Polyunsaturated fatty acid-derived lipid mediators as potential biomarkers for leprosy among individuals with asymptomatic Mycobacterium leprae infection

Abstract

Intra-household contacts of leprosy patients (HCs) are at increased risk of infection by Mycobacterium leprae and about ~ 5–10 % will develop active disease. A prognostic tool to identify HCs with the greatest risk of progressing to active disease would enhance early leprosy diagnosis and optimize prophylactic intervention. Previous metabolomics studies suggest that host lipid mediators derived from ω-3 and ω-6 polyunsaturated fatty acids (PUFAs) are potential biomarkers for leprosy. In this study, we investigated retrospective sera of leprosy HCs by liquid chromatography-mass spectrometry and enzyme-linked immunoassay to determine whether circulating levels of ω-3 and ω-6 PUFA metabolites were altered in HCs that developed leprosy (HCDLs) in comparison to those that did not (HCNDLs). Sera were collected from HCs at the time of index case diagnosis and before clinical signs/symptoms of leprosy. Our findings showed that HCDL sera exhibited a distinct metabolic profile in comparison to HCDNL. Specifically, arachidonic acid, leukotriene B₄,11-hydroxyeicosatetraenoic acid, prostaglandin D2 and lipoxin A₄ were elevated in HCDL. In contrast, prostaglandin E2 levels were reduced in HCDL. The ω-3 PUFAs, docosahexaenoic acid and eicosapentaenoic acid, and the docosahexaenoic acid-derived resolvin D1 and maresin-1 were also elevated in HCDL individuals compared to HCNDL. PCA analyses provided further evidence that lipid mediators could serve as an early biomarker for progression to active leprosy. A logistics model identified resolvin D1 and D2, and prostaglandin D2 as having the greatest potential for early detection of HCs that will manifest leprosy.

Graphical Abstract

Leprosy, also known as Hansen’s disease, is a chronic infectious disease that affects peripheral nerves and skin. Despite effective antimicrobial treatments against the main infecting pathogen, Mycobacterium leprae, leprosy remains prevalent in some parts of the world ¹. The disease presents as a spectrum of clinical forms that are defined by the host’s immune response to the pathogen ². Leprosy patients are broadly classified as paucibacillary (PB) (typically individual with one to five skin lesions) or multibacillary (MB) (individual with more than five skin lesions) ³. PB patients exhibit a more dominant T helper 1 (Th1) immune response that limits the growth of M. leprae, while MB patients display a weaker Th1 response resulting in permissive bacterial replication and high bacterial loads ². Leprosy diagnosis and classification mainly relies on clinical evaluation, light microscopy for detection of acid-fast bacilli in slit-skin smears, and the histopathology of skin biopsies ^{2, 4}. Limited access to healthcare and a lack of clinical expertise in leprosy can cause diagnostic delays that contribute to the disease progression and nerve damage ⁴.

Leprosy is primarily transmitted between humans via aerosol droplets. However, evidence of leprosy in nine-banded armadillos and red squirrels also makes them potential vectors for transmission ^{5, 6}. Major risk factors for contracting leprosy include close contact with MB patients, consanguinity, and intra-household contact ⁷. About 5–10% of household contacts of leprosy patients (HCs) will progress to active disease ^8–10. However, the early and accurate identification of HCs that will progress to active leprosy remains a challenge ¹¹. The majority of studies targeting the early detection of individuals with leprosy are based on immune responses to M. leprae or its antigens^{12, 13}. Such studies demonstrate that the induction of IFN-γ and other cytokines in peripheral blood cells stimulated with M. leprae whole-cell sonicate or recombinants proteins can be used as a tool to assess M. leprae exposure. Serum antibody responses to M. leprae products such as phenolic glycolipid I (PGL-I), the fusion protein LID-1 (a combination of ML0405 and ML2331) and the NDO-LID1 (a synthetic combination of LID-1 with the PGL-I) are also useful in identifying individuals exposed to or infected with M. leprae ^{8, 11, 14, 15}. However, longitudinal serological assays for PGL-I in HCs were unable to predict the onset of leprosy ¹¹. It was observed that positive detection of both anti-PGL-I-IgM and the M. leprae-specific repetitive sequence (RLEP) DNA region did indicate latent disease ⁸. van Hooij et al. ¹⁵ demonstrated that the combined measurement of humoral (anti-PGL-I) and cellular markers (C-reactive protein, CCL4, and IP-10) in blood samples improves the detection of leprosy patients, including those with PB form. These findings suggested for the first time that the detection of M. leprae antigen and host markers could be applied to a fingerstick blood test, a less invasive test that can be used for the diagnosis of leprosy. Only recently have host markers been targeted as a means to identify HCs that are at the greatest risk of developing active leprosy and identified a four gene transcriptional signature in HCs that progressed to leprosy, 5-years prior to clinical diagnosis ¹⁶.

Beyond transcriptional signatures, the use of hosts’ metabolic profiles (metabolomics) offers a new perspective in the study of human infectious diseases and the ability to develop predictive biomarkers ¹⁷. A limited number of studies have evaluated the metabolic profiles of leprosy patients^18–21. Interestingly, a common theme emerging from the collective results of these studies is a marked alteration in the serum levels of ω-3 and ω-6 polyunsaturated fatty acid (PUFAs)-derived lipid mediators that occurs in leprosy patients and between the clinical forms of disease ^{18, 19}. Recent findings from our group demonstrated that leprosy patients with type 1 reactions (T1R, an acute localized cell-mediated inflammatory event ⁴), as compared to those without T1R exhibit a significant increase in the levels leukotriene B₄ (LTB₄) and a significant decrease in resolvin D1 (RvD1) ²¹. It was also observed that patients with erythema nodosum leprosum (ENL), a systemic immune-mediated inflammation that may be recurrent or chronic, present high levels of PUFA-derived lipid mediators during the entire course of the disease (at multibacillary leprosy diagnosis, ENL onset and treatment follow-up) ²⁰. These findings reinforce the hypothesis that the nature and magnitude of immune responses in leprosy patients are related to alterations in PUFA metabolism. However, it is unknown whether subclinical M. leprae infection also alters host metabolism and whether these alterations are linked with susceptibility to leprosy. In this current study we explored whether differential ω-3 and ω-6 PUFA metabolism occurs in HCs and whether these alterations correlated with progression to leprosy.

