Comparing in vivo bioluminescence imaging and the Multi-Cruzi immunoassay platform to develop improved Chagas disease diagnostic procedures and biomarkers for monitoring parasitological cure

Background Chagas disease is caused by the protozoan parasite Trypanosoma cruzi and is a serious public health problem throughout Latin America. With 6 million people infected, there is a major international effort to develop new drugs. In the chronic phase of the disease, the parasite burden is extremely low, infections are highly focal at a tissue/organ level, and bloodstream parasites are only intermittently detectable. As a result, clinical trials are constrained by difficulties associated with determining parasitological cure. Even highly sensitive PCR methodologies can be unreliable, with a tendency to produce “false-cure” readouts. Improved diagnostic techniques and biomarkers for cure are therefore an important medical need. Methodology/Principal findings Using an experimental mouse model, we have combined a multiplex assay system and highly sensitive bioluminescence imaging to evaluate serological procedures for diagnosis of T. cruzi infections and confirmation of parasitological cure. We identified a set of three antigens that in the context of the multiplex serology system, provide a rapid, reactive and highly accurate read-out of both acute and chronic T. cruzi infection. In addition, we describe specific antibody responses where down-regulation can be correlated with benznidazole-mediated parasite reduction and others where upregulation is associated with persistent infection. One specific antibody (IBAG39) highly correlated with the bioluminescence flux and represents a promising therapy monitoring biomarker in mice. Conclusions/Significance Robust, high-throughput methodologies for monitoring the efficacy of anti-T. cruzi drug treatment are urgently required. Using our experimental systems, we have identified markers of infection or parasite reduction that merit assessing in a clinical setting for the longitudinal monitoring of drug-treated patients.


Introduction
Chagas disease results from infection with the insect-transmitted, obligate intracellular parasite Trypanosoma cruzi, and is a major public health concern in many areas of Latin America [1,2]. In humans, the initial acute stage is characterised by widespread distribution of T. cruzi in tissues and organs, and by a readily detectable parasitemia. The induction of a cellular immune response then results in a major reduction in the parasite load, and by 6-8 weeks post-infection, detection of bloodstream parasites becomes problematic. However, sterile clearance does not occur, and typically the infection transitions to a chronic life-long state [3]. In the acute stage, symptoms are usually mild and transient, although they can occasionally present as myocarditis or meningoencephalitis [4,5]. 30-40% of infected people eventually develop chronic pathology, with cardiomyopathy and/or digestive megacolon syndromes being the most common outcomes [6,7]. Chronic symptomatic disease can take decades to become apparent, and the lack of reliable predictive markers to identify those most at risk of disease progression has highlighted the need for new approaches in this area [8,9].
There is strong evidence that the tissue damage associated with chronic Chagas disease is cumulative and dependent on parasite persistence [10][11][12][13]. This provides a rationale to underpin curative drug treatment as a strategy to block or alleviate the development of chronic pathology. The oral nitro-heterocyclic pro-drugs benznidazole and nifurtimox are the only chemotherapeutic treatments available for T. cruzi infections [14,15]. Although treatments show efficacy in the chronic indeterminate stage of the disease, long treatment periods (~60 days) and frequent toxic side-effects impact negatively on patient compliance [16][17][18]. In addition, few of those in the acute or asymptomatic chronic stages are diagnosed, and the percentage of infected individuals offered anti-parasitic treatment, at the time-point when it would be most beneficial, is very small [19]. In response to the urgent need for more effective anti-T. cruzi therapeutics, there is now a global effort involving the academic, commercial, and not-for-profit sectors [20]. Furthermore, clinical trials aimed at optimising dosing regimens of the currently available drugs are on-going [21].
During the chronic stage, parasites are only intermittently detectable in the bloodstream and infection foci are rare and highly focal at a tissue/organ level [22]. As a result, unequivocal confirmation of parasitological cure is difficult to demonstrate. Even highly sensitive PCR methodologies require long-term follow-up to provide reliable diagnosis [16][17]. Serologybased approaches to assess parasitological cure have also proved to be problematic with seroconversion often requiring many years [23][24]. The identification of robust biomarkers of parasitological cure is therefore crucial to facilitate clinical trials and the subsequent roll-out of new drugs at a community level. The long-term nature and complexity of Chagas disease in humans has meant that predictive experimental models have played a key role in research [25]. One such example is the murine bioluminescence imaging system, based on infections with genetically modified T. cruzi that express a red-shifted luciferase, which facilitates highly sensitive longitudinal monitoring of parasite burden in a non-invasive manner [26][27][28]. This has become an important tool for drug testing and for studies on disease pathogenesis. Here, we have combined this pre-clinical model with an antibody multiplex assay system (Multi-Cruzi, InfYnity Biomarkers, Lyon, France) [29][30][31] as a means of developing improved serological procedures for diagnosis of Chagas disease and for confirming parasitological cure of chronic T. cruzi infections.

