Chagas disease (CD) is a parasitic disease caused by the protozoan Trypanosoma cruzi and is one of the designated neglected tropical diseases [1]. T. cruzi is endemic to the Americas and infects 7–8 million people worldwide [1]. An estimated 300,000 infections have been recorded in the United States due to a large Latin American immigrant population and endemic transmission [2–4]. CD is primarily transmitted through triatomine insects of the Triatoma and Rhodnius genera [2,5]. Non-vectorial modes of transmission involve blood transfusion, transplacental transmission, and food and drink contaminated with T. cruzi [1]. The T. cruzi life cycle includes three main stages: epimastigotes, trypomastigotes and amastigotes. T. cruzi in the insect vector undergoes transformation from trypomastigotes to epimastigotes in the midgut, and then migrates to the hindgut and differentiates into infective trypomastigotes [1]. Upon triatomine defecation on the human host, the infective trypomastigotes enter the host through scratching or rubbing of the bite wound, or through eyes and mucosal surfaces [1]. Following mammalian host cell infection, trypomastigotes differentiate into amastigotes, which proliferate and subsequently transform into trypomastigotes [1].

CD has two disease stages: acute and chronic [1,2,5]. The acute stage, with high parasite load, is usually asymptomatic, or presents with non-specific symptoms (fever, malaise) [1,2,5]. A minority of infected individuals (20–30%) will then progressively develop clinical manifestations of chronic CD, including cardiomegaly, cardiac arrhythmias, apical aneurysms, megacolon, and megaesophagus [2]. In contrast to acute stage CD, chronic CD presents with low to no parasitemia [5]. T. cruzi infections are treated with either benznidazole or nifurtimox; however, these treatments cause significant adverse effects, to the point that up to 30% of treated individuals fail to complete the full treatment course [6,7].

CD was previously considered to have an autoimmune etiology, but parasite persistence has now conclusively been demonstrated to be required for disease pathogenesis [8]. Along with parasite persistence, chronic pro-inflammatory responses, including cytokine release and CD8+ T cell- mediated cytotoxicity, contribute to tissue damage [9]. A heterogeneity of interacting parasite-host factors, including T. cruzi strain, load and tissue tropism, host genetic background, and mode of infection, influence the clinical outcomes of the disease [10,11]. However, CD pathogenesis is not yet completely understood [2]. A holistic understanding of the molecular pathways involved in disease progression could help identify new drug development avenues.

Metabolites are the final products of mRNA and protein expression and of protein activity, thus providing information closely linked to phenotype [12]. Metabolic pathways are druggable. They also change dynamically in response to disease [13,14]. As such, an improved understanding of metabolism in CD may lead to new avenues for drug development and CD patient monitoring. Acute T. cruzi-infection affects in vitro and in vivo host metabolic pathways, including decreasing mitochondrial oxidative phosphorylation-mediated ATP production [9,15–17]. In addition, acutely T. cruzi-infected mice heart tissue and plasma showed significant changes in certain metabolic pathways, such as glucose metabolism (glucose levels elevated in heart tissue and lowered in plasma over time), tricarboxylic acid cycle (TCA) (decrease in select TCA metabolites in the heart tissue and a decrease in all detected TCA metabolites in plasma), lipid metabolism (increased long-chain fatty acids in the heart tissue and decreased long-chain fatty acids in plasma), and phospholipid metabolism (high accumulation of phosphocholine precursor metabolites in the heart) [15]. Prior analysis of hearts from acutely infected mice also showed that cardiac metabolite profiles reflected disease severity, with changes in cardiac acylcarnitines and glycerophosphocholines predictive of acute infection outcome [9]. Metabolomic analysis of chronic CD has been limited to serum and gastrointestinal tract samples [18,19]. Serum analysis demonstrated significant changes in amino acid and lipid metabolism, particularly acylcarnitines, sphingolipids, and glycerophospholipids [19]. Analysis of GI tract samples observed persistent metabolic perturbations in the oesophagus and large intestine in chronic CD, including infection-induced elevation of acylcarnitines, phosphatidylcholines and amino acid derivatives [18]. However, metabolic changes in the heart may differ from those in the circulation or in the GI tract [15]. It is therefore essential to perform metabolomic analysis of tissues collected from the heart in chronic CD.