Results

HCs individuals that manifested leprosy displayed higher levels of anti-PGL-I IgM and the corresponding index cases showed increased BI.

The study population included 17 HCs individuals who remained healthy (HCNDL) and 17 HCs that manifested clinical leprosy symptoms (HCDL) with no statistical difference in male to female proportion (HCNDL, 58.8% female; HCDL, 64.7% female; p=1.00). There was also no significant difference in the mean age (mean of 45.2 ± 15.8 for HCNDL and 37.1 ± 14.8 for HCDL; p=0.30) and in the number of BCG scars (p=0.84). The levels of anti-PGL-I IgM antibody at the time of diagnosis of the index cases were significantly higher in HCDL individuals (HCDL, OD₄₅₀ mean = 0.43 ± 0.28; HCNDL, OD₄₅₀ mean = 0.11 ± 0.11, p<0.001) (Figure 1A). In addition, the percentage of HCDL individuals positive for anti-PGL-I IgM antibody was 70.6% (12 out of 17) against 5.9% of HCNDL individuals (1 out of 17). Within HCDL group, no significant difference was found when we compared the levels of anti-PGL-I IgM before and after the development of leprosy symptoms (at T0, OD₄₅₀ mean = 0.46 ± 0.29; after T0 OD₄₅₀ mean = 0.51 ± 0.23, p=0.32). No difference regarding the distribution of clinical forms of the index cases between HCNDL and HCDL groups was observed, but the index cases of the HCDL group exhibited significantly higher BI mean (HCNDL, mean 1.35 ± 1.65; HCDL, mean 3.14 ± 1.14; p=0.0019) (Figure 1B). Interestingly, anti-PGL-I IgM antibody levels from HCs individuals together with the BI from the index cases could be used to separate the HCDL contacts group from the HCNDL group (Figure 1C).

Figure 1. Anti-PGL-I IgM titers from HCs and the bacillary index of the index cases differ between HCs that developed leprosy and those that remained healthy.

(A) Serum IgM reactivity to PGL-I for HCNDL (n=17) and HCDL (n=17) as measured by ELISA. The IgM responses were not available for two individuals from the HCNDL group and for three individuals from the HCDL group. The solid black lines represent the mean values. The dashed line indicates the cutoff value for a positive response (D.O.₄₅₀ = 0.30). (B) Bacillary index of the index cases of HCNDL (n=17) and HCDL (n=17) individuals. Statistical comparisons were performed using a nonparametric Mann-Whitney test. * Represent p<0.05. (C) Plot of the IgM anti-PGL-I responses of the HCs versus the bacillary index for the respective index case. Orange circles and purple triangle represent HCNDL and HCDL individuals, respectively. The dashed lines were included to highlight the differential characteristics of the HCDL subjects from HCNDL individuals.

HCDL individuals manifested several forms of leprosy at least one year after the diagnosis of the index cases.

Serum samples from HCNDL and HCDL subjects were obtained at the time of the diagnosis of the index cases. It is important to highlight that when HCDL samples were obtained the HCDL subjects did not display any clinical symptoms or signs of leprosy as previously explained in methods section. Among HCDL group, the time to clinical leprosy presentation after diagnosis of the index case varied widely (from 1 to 10 years, with a mean of 3.4 years and a median of 2 years; Supplementary Table 1). Seven individuals (41.2%) developed leprosy within one year of the index case diagnosis, while four (23.5%) manifested the disease within two years, and six (35.3%) displayed leprosy symptoms between 4–10 years. As presented in Supplementary Table 1, several forms of leprosy emerged in the HCDL group (indeterminate leprosy [I], 41.2%; borderline tuberculoid [BT], 17.6%; borderline borderline [BB], 11.7%; borderline lepromatous, 5.9%; lepromatous leprosy [LL], 11.8%; pure neural leprosy [PNL], 11.8%.

Healthy leprosy household contacts that evolved to active disease exhibited a serum metabolome profile distinct from those that remained healthy.

Raw LC-MS data from individual samples were processed by XCMS and a total of 5,540 features in negative-ionization mode and 12,361 features in positive-ionization modes were obtained. After the filtering steps, 2,280 and 6,876 negative- and positive-ion features were applied to the comparative analyses. A total of 3,245 features (1,137 and 2,108 negative- and positive-ion features, respectively) displayed log₂ abundance FC ≥1.0 and p<0.05 between the HCDL and HCNDL groups (Supplementary Table 2). Principal Component Analyses (PCA) of the negative- and positive-ion data indicated differences between the metabolic profiles of HCDL and HCNDL subjects (Figure 2A and 2B). The potential differences in metabolic pathways between these two groups were interrogated using MetaboAnalysts ²² and more than 20 metabolic pathways were identified as being enriched (Supplementary Table 3). Among these, arachidonic acid (AA) metabolism was significantly enriched (p<0.05) and exhibited a significant IV of 0.81 (Figure 2C). A heatmap generated via Metaboanalyst revealed that features (m/z – RT values) assigned as putative metabolites from arachidonic acid metabolism were in greater abundance in HCDL than in HCNDL group (Figure 2D). The linoleic acid (LA, a second ω-6 PUFA) metabolism was also significantly enriched (p<0.05) with an IV of 0.34 (Figure 2C and Supplementary Table 3). The linoleic acid (LA, a second ω-6 PUFA) metabolism was also significantly enriched (p<0.05) with an IV of 0.34 (Figure 2C and Supplementary Table 3). In contrast, the metabolism of the ω-3 PUFA α-linolenic acid (ALA) was not significantly enriched (Figure 2C, and Supplementary Table 3), but yielded a significant IV of 0.2 (Figure 2C). To further interrogate the ω-3 PUFA metabolic pathway, putative features corresponding to precursors of ω-3 PUFA metabolism (eicosapentaenoic acid [EPA] and docosahexaenoic acid [DHA]) and downstream lipid mediators (17(S)-hydroperoxy docosahexaenoic acid, D5-D6 resolvins, MaR1 and D1-D4 resolvins) were extracted from the LC-MS data and their relative abundances compared between the HCDL and HCNDL groups (Supplementary Table 4). This analysis further indicated increased ω-3 PUFA metabolism in the HCDL group.