Ethics statement
Experiments were performed under UK Home Office project licenses PPL 70/8207 and P9AEE04E4, with consent of the LSHTM Animal Welfare and Ethical Review Board (AWERB). All protocols and procedures were conducted in accordance with the UK Animals (Scientific Procedures) Act 1986 (ASPA).

Mice and parasites
Experimental infections were carried out using female BALB/c mice, purchased from Charles River (UK). CB17-SCID mice were bred in-house. They were maintained in individually ventilated cages, under specific pathogen-free conditions, with a 12-hour light/dark cycle, and given food and water ad libitum. A T. cruzi CL Brener line (Discrete Typing Unit (DTU) 6), that constitutively expresses the red-shifted luciferase PpyRE9h [26], was used for infections. BALB/c mice (6-8 weeks old), were injected i.p. with 1x10 3 bloodstream trypomastigotes obtained from immunodeficient CB17-SCID mice, as outlined previously [27,28].

Assessment of drug treatment and bioluminescence monitoring
Chronically infected mice were dosed by oral gavage with benznidazole (Epichem Pty Ltd.) in an aqueous suspension vehicle containing 0.5% (w/v) hydroxypropyl methylcellulose (HPMC), 0.4% (w/v) Tween 80 in Milli-Q H 2 O. For in vivo imaging, mice were injected with 150 mg kg -1 d-luciferin i.p., and anaesthetized with 2.5% (v/v) gaseous isoflurane in oxygen [27,28]. They were then placed in IVIS Lumina or Spectrum imaging systems (Caliper Life Science). Images were acquired 10-20 minutes after d-luciferin administration using Living Image 4.7.2. Exposure times were between 1 and 5 minutes, depending on the signal. To estimate parasite burden, whole body regions of interest were drawn using Living Image 4.5.5 to quantify bioluminescence expressed as total flux (photons/second; p/s). The detection threshold was established from uninfected mice.
For ex vivo imaging, mice were injected i.p. with 150 mg kg −1 d-luciferin and then sacrificed by exsanguination under terminal anesthesia 5 minutes later [32]. They were perfused with 10 ml d-luciferin at 0.3 mg ml -1 in DPBS (Dulbecco's phosphate-buffered saline) via the heart. Organs and tissues were excised, transferred to a dish, soaked in 0.3 mg d-luciferin ml -1 in DPBS, and then imaged as for the live mice. At various time points, blood samples (minimum 40 μl) were taken from tail veins into individual sterile tubes containing 10 μl 0.5 M EDTA. They were spun at 1,800 g for 10 minutes (4˚C) and the serum fractions collected. Aliquots were stored at -20˚C until shipment on dry ice.

Multiplex Chagas assay
The Multi-Cruzi immuno-assay used here is equivalent to the multiplex system described previously [29][30][31], adapted for mice serum samples. The technology allows multiple antigens to be combined in single wells of a 96-well microplate, using a printing process based on noncontact piezo electric impulsion of a defined volume of an antigenic solution. We used a noncontact volume dispensing system to print 16 T. cruzi antigens (Fig 1) at the bottom of each well at precise X-Y coordinates, under controlled humidity and temperature. Antigens, selected for proven immunogenic properties, were obtained synthetically and printed in duplicate [29,30]. In addition, four Positive Control (PC) spots were printed in order to define the spatial orientation and validate the distribution of all the materials (mouse serum samples, conjugate and substrate) in the correct order. The ELISA assay was carried out as previously described [29] using the anti-mouse-IgG secondary antibody conjugated to horseradish peroxidase.
Each plate was read and analysed using a specific microarray reader which acquires highresolution digital images. An integrated software calculates the pixel intensity for each spot with the background noise subtracted. We calculated the mean value of the duplicated spots to get the net intensity of each antigen. The data were incorporated into a statistical analysis package and analysed accordingly.