Many sudden fatalities due to chronic cardiac CD are often attributed to apical aneurysms which occur at the bottom of the heart [20,21]. Importantly, parasite load is low, spatially heterogeneous and poorly correlated to the magnitude of tissue damage in chronic CD, including in clinical samples [22,23], indicating possible spatial disconnect between CD-induced metabolic alterations and tissue parasite load. We therefore focused on liquid chromatography-tandem mass spectrometry-based metabolomic analysis of transversely sectioned hearts from mice chronically infected with T. cruzi strains CL and Sylvio X10/4. These samples were previously collected but only previously analyzed with regards to parasite load and differences in overall metabolome between heart apex and heart base (infected and uninfected samples combined), without characterizing the impact of chronic infection on the heart metabolome [9]. In contrast, this study focuses on metabolic changes associated with chronic (90 and 147 days) T. cruzi strains CL and Sylvio X10/4 infection, compared to matched uninfected controls. Overall, we observed significant localized chemical differences associated with infection, with a disconnect between parasite localization and overall positional metabolic perturbations. Our data also showed infection-induced variations in acylcarnitine and glycerophosphocholine chemical families.

The purpose of this study was to compare the metabolic impact of chronic CD in the mouse heart between divergent T. cruzi strains and between cardiac regions. To do so, we analyzed previously-collected positive mode LC-MS/MS data [9]. While this prior study focused on the impact of acute infection (12 days post-infection in mice) on the cardiac metabolite profile, here we specifically focused on the impact of 90 and 147day infection (chronic stage CD in mice) on the cardiac metabolite profile, which had not been studied before.

We observed a clear localized impact of T. cruzi infection on the overall metabolite profile (Figs 1, S2 and S3). As previously described [9], parasite burden was highest at the base of the heart (position A) for strain CL and central positions (position C) for strain Sylvio X10/4 (Fig 1A). PERMANOVA analysis indicated that the highest significant perturbation in the overall metabolite profile occurred at central positions for strain CL infection (PERMANOVA analysis of Bray-Curtis-Faith distance matrix R² = 0.20813, p-value = 0.004 at position C) and at apical positions for strain Sylvio X10/4 infection (PERMANOVA R² = 0.27923, p-value = 0.014 at position D) (Fig 1B). Strikingly, in both cases chemical disturbance was greatest at sites distinct from the site of highest parasite burden, which corroborates our observations in the context of chronic gastrointestinal T. cruzi infection in mice [18]. The localization of chemical disturbance also provides a molecular mechanism explaining the apical aneurysms observed in CD patients and the fact that these happen even though cardiac tissue parasite burden is low [22].

To identify the specific cardiac metabolites spatially perturbed by infection, initially we built a random forest classifier for each position, each strain and each extraction method, comparing to uninfected matched control samples (S1–S4 Tables). We first assessed the overlap between the top-ranked most differential metabolites by random forest for the two different strains, at each position, as described in Methods. Limited overlap of these significant metabolites was observed between strains (Fig 2A–2D). To address the possibility that a given metabolite was modified by each strain at different positions, we also performed this analysis for all positions combined (Fig 2E and 2F, S5 and S6 Tables). Indeed, a higher overlap was observed between strains under these conditions, but still only representing a fraction of all differential metabolites. However, annotation of these differential metabolites using molecular networking through the GNPS platform [30] revealed that while differing in terms of m/z, many were part of the same chemical families, including acylcarnitines and glycerophosphocholines (S1–S6 Tables, S4 and S5 Figs).

Random forest classifier identified several acylcarnitines and glycerophosphocholines as impacted by infection (S1–S5 Tables). Both chemical families play a major role in several biochemical pathways. Carnitine serves as a shuttling mechanism for fatty acids, in the form of acylcarnitines, from the cytosol into the matrix of the mitochondria for beta-oxidation [38]. Glycerophosphocholines are major components of lipid metabolism, cell membrane structure, and choline production, the latter of which is essential for select amino acid and neurotransmitter synthesis [39,40].

Total acylcarnitines in the central positions of the heart were decreased by strain Sylvio X10/4 infection compared to the uninfected group (Fig 3A and 3B, Mann-Whitney p<0.05). A similar pattern was observed for total acylcarnitines following strain CL infection when compared to matched uninfected samples at the heart base (Fig 3A and 3B, Mann-Whitney p<0.05). Normal levels and distributions of acylcarnitines in the heart are represented by uninfected samples (Fig 3). Previous studies demonstrated that acylcarnitines of different lengths were associated with infection outcome in acute (12 days post-infection) T. cruzi mouse models [9]. Therefore, we sought to understand how different length acylcarnitines were affected by chronic (90 and 147 days post-infection) infection. Acylcarnitines are classified based on the number of carbons in their fatty acid chain as short- (≤C4), mid- (C5 -C11), and long-chain (≥C12) acylcarnitines. In the case of CL strain infection, when compared to uninfected samples, short chain acylcarnitines were significantly decreased at the heart base (p<0.05 Mann-Whitney, Fig 3C and 3D). Strain Sylvio X10/4 infection significantly decreased mid-chain acylcarnitine at all positions (Mann-Whitney p<0.05, Fig 3E and 3F) and significantly decreased long-chain acylcarnitines at heart base, center and apex (Mann-Whitney p<0.05, Fig 3E–3H).

