dTHP1 cells as time progressed from 48 to 96 h (9.9- to 25.9-fold, adjusted p-value < 0.001 compared with undifferentiated cells, Fig. 4F), whereas KYN levels did not change in undifferentiated THP1 cells. Following the increase in KYN level, QUNA levels in the dTHP1 cell supernatant were significantly increased from 72 h after PMA treatment (adjusted p-value < 0.001 compared with undifferentiated cells, Fig. 4F). These results suggest that the differentiation of monocytes to macrophages enhances TRP metabolism via the KYN pathway and leads to the secretion of its metabolites, such as KYN and QUNA, into the extracellular space.

As IFNγ induced IDO1 expression at the highest level in dTHP1 cells (Additional file 3: Fig. S8B), we analyzed the effects of treatment with IFNγ for 24 h on the levels of TRP, KYN, and QUNA in the supernatant of dTHP1 cells. TRP levels in the supernatant of dTHP1 cells were dramatically reduced by IFNγ treatment (p-value < 0.001, Fig. 4G). In contrast, KYN levels in the supernatant of dTHP1 cells were not changed by IFNγ treatment (Fig. 4G). Nevertheless, extracellular QUNA levels were increased by IFNγ treatment (p < 0.01, Fig. 4G). As clear increases in KYN level were not observed in the supernatants, we also analyzed the levels of TRP metabolites in the whole cell lysates of dTHP1 cells treated with IFNγ (Fig. 4G). Consistent with the findings using supernatants, TRP levels were significantly decreased by IFNγ treatment (p-value < 0.001, Fig. 4H). Moreover, a significant elevation in KYN (p-value < 0.01) and QUNA (p-value < 0.001) levels occurred after IFNγ treatment (Fig. 4H). Similar experiments were performed using dU937 cells (Additional file 3: Fig. S9). IFNγ-based activation of the KYN pathway was also observed in dU937 cells, although the extent of activation differed from that in dTHP1 cells (Additional file 3: Fig. S9).

Taken together, our findings imply that inflammation in the lungs of patients with acute DILD stimulates IFNγ signaling, leading to IDO1 production in macrophages differentiated from monocytes, thereby inducing the metabolism of TRP to KYN and QUNA and contributing to the increase in KYN and QUNA levels in the systemic circulation.

IFNγ stimulation induces the production of KYN but not QUNA in human lung microvascular ECs

In addition to in macrophages, the basal protein expression of IDO1 occurs in ECs in normal human lung tissues [37]. Additionally, predominant IDO1 expression can be observed in lung ECs of COVID-19 patients [38, 39], in whom elevated serum KYN and QUNA concentrations have been reported [40]. To clarify whether lung ECs contribute to the production of KYN and QUNA upon changes in immunological conditions, we investigated the effect of various inflammatory and anti-inflammatory stimuli on IDO1 mRNA expression using immortalized normal human lung microvascular ECs (HULEC-5a cells) (Additional file 3: Fig. S8C). The significant induction of IDO1 mRNA expression was observed upon stimulation with IFNβ, IFNγ, and TNFα (adjusted p-value < 0.01). Among these cytokines, IFNγ was the most potent inducer of IDO1 mRNA expression; an over 20,000-fold increase in IDO1 mRNA levels was observed after its application (Fig. 5A and Additional file 3: Fig. S7C). However, a decrease in mRNA levels upon anti-inflammatory stimulation was not observed.

Fig. 5

Activation of KYN pathway in human lung ECs upon IFNγ stimulation. A Fold changes of IDO1 mRNA expression levels upon vehicle (10% FBS-PBS) or IFNγ (10 ng/mL) treatment were examined using HULEC-5a cells. B Western blotting analysis of IDO1 expression in HULEC-5a cells upon treatment with vehicle or IFNγ (10 ng/mL) for 48 h. GAPDH expression was analyzed as a loading control. Uncropped images are shown in Additional file 5. C Relative levels of TRP, KYN, and QUNA in supernatant of HULEC-5a cells treated with vehicle or IFNγ (10 ng/mL) for 48 h are shown. D Relative mRNA expression levels of major metabolic enzymes of KYN pathway (KMO, KYNU, and 3-HAO) in differentiated macrophage-like cell lines (dTHP1 and dU937) and HULEC-5a cells. Each bar represents the mean ± standard deviation of three independent experiments. Statistical significance of mean values was tested using Student’s t-test (ns, not significant; **p-value < 0.01; ***p-value < 0.001). Bonferroni correction was conducted for multiple comparisons

The protein expression of IDO1 was analyzed using HULEC-5a cells treated with IFNγ. Western blotting showed no basal IDO1 protein expression in control HULEC-5a cells but increased protein expression levels after IFNγ treatment (Fig. 5B). Furthermore, we examined the effects of IFNγ treatment on the concentrations of TRP, KYN, and QUNA in the supernatant of HULEC-5a cells (Fig. 5C). TRP levels were significantly decreased to 3.9%, whereas KYN levels were increased (34.1-fold) by IFNγ treatment (p-value < 0.0001, Student’s t-test, Fig. 5C). However, contrary to the results for differentiated macrophage-like cell lines, QUNA levels in the cell supernatant of HULEC-5a were not affected by the treatments (Fig. 5C).

To clarify the reasons for the unchanged levels of QUNA in HULEC-5a cells, we analyzed the mRNA expression of metabolic enzymes involved in QUNA generation from KYN, such as KMO, KYNU, and 3-HAO (Fig. 2A). Interestingly, the mRNA levels of these enzymes