Figure 2. HCs that progressed to leprosy display distinct serum metabolomic profiles compared to HCs that remained healthy.

Principal component analyses were performed with features obtained by LC-MS and that yielded a log₂ FC ≥ 1.0 and p<0.05 between the HCDL (n=17) and HCNDL (n=17) groups. PCA was performed separately on features collected in the negative ionization (A) and positive ionization (B) modes. (C) Pathway analysis via Metaboanalyst ²⁵ was performed using the m/z values of all 3245 features displaying a log₂ FC ≥ 1.0 and p<0.05 between the HCNDL and HCDL groups. Metabolic pathways with −log₁₀(p) value > 3.0 (left) and/or impact values > 0.10 (right) are shown. (D) Heatmap generated via Metaboanalyst with peak areas of LC-MS features (m/z - RT) assigned to arachidonic acid metabolism (see Supplementary Table 6). The −log₁₀(p) and impact values for all pathways can be found in Supplementary Table 3.

Confirmation that AA and ALA metabolism is altered in HCs that developed leprosy.

To confirm that AA metabolism was altered in HCs that developed leprosy, the presence of AA and two products derived from AA, LTB₄ and 11-hydroxyeicosatetranoic acid (11-HETE) were first validated in pooled HCs serum through LC-MS/MS. These data demonstrated that the targeted feature at m/z 303.23 [M-H]⁻ and the AA standard yielded the same retention time (20.59 min) and nearly identical fragment ions (Supplementary Figure 1). Likewise, the retention time and MS/MS spectra of the features at m/z 335.22, 16.90 min and m/z 319.22, 18.36 min corresponded to standards of LTB₄ and 11-HETE, respectively (Supplementary Figures 2 and 3). The primary precursors of ω-3 PUFA metabolism, EPA and DHA were observed as the features m/z 301.21, 19.97 min and m/z 327.23, 20.34 min, respectively, and corresponded to the commercial standards (Supplementary Figures 4 and 5). Based on the validated structures, the average of normalized peak areas corresponding to these five metabolites were extracted from the LC-MS data of individual samples and compared across groups. All five metabolites showed significant increase in the HCDL group with log₂ FC values (HCDL vs HCNDL) ranging from 1.79 to 4.00 (Table 1).

Table 1.

Lipid mediator’s structures confirmed by MS/MS spectra and their respective log₂ fold change (FC) values between HCDL and HCNDL comparison

Target Compounds (m/z - RT) a	Relative abundance between HCDL and HNCDL b
	log2FC	p values
Metabolism of ω-6 PUFAs
Arachidonic acid (303.23 – 20.59)	+2.43	<0.0001
Leukotriene B4 (335.22 – 16.90)	+2.18	0.0044
11-hydroxyeicosatetranoic acid (319.22 – 18.36)	+4.00	0.0007
Metabolism of ω-3 PUFAs
Eicosapentaenoic acid (301.21 – 19.97)	+2.28	<0.0001
Docosahexaenoic acid (327.23 – 20.34)	+1.79	<0.0001

Abbreviation: RT, retention time.

^aThe identification of the targeted metabolites were made in pooled sera by LC-MS/MS. The observed (m/z - RT) values are for the commercial standards used to identify the targeted metabolite in pooled sera.

^bThe comparison of each targeted metabolite between HCDL and HCNDL. LC-MS of individual samples was performed and the mean relative abundances (peak area calculated by XCMS software) for each metabolite was calculated. The fold changes between HCDL and HCNDL were log₂ transformed. Targeted metabolites that differed between the HC groups with log₂FC ≥ +1.0 and p<0.05 were considered as significantly altered.

HCs individuals that progressed to leprosy presented increased levels of lipid mediators derived from both ω-6 and ω-3 PUFA metabolism.

The metabolism of ω-6 and ω-3 PUFA metabolism result in biologically active pro-inflammatory or specialized pro-resolving lipid mediators ^{23, 24}. To gain further insight into the changes associated with these two classes of lipid mediators, the serum levels of the pro-inflammatory (LTB₄, PGD₂ and PGE₂) and pro-resolving (LXA₄, RvD1, RvD2; and MaR1) lipid mediators were measured by EIA in each individual serum sample and compared between the HCNDL and HCDL groups. Like the data obtained for LTB₄ by LC/MS, the EIA data demonstrated a significant increase (log₂FC= +1.70; HCDL vs HCNDL) of LTB₄ in the HCDL patient population (Figure 3A). The serum levels of PGD₂ were also significantly increased (log₂FC= +3.00; Figure 3B) in the HCDL group in comparison to HCNDL. In contrast the levels of PGE₂ were reduced in HCDL (log₂FC= −1.57; Figure 3C). The distribution of the concentration values of the pro-inflammatory lipid mediators within HCDL sera group did not differ between HCDL individuals that developed leprosy within 1 year (black circle) as compared to those that manifested the disease after 2–10 years (gray square) (Figure 3). A high proportion of individuals (41.2%) in the HCDL group developed the indeterminate form of leprosy. However, this subset of individuals was not driving the observed changes in the pro-inflammatory lipid mediators (Supplementary Figure 6).

Figure 3. HCDL individuals exhibited increased levels of pro-inflammatory lipid mediators.

Serum levels of (A) LTB₄; (B) PGD₂ and (C) PGE₂ were measured by enzyme-linked immunoassay (EIA) for HCDL (n=17) and HCNDL (n=17) individuals. In the HCDL group the black circles and grey squares represent individuals that developed leprosy withing one year after diagnosis of the index cases, and 2–10 years after diagnosis of the index cases, respectively. *adjusted p value less than 0.05.

Significant differences were also observed for the levels of specific pro-resolving lipid mediators. Specifically, the levels of LXA₄, RvD1 and, MaR1 were significantly higher in HCDL individuals versus HCNDL (Figure 4 A, B and C) with log₂ fold differences of +1.63, +2.86 and +0.76, respectively. RvD2 followed the same trend as the other pro-resolving lipid mediators but did not meet statistical significance (Figure 4D). As observed for the pro-inflammatory lipid mediators, no difference in the distribution of the values between HCDL individuals that developed leprosy after 1 year (black circle) from those individuals that manifested the disease after 2–10 years (gray square) were observed. The high proportion of HCDL individuals that developed the indeterminate form of leprosy were also not responsible for the higher levels of the pro-resolving lipid mediators within this group (Supplementary Figure 6). Altogether, the results presented indicate that several lipid mediators exhibited increased circulating levels in HCDL individuals and therefore the levels of these compounds could be used to differentiate HCs individuals that will progress to active disease from those that will remain healthy.

Figure 4. The HCDL individuals exhibited increased levels of the pro-resolving lipid mediators.

Serum levels of (A) LXA₄, (B) RvD1, (C) MaR1, and (D) RvD2) were measured by enzyme-linked immunoassay (EIA) for HCDL (n=17) and HCNDL (n=17) individuals. In the HCDL group the black circles and grey squares represent individuals that developed leprosy withing one year after diagnosis of the index cases, and 2–10 years after diagnosis of the index cases, respectively. *adjusted p value less than 0.05.

PGD₂ exhibits a positive correlation with anti-PGL-I titers.

The circulatory levels of the lipid mediators determined by EIA were assessed for relationships or correlations to the clinical characteristics of HCs individuals (Supplementary Table 5). For these analyses, the HCNDL and HCDL groups were not separated, but considered as a single group. Based on a Welch’s T-test no significant difference existed between genders and the levels of the individual lipid mediators. Using a Spearman’s correlation there was also no correlation between lipid mediators levels and age. There was a positive correlation between PGD₂ and anti-PGL-I IgM levels (ρ=+0.75) and bacillary index of the index case (ρ=+0.58). LXA₄ also showed a moderate correlation with anti-PGL-I IgM levels (ρ=+0.54).

The levels of lipid mediators provide a potential tool for the early detection of leprosy.

To evaluate the ability of the seven lipid mediators measured by EIA to differentiate between HCDL and HCNDL individuals PCA and PCA biplots were performed using the circulating levels of the lipid mediators. The PCA displayed two primary clusters, most of the HCDL samples (n=12) clustered together, with four HCDL samples clustering with the HCNDL group, and only one of the HCNDL sample clustering with the HCDL group (Figure 5A). Interestingly, the PCA biplot revealed that PGE₂ was the lipid mediator that influenced the clustering of HCNDL samples, while the other lipid mediators (LTB₄, LXA₄, PGD₂, RvD1, RvD2, and MaR1) drove the clustering of HCDL samples (Figure 5B). Finally, we constructed a logistic model to identify those lipid mediators that could classify HCNDL and HCDL samples (Figure 5C). A set of three lipid mediators (RvD1, RvD2, and PGD₂) with the potential to predict HCs that would progress to leprosy best fit the classification model. The model also displays correlations between individual lipid mediators within the HC groups (Figure 5C). Specifically, before the manifestation of the active disease, HCDL individuals generally had higher levels of PGD₂ (x-axis) and RvD1 (symbol size) in comparison with HCNDL individuals. We also observed two subsets within the HCDL group. One subset was defined by lower RvD2 (y-axis) and PGD₂ levels. The other subset was more distributed but could be defined by either elevated levels of RvD2 or PGD_2, or elevated levels of both lipid mediators.

Figure 5. Serum levels of the lipid mediators allow unsupervised and supervised separation of HCDL and HCNDL individuals.

(A) Serum levels of the lipid mediators LTB₄, LXA₄, PGD₂, PGE₂, RvD1, RvD2, and MaR1 were applied to PCA) PCA data points corresponding to HCNDL and HCDL individuals are indicated with orange circles and purple triangles, respectively. (B) A PCA biplot indicates that PGE₂ negatively correlates to the other lipid mediators and primarily drives the clustering of the HCNDL individuals. (C). Visualization of a logistic model for classifying HCNDL and HCDL based on the PGD₂ (x-axis), RvD2 (y-axis), and RvD1 (dot size proportional to the amount of RvD1) levels in HCs’ sera.

Discussion

The current study utilized a metabolomics approach to assess biochemical differences between asymptomatic M. leprae infection in HCs that developed leprosy or not in a 10-year period following exposure to the index case. These findings demonstrated that at the time of index case detection the serum metabolic profile of HCDL individuals differed in comparison to HCNDL individuals. Multiple metabolomics studies have evaluated metabolic differences among the presentations of leprosy and associated immunologic reactions^18–21. Common among these previous findings is the enrichment of metabolic pathways associated with the metabolism of PUFAs and the resulting lipid mediators that are known to modulate the inflammatory response ^25–31. Thus, it was not surprising that the most significantly enriched metabolic pathways between HCDL and HCNDL individuals were arachidonic acid and linoleic acid metabolism. However, what was remarkable is that these metabolic differences were present months to years before the susceptible individuals developed active leprosy. We also note that the lipid mediators targeted in our studies did not separate HCDL individuals based on whether they developed leprosy within 1 year or 2–10 years of index case diagnosis. The frequency of HCDL individuals (41%) that developed the indeterminate form of leprosy in our study is at least 3.7x higher than in previous reports (11–26%) ^{32, 33}. This potentially important factor was investigated; however, the levels of targeted lipid mediator within the HCDL group did not differentiate individuals that developed the indeterminate form of leprosy from others. Interrogation of clinical data further demonstrated no difference in the mean age, the proportion of male:female, or the number of BCG vaccination scars between HCNDL and HCDL groups, indicating that these factors possibly did not influence our metabolomics based results. However, we worked with a small sample size and therefore the influence of these factors cannot be excluded. Collectively, these data reinforce the hypothesis that lipid mediator metabolites can serve as a biomarker for risk of disease progression in all HCs.

Specifically, HCDL individuals exhibited higher levels of AA, 11-HETE, LTB₄, LXA₄, PGD₂, RvD1 and MaR1, but lower levels of PGE₂ in comparison with HCNDL subjects. These findings suggest that immune-inflammatory mechanisms differ at the asymptomatic stage of infection between HCs that presumably clear infection versus those that progress to disease. It is important to highlight, however, that complementary immunological data and extensive longitudinal sample collection for these same individuals are limitations of the current study. Nevertheless, the hypothesis that the metabolic and immunological states at the time of exposure to M. leprae correlate with an individual’s risk of developing leprosy should be pursued in future studies to elucidate the link between host predisposition to leprosy and the role of lipid mediators.

The observed increase in AA in HCDL individuals is consistent with a correlative increase in the downstream metabolic products LTB₄, PGD₂, and LXA₄. In contrast another AA derived lipid mediator, PGE_2, was lower in the HCDL patient group. Prostaglandins are produced via two isoforms of cyclooxygenase (COX); the constitutively expressed COX-1 and the inducible COX-2 ²⁴. Both enzymes convert AA into PGH₂, which is subsequently converted to PGE₂ and PGD₂ by PGE synthases and PGD synthases ²³. Based on this pathway we hypothesize that HCDL individuals at the time of M. leprae exposure will exhibit increased PGD₂ synthase or decreased PGE₂ synthase expression and activity, thus shunting PGH₂ to PGD₂ production and away from PGE₂. Although LTB₄ and LXA₄ are pro-inflammatory and pro-resolving lipid mediators, respectively, the formation of both depends on 5-lipoxygenase (5-LO) activity that converts AA to the shared precursor, LTA₄ ²³. Hence, increased expression/activity of 5-LO, could explain the increased circulatory levels of LTB₄ and LXA₄ in HCDL individuals. This hypothesis is reinforced by the higher levels of RvD1 observed in HCDL individuals as RvD1 formation is also dependent on 5-LO activity ³⁴. Increased levels of LXA₄ could also be reflective of increased 15-LO expression/activity that converts AA to15-HPETE, which is then used to form LXA₄ via 5-LO. Regarding 11-HETE, the increase of this compound observed in HCDL sera could result from the action of COX-2 ³⁵ on AA, or from the non-enzymatic oxidation of AA ³⁶.

Higher levels of pro-resolving RvD1 and MaR1 in HCDL serum agreed with the increased levels of the ω-3 PUFA, DHA. We also observed a trend for the increase in RvD2 levels in HCDL samples, but the difference between HCNDL and HCDL was not significant. MaR1 is produced by macrophages, a primary cellular target of M. leprae, via 12-LO that converts DHA into MaR1 ³⁷. The higher levels of MaR1 in HCDL individuals suggest that 12-LO expression/activity in macrophages is potentially linked with the susceptibility to leprosy. Likewise, increased RvD1 in this same patient group would indicate either increased COX-2 or 15-LO activity as both result in the conversion of DHA to 17-hydroperoxydocosahexaenoic acid (17-H(p)DHA), the initial precursor in RvD1 formation. E-series resolvins (RvEs) were not investigated in the current work but the increased levels of their precursor EPA ²⁴ in HCDL samples suggest that the circulatory levels of RvEs would possibly be elevated during asymptomatic M. leprae infection.

In addition to the above biochemical explanation of lipid mediator differences among the HCs of leprosy patients, the current results offer biological insights to the protective or permissive host responses at early stages of M. leprae infection. The significantly higher levels of lipid mediators in HCDL samples may reflect a response to high bacterial loads in HCDL individuals at the asymptomatic stage of M. leprae infection. As noted, we observed that the serum level of anti-PGL-I IgM together with the bacillary load of the index cases distinguished HCDL from HCNDL individuals. Higher LTB₄ levels may promote the inflammatory response to M. leprae infection through the induction of chemotaxis and the production of cytokines derived from Th1 cells (IFN-γ, TNF-α and IL-12p70) ³⁰. However, it seems contradictory that HCs individuals susceptible to leprosy would exhibit increased levels of pro-inflammatory lipid mediators that promote host responses unfavorable to M. leprae. On the other hand, it is important to remember that LXA₄ and RvD1 were also increased in the HCDL individuals. Silva and Belisle ²⁴ discussed how LXA₄ and RvD1 might counteract the inflammatory effects of LTB₄ by down-modulating Th1-derived cytokines, stimulating the production of the IL-10, the activity of Treg and, blocking the production of LTB₄. Based on previous findings ^25–27. The authors also suggested that increasing levels of LXA₄ (and of RvD1 and MaR1) during subclinical M. leprae infection could result in a reduction of the levels of LTB₄ over time, favoring the onset of leprosy in HCDL individuals. In vitro and longitudinal clinical studies of M. leprae infection that further assess lipid mediator levels and the expression of key enzymes such as COX-2, 5-LO, 12-LO and 15-LO are needed to confirm this hypothesis.

PGD₂ is well-known for exhibiting a dual role during the inflammatory process. This prostaglandin was already shown to favor the production of IFN-β and inhibit the inflammasome activation during neurotropic coronavirus infection through DP1 receptor signaling ³¹. As shown with Histoplasma capsulatum infection of alveolar macrophage, PGD₂ can also reduce the secretion of IL-10, which would favor the Th1 response, and the expression of the leukotriene B₄ receptor BLT1 ²⁸. These findings point to a pro-inflammatory role of PGD₂. On the other hand, previous work showed that PGD₂ contributes to the resolution of the inflammation by inducing M2 polarization of macrophages in several acute inflammatory models ³⁸. The potential effects of PGD₂ during M. leprae infection have not been investigated, but Silva and Belisle ²⁴ suggested that PGD₂ might have an immunesuppressive role in lepromatous patients. In contrast, they also raised the possibility that the higher levels of PGD₂ in lepromatous patients could be linked with nerve damage since individuals with severe nerve damage, such as leprosy patients with type 1 reaction (T1R), displayed higher levels of PGD2 ²¹. More specifically, they suggested that PGD₂ might control the damaging effects that occurs during M. leprae infection since this prostaglandin participate in the nerve regeneration and remyelination as showed in previous reports ³⁹

It was interesting to observe that PGE₂ was reduced in HCDL individuals and significantly influenced the PCA clustering of the HCNDL individuals. Previous studies in leprosy suggest that PGE₂ decreases the efficiency of the host response against M. leprae ²⁴. These studies only provide a glimpse at the possible roles of PGE₂ in leprosy. In contrast, it was found that PGE₂ produced by dendritic cells treated with the M. tuberculosis rpfE protein could induce strong Th1 and Th17 cell responses ⁴⁰. Furthermore, Kaul and collaborators demonstrated that a knockout-mouse for EP2, one of the four receptors that recognize PGE₂ (EP1, EP2, EP3 and EP4), was more susceptible to aerosol infection with M. tuberculosis than wild-type mice due to a higher activity of Treg and Th17 cells ⁴¹. These studies with M. tuberculosis potentially indicate that PGE₂ enhance the host response against mycobacteria at early stages of infection. Hence, reduced levels of PGE₂ early in M. leprae infection might promote an environment permissive to active disease.

The measurement of lipid mediators in HCs also holds potential as a biomarker of risk of leprosy. PCA and PCA biplots of circulatory lipid mediator levels assessed by EIA allowed differentiation of HCDL and HCNDL individuals. In addition, a classification model suggested that the combination of PGD₂, RvD1 and RvD2 might serve as a biomarker for risk of leprosy. However, this model was made with a small sample size and therefore is subject to over-fitting. Correlations between the serum levels of the lipid mediators and the clinical data of HCs individuals were investigated. Previous findings showing that the risk of developing leprosy among HCs positively correlates with the bacillary load of the index case ⁷ and the anti-PGL-I IgM titer of the HC ⁴². Our current studies corroborated these findings and led us to speculate that metabolic differences among HCs may be related to increased M. leprae exposure in the HCDL group. Indeed, our correlation analyses reinforced this idea as a positive correlation of PGD₂ levels with anti-PGL-I IgM levels was observed. Additional studies with larger sample cohorts are needed to confirm this correlation and to further establish the link between bacillary loads/exposure and metabolic changes in HCs individuals, as well as to further investigate and test the generalizability of predictive models.

Overall, the current study opens new possibilities for the development of prognostic biochemical tests capable of assessing an individual risk for developing leprosy. Our findings alsio suggest an important role of the pro-resolving lipid mediators in determining the disease outcome during the early stages of M. leprae infection. There is a lack of in vitro studies that link M. leprae infection of macrophages and Schwann cells with increased levels of ω-3 and ω-6 PUFA-derived lipid mediators. Such in vitro studies are needed to 1) elucidate molecular mechanisms leading to lipid mediator production, 2) understand the association between lipid mediator production and high versus low bacillary exposure, and 3) to demonstrate which lipid mediators result in a protective versus permissive micro-environment for the pathogen. Further in vitro analyses will help establish how lipid mediator responses are influenced by or influence the production of other host products, such as cytokines, during an M. leprae infection.

Methods

Ethics statement

All individuals were recruited on a volunteer basis. Written informed consent was obtained from all subjects. The institutional review boards of Colorado State University and the Oswaldo Cruz Foundation (CAAE number: 34239814.7.0000.5248) approved the use of serum samples for the reported study. In addition, this research was conducted using approved ethical protocols that were in accordance with the Declaration of Helsinki.