Study design
Three cohorts of 15 BALB/c mice were infected with bioluminescent T. cruzi (Methods). Cohort size was selected to provide for the expected variability of the immune response, identified in preliminary experiments. When mice had reached the chronic stage (101 days postinfection), they were treated once daily with benznidazole at either 30 or 100 mg kg -1 for 5 days, dosing regimens that are generally non-curative or curative, respectively. These curative outcomes were based on mice being bioluminescence negative by both ex vivo and in vivo imaging following immunosuppressive treatment [33]. In the current experiment, mice were monitored by bioluminescence imaging, with blood samples taken at regular intervals: at days 101, 138, 182, 202, 259 and 300, relative to the day of infection (day 0).

Statistical analysis
Linear Mixed Model (LMM) analysis was used for the statistical analysis of the longitudinal data, since a mixed effects model has both random and fixed effects. In the current model, treatment group is a fixed effect, while time and mouse are random effects. Longitudinal data are described by the response variable (biomarker intensity), which was repeatedly measured at each group and time. The data prior to start of treatment were disregarded from the LMM analysis. There were no missing value observations. Response variables were the antigen reactivities of 16 biomarkers in the Multi-Cruzi assay and the log-transformed bioluminescence total flux. Independent variables were time (in days after infection) and treatment group (vehicle; benznidazole 30 mg kg -1 ; benznidazole 100 mg kg -1 ). Variance and co-variance were modelled as unstructured (UN). For each subgroup, and for each antigen or bioluminescence variable, slopes and intercepts were calculated and compared. SAS Proc Mixed was used for LMM analysis, from SAS 9.4 (SAS Institute Inc., Cary, NC, USA). The mixed effects model allows the average intercept and slope to be fitted as fixed effects, while taking account of differences between mice and time (random effects).

Identification of reactive antigens
Our aim was to use the Multi-Cruzi serology assay to generate immune signatures of disease status before and after benznidazole treatment of T. cruzi-infected mice (5 days, at either 30 or 100 mg kg -1 ), and to identify biomarkers of chronic disease and curative outcomes. For that purpose, mouse serum samples were screened using the multiplex immuno-assay system containing 16 antigens arrayed in microtiter plate wells as described (Fig 1). The linear mixed model (LMM) analysis results are shown in Table 1. Blue: significant p-values (< 0.05); Red: high intensity. We have data on several mice where we have treatment group and biomarker reactivity measured at different time points. Where all mice have the same slope and intercept relating reactivity to time and treatment group, a regular multiple linear regression model can be fitted with time and treatment group as the predictor, and reactivity as the response. Biomarker reactivities < 10 were considered below the positive reactivity threshold and were regarded as non-reactive. Slopes compared between treatment groups were assigned as different based on a 5% significance level. At the initial screening time-point, 35 days post-infection, there was diversity in the antigen-specific response. Some antigens were detected by sera from all mice, whereas with other antigens, the response was universally weak or insignificant at this time-point (Fig 2A and 2B). Some antigens showed high signal intensities until the end of the monitoring period (300 days) (Fig 2C). These highly reactive antigens were IBAG39 (SAPA-shed acute phase antigen, a member of the trans-sialidase super-family expressed by infectious forms of the parasite) and IBAG257 (the B-cell superantigen Tc24, a flagellar calcium-binding protein), followed by IBAG38 (TcD) ( Table 1 (see intercept), Figs 2C and 3). IBAG38 and IBAG39 antigen reactivities showed a significant decline over time in the 100 mg kg -1 benznidazole treated mice, but