Select glycerophosphocholines were increased at specific sites upon infection. CL strain infection significantly increased total glycerophosphocholines, at central heart positions compared to uninfected samples (Mann-Whitney p<0.05), as did strain Sylvio X10/4 at the heart apex (Mann-Whitney p<0.05, Fig 4A and 4B). Further analysis based on glycerophosphocholine m/z was performed, because previous studies showed differences in glycerophosphocholine m/z range between fatal and non-fatal acute mouse infection [9]. Glycerophosphocholines were categorized into three mass ranges: short (400–599.99 m/z), mid (600–799.99 m/z), and long (>800 m/z). Significantly increased glycerophosphocholines were observed for CL strain infection in short (m/z 401–599.99) glycerophosphocholines at heart center and apex (Mann-Whitney p<0.05, Fig 4C and 4D). Sylvio X10/4 strain infection significantly increased short (m/z 401–599.99) and long (m/z>800) glycerophosphocholines at the heart apex when compared to uninfected samples (Mann-Whitney p<0.05, Fig 4C, 4D, 4G and 4H).

There are 7 T. cruzi discrete typing units infectious to humans (DTUs TcI—TcVI and Tcbat). These DTUs, while still currently considered the same species, nevertheless present significant genetic differences [41,42]. Select T. cruzi strains are also more virulent than others or require a greater dose to establish infection. The inocula used in this study for both strains are standard and based on their relative degree of acute-stage lethality [9,43]. Specifically, strain Sylvio X10/4 is a low virulence strain, necessitating a higher infectious inoculum to observe cardiac pathology at chronic timepoints [43]. In contrast, susceptible mice infected with 1000 CL trypomastigotes survive the acute stage to develop chronic cardiac symptoms, while mice infected with higher inocula only survive a few weeks [9]. Timing and magnitude of induced disease may also differ between T. cruzi strains. Ninety days timepoint for strain CL infection enables comparison with Hossain et al, which analyzed impact of infection on the gastrointestinal metabolome at 89 days post-infection [18]. In addition, the different doses and timepoints enabled us to find commonalities across infection systems, mimicking the clinical situation where patients do not know how long ago they were infected, or with which strain. Cross-strain, -dose and -timepoint comparisons are thus important to guide drug development.

However, pathogenic processes are overall similar in cardiac CD across T. cruzi strains, with accumulation of fibrosis and inflammation [42,44]. These similarities are reflected in the common metabolomic changes observed for strain Sylvio X10/4 (TcI) and strain CL (TcVI)-infected heart tissue in this study, including chronic infection-induced increases in glycerophosphocholines and decreases in acylcarnitines. These results concur with independent findings with regards to select acylcarnitines in the serum of CD1 mice chronically infected with T. cruzi strain Brazil (TcI) and to short-chain acylcarnitines and glycerophosphocholines in the heart of BALB/c mice acutely infected with T. cruzi strain Y (TcII) [15]. They also concur with the negative relationship between mid-chain acylcarnitines vs fibrosis and disease progression-associated cytokine PDGF, and the positive relationship between glycerophosphocholines in mass ranges 400–499, 500–599 and over all mass ranges vis-à-vis of inflammation, fibrosis and progression-associated cytokines in the heart of BALB/c mice chronically infected with T. cruzi strain H1 (TcI) [45].

Differences in pathogenesis between strains may be due to differential strain tropism. Indeed, TcI strains tend to produce cardiomyopathy, while TcVI strains commonly produce megacolon and megaesophagus, although cardiomyopathy can still occur [46]. Our results indicate a disconnect between sites of highest parasite burden and sites of metabolic perturbation. Although parasite levels were highest in central heart segments following strain Sylvio X10/4 infection, we observed statistically significant perturbations in metabolism at the apex of the heart (Fig 1). Apical aneurysms are one of the major symptoms in chronic CD patients [47]. In addition, lateral heart wall damage is also common among chronic CD patients, in central regions of the heart [48], and we observed significant perturbations in cardiac metabolism at lower central heart positions in strain CL infection (Fig 1B) [48]. These results are consistent with clinical findings of low cardiac tissue parasite burden in CD in humans. Although parasite persistence is required for Chagas disease progression, nevertheless, cardiac tissue parasite load was not correlated with the magnitude of tissue damage, in patient-derived samples [22]. Based on these results, we propose a concept of spatial disease tolerance, whereby some tissue regions are more affected by infection, while others are less functionally affected. This is likely due to a combination of host and pathogen factors, given the differences we observe here between strain CL and strain Sylvio X10/4 infection in the same C3H mouse genetic background. Importantly, the localization of maximal metabolic perturbation in acute strain Sylvio X10/4 infection was also the heart apex, indicating that the spatial course of disease may be set early in CD [9]. Likewise, host factors likely contribute, such as the higher production of antiparasitic but tissue-damaging IFNγ at the heart apex or specific cardiac regions being more prone to microvasculature disruptions [9].