Household contacts and serum samples

All serum samples were obtained from HCs that attended at the Souza Araujo Outpatient Unit (Leprosy Laboratory, Fiocruz, Rio de Janeiro, Brazil). Sera were stored at −80°C until subsequent use in the laboratory. Specifically, sera from HCs who never developed leprosy (HCNDL, n=17) and from HCs who manifested clinical symptoms (HCDL, n=17) were collected at the time of the diagnosis of the index cases, considered as time zero (T0). The samples obtained from HCDL individuals were collected within the period of August,1996 – July, 2011 and the serum samples from HCNDL subjects were obtained between May, 2011 – December, 2011. Those contacts that presented previous illness due to cancer, tuberculosis, or any other type of infectious contagious disease were removed from the study, as well as pregnant, lactating and puerperal women. Prior to sample collection, specialized dermatologists and neurologists examined the HCs individuals to check for clinical signs of leprosy. None of individuals in either group exhibited signs or clinical symptoms of leprosy at T0. All HCs individuals were instructed to return to the clinic in case of the onset of leprosy symptoms and/or skin lesions and for a routine follow-up within a year post-T0. At the routine follow-up examination, none of HCNDL individuals were diagnosed with leprosy symptoms. In contrast, HCDL individuals developed leprosy symptoms within 1–10 years after T0. To confirm that HCNDL individuals had not developed leprosy, a search in the Brazilian Disease Notification System (SINAN) in the period from January 2007 to December 2020 was performed, as described in Supplementary methods. None of HCNDL subjects were identified as leprosy patients. Leprosy diagnosis and classification of the index cases and the HCDL were established according to the Ridley and Jopling criteria ².

The clinical and demographic data for the HCs and their index cases are shown in Table 2, including the bacillary index (BI), the clinical form of leprosy and the levels of anti-PGL-I immunoglobulin M (IgM). Positive responses for the levels of anti-PGL-I IgM were defined as an O.D.₄₅₀ > 0.3 (cutoff). The cutoff was determined by the following formula A _O.D.450 + 2 × SD _O.D.450 in which A _O.D.450 is the O.D.₄₅₀ average obtained from endemic control sera and the SD _O.D.450 the standard deviation from the same set of samples. The endemic control samples (100 individuals) were collected from individuals that lived in an endemic area, Rio de Janeiro, with no history of leprosy or tuberculosis infection. Fisher’s exact test was used to evaluate whether a significant difference occurred between the proportion of males to females, the proportion of the clinical forms of the index case, and the number of BCG scars⁴³. To determine the statistical difference if the mean ages between HCNDL and HCDL, a two-tailed Welch’s t-test was performed. An unpaired Mann-Whitney was used to detect differences between HCNDL and HCDL in the levels of anti-PGL-I IgM. A paired t-test was also applied to compare the levels of anti-PGL-I IgM of HCDL individuals before-after the development of leprosy. The BI of the index cases was compared by a Mann-Whitney test. The values of p<0.05 were considered as statistically significant. Supplementary Table 1 shows the period of time that HCDL individuals took to manifest leprosy, the clinical form observed, the BI, and levels of anti-PGL-I IgM.

Table 2.

Demographic and clinical characteristics of household contacts and their index cases

ID	Sex	Age, years	Clinical form of the index casea	Bacillary index of the index case	Number of BCG scars b	Anti-PGL-I (OD450 value)c	Positive/N egative for anti-PGL-I
Household contacts that did not develop leprosy (HCNDL)
HCNDL01	Female	62	BT	0	ns	0.113	negative
HCNDL02	Female	66	BL	1.00	ns	0.19	negative
HCNDL03	Male	27	LL	5.50	2	0.292	negative
HCNDL04	Male	55	BL	1.00	ns	0.011	negative
HCNDL05	Male	50	TT	0	1	0.028	negative
HCNDL06	Female	62	BL	1.00	ns	0.466	positive
HCNDL07	Female	31	BL	0.50	1	0.15	negative
HCNDL08	Female	55	BL	1.00	1	0.031	negative
HCNDL09	Male	60	BT	0	ns	0.008	negative
HCNDL10	Male	22	BL	0.50	2	0.029	negative
HCNDL11	Female	45	BL	4.25	1	0.121	negative
HCNDL12	Female	21	BT	0	2	0.100	negative
HCNDL13	Female	26	BL	1.00	2	0.034	negative
HCNDL14	Male	42	BL	1.00	ns	0.055	negative
HCNDL15	Female	48	BL	1.00	1	0.149	negative
HCNDL16	Female	65	LL	4.25	1	0.079	negative
HCNDL17	Male	32	BL	1.00	1	0.050	negative
Household contacts that developed leprosy (HCDL)
HCDL01	Female	54	BL	1.83	1	0.13	negative
HCDL02	Female	57	BL	3.33	ns	0.46	positive
HCDL03	Female	67	BL	3.00	ns	0.22	negative
HCDL04	Male	24	LL	2.16	1	0.390	positive
HCDL05	Female	24	LL	2.50	2	0.840	positive
HCDL06	Male	21	LL	4.75	1	0.470	positive
HCDL07	Male	48	LL	4.25	ns	0.63	positive
HCDL08	Male	21	BL	3.00	2	0.410	positive
HCDL09	Male	18	LL	4.16	1	0.32	positive
HCDL10	Female	34	BB	1.33	ns	1.26	positive
HCDL11	Female	37	LL	3.66	1	0.50	positive
HCDL12	Male	55	LL	4.25	ns	0.51	positive
HCDL13	Female	31	LL	4.66	ns	0.200	negative
HCDL14	Female	42	BL	3.00	ns	0.340	positive
HCDL15	Female	23	BB	1.75	1	0.179	negative
HCDL16	Female	39	BL	1.50	ns	0.350	positive
HCDL17	Female	36	LL	4.25	1	0.114	negative

Abbreviations: na - not available; ns - no scars were found it; TT - tuberculoid leprosy; BT - borderline tuberculoid; BB - borderline borderline; BL - borderline lepromatous; LL - lepromatous leprosy. The sera samples were collected from the household contacts (HCDL and HCNDL) at the diagnosis of index case (considered as time zero). According to the National Notifiable Conditions System database (SINAM; ministry of health of Brazil) none of the HCNDL individuals exhibited clinical signs or symptoms of leprosy during the period of January/2007 – December/2020.

^aclinical manifestations of leprosy were classified according to Ridley-Jopling criteria ².

^bSimilar to many previous studies, we used the number of BCG scars as a proxy for BCG vaccination record (Supplementary methods).

^cSera collected were assayed for the presence of anti-phenolic glycolipid 1 immunoglobulin M by a specific enzyme-linked immunosorbent assay. The antigen used in ELISA was NT-P-BSA (synthetic native trisaccharide of PGL-1 coupled to BSA through a 3-phenylpropanoyl). A cutoff value of 0.30, at an optical density at 450 nm (OD₄₅₀), was set for positive responses.

Liquid chromatography-mass spectrometry (LC-MS)

An aliquot of serum (80 μL) was used for metabolite extraction as described in Supplementary methods. Extracted metabolites were separated by reversed-phase chromatography on a C18 column (XBridge BEH C18 XP; 2.5 μm particle size, 2.1mm ×100 mm) (Waters Millford, MA, USA). The solvent gradients are described in the Supplementary methods. MS ionization of eluting metabolites was performed in negative and positive ion modes (see Supplementary methods). Each serum sample was analyzed three times for both ion modes. The order of sample analysis, including the triplicate injections, was randomized. Quality control (QC) serum was used to monitor the stability of the system and to check the reliability of the features (i.e., LC-MS signals with specific retention time and mass to charge ratio m/z) (additional detail in Supplementary methods).

LC-MS data processing and statistical analyses

Raw LC-MS data were converted using ProteoWizard MS Convert version 3.0.6478 64 bit. Retention-time correction, chromatogram alignment, and features annotation were performed using XCMS software version 1.46 in R version 3.2.2 ⁴⁴ (for additional details see supplementary methods). After data processing, the relative abundances of all features were normalized by the median fold change ⁴⁵ using the median of QC serum as a reference sample. For statistical analysis, a final set of features were selected based on several filtering steps (Supplementary methods). Before statistical analyses, missing values were imputed with half of the overall minimum value for all normalized intensities found in the final dataset. The feature peak area for each biological sample was summarized by the mean of the triplicate values and converted to log₂. These features were compared using a linear models’ approach with t-statistics having an empirical Bayes moderation, implemented in the R package limma ^{46, 47}. The p values were adjusted for false discovery rate ⁴⁸. Features that differed between the HC groups with log₂FC ≥+1.0 (HCDL - HCNDL) and p<0.05 were considered as significantly altered. Principal component analyses (PCA) were conducted on scaled and centered values of all features that remained after filtering.

Metabolic pathway analysis and metabolite identification

The accurate mass of each feature with a log₂FC and p<0.05 in the comparisons was queried against the Human Metabolome Database (HMDB, www.hmdb.ca) (mass error tolerance of +/− 10 ppm) ⁴⁹. The metabolite lists generated by HMDB were imported into Metaboanalyst ²² (http://www.metaboanalyst.ca/) to map them against the Kyoto Encyclopedia of Genes and Genomes (KEGG) database to perform pathway analysis. Metaboanalyst combines over-representation (ORA) and topology analysis to determine which metabolic pathways are the most disturbed in a disease or experimental condition ²². The ORA analysis uses a given list of significantly altered metabolites to test whether a particular metabolic pathway is represented more than expected by chance within the input list ⁵⁰. ORA analysis was performed by hypergeometric tests and metabolic pathways with adjusted p values <0.05 were considered as significantly enriched (Holm-Bonferroni correction). The topology analysis uses the structure of a given metabolic pathway to evaluate the relative importance of the metabolites in a pathway based on their relative locations and is represented by the pathway impact value (IV); also known as “Pathway Impact Score” ⁵¹. Topology analysis was based on the relative betweenness centrality ⁵². The pathway IV threshold was set to > 0.10 ⁵¹. IVs closer to 1.0 indicate a more perturbed pathway.

Metabolite identification by LC-MS/MS

For metabolite structural confirmation, individual serum samples were pooled and submitted to LC-MS/MS analysis. The tandem MS (MS/MS) fragmentation spectrum obtained for each peak was compared with the MS/MS data available in the Metlin database or compared to the experimentally acquired fragmentation spectrum of a commercial standard (see Supplementary methods). Pooled serum samples were also spiked with commercial standards to confirm that the targeted peak exhibited similar LC retention time of the commercial standard.

Enzyme-linked immunoassay (EIA)

To investigate the circulatory levels of LTB₄, prostaglandin D₂ (PGD₂), prostaglandin E₂ (PGE₂), RvD1, resolving D2 (RvD2), and maresin-1 (MaR1), EIA kits were purchased from Cayman Chemical. The LXA₄ EIA kit was from Neogen (Lexington, USA). For EIA assays, serum samples that had never been thawed and had been stored at −80°C were used. LXA₄, RvD1, RvD2, and MaR1 were extracted from serum samples using C₁₈ Sep-Pak columns (Waters; Elstree, UK) before the assay. Extraction and EIAs were performed following the manufacturer’s protocol. Lipid mediators were compared between HCNDL and HCDL by using univariate linear mixed models to account for multiple subjects from the same household. The p-values were adjusted for multiple comparisons (Holm-Bonferroni adjustment ⁵³) within each type of comparison. To stabilize variance where necessary, the circulatory concentrations of the lipid mediators were transformed by log₂ using the caret package version 6.0–88 to optimize parameters ⁵⁴, prior to the statistical analysis. Correlations between lipid mediators’ levels and the available clinical data were calculated by Spearman’s rank correlations due to outliers. Statistical analyses were conducted in R open software version 4.1.1. ⁵⁵. Logistic mixed models were also fitted to different combinations of lipid mediators to determine whether a select group of features could classify the groups, and the lipid mediators in the best-fitting model are reported here.