Our results also highlight the importance of considering metabolic changes at the level of chemical families, beyond just individual metabolites. While there was little overlap of highly significant metabolite m/z at each position between strains, most differential annotatable metabolites were from these two chemical families. McCall et al. described these two chemical families as discriminatory compounds between fatal and non-fatal acute T. cruzi infected heart tissue [9]. Considering acute stage infection progresses into chronic stage infection, it is not surprising that changes in the relative abundance of these molecules are also observed in chronic CD.

Glycerophosphocholines have been linked to coronary heart disease due to production of lysophosphatidylcholines and choline [39,49]. Increased acylcarnitine levels have been linked to cardiovascular disease as well as cardiac symptoms in non-CD cardiac disease [50,51]. However, our results show the opposite pattern for acylcarnitines compared to non-infectious heart disease, highlighting the need to specifically study CD rather than extrapolate from other cardiac conditions. Indeed, this divergence was also observed at the level of gene expression profiles: upregulation of lipid metabolism gene expression was observed in heart samples of human cardiac CD patients compared to controls, while downregulation was seen in non-infectious dilated cardiomyopathy patient samples [52]. Higher lipid metabolism would increase acylcarnitine catabolism and thus decrease overall acylcarnitine abundance. Decreased carnitine palmitoyltransferase and acetyltransferase levels, as observed by proteomic analysis of infected mouse heart tissue [53], may alternatively also contribute to the decreased acylcarnitine levels we observed. Interestingly, a few infection-perturbed metabolites were annotated as odd chain saturated fatty acids. While rare, odd chain saturated fatty acids have also been linked to protection from coronary heart disease, atherosclerosis and type II diabetes [54–56].

These results set a foundation for host-directed therapeutic development. CD may be particularly amenable to such treatment strategies, due to the contribution of host-mediated tissue damage to CD pathogenesis [1,8]. Indeed, we have previously shown that carnitine supplementation can be used to treat acute CD [18]. These findings also demonstrate a causal role in disease pathogenesis for the metabolic alterations observed in our studies. Our observation of decreases in cardiac acylcarnitines in chronic CD indicate that carnitine-based treatment regimens may also be useful to treat chronic CD. Importantly, the fact that acylcarnitines are affected in both chronic CL and Sylvio X10/4 infection suggests cross-strain applicability. Other studies have emphasized the impact of metabolism modulators on CD progression. High fat diet reduces parasite levels and increases survival in acute CD mouse models [57]. Treatment of acutely T. cruzi infected mice with metformin (a metabolic modulator used to treat diabetic patients) also led to an increase in overall survival rate and decreased blood parasitemia [58].

Due to the low parasite burden in chronic Chagas disease and instrumental limits of detection, we anticipate most if not all detected metabolites to be host-derived, supported by their detection in uninfected tissues. As such, this study is focused on the impact of T. cruzi infection on host metabolism. A further limitation is that many of the differential metabolites were not annotatable, as is usual in metabolomic studies [59]. Nevertheless, we were able to annotate metabolites affected by chronic infection that make up important host biochemical pathways. We only analyzed samples in positive mode due to the greater availability of positive mode reference libraries, leading to better annotation rates (35.4% in positive mode vs 10.2% in negative mode [18]). A further limitation is that free fatty acids do not ionize well under our instrument conditions, so that we cannot link the changes in PCs to other lipids [60]. Lastly, effect size calculations on our total acylcarnitine and glycerophosphocholine data indicate that we were only adequately powered to detect large differences (Hedges’ g >0.6–0.8).

Overall, our study highlights the importance of identifying overall differences but also positional metabolic differences associated with infection, and the need to study multiple T. cruzi strains. Results also show the strength of systematic chemical cartography in understanding disease tropism and how it differs from pathogen tropism. These results will serve as stepping stones for further CD drug development, something that is urgently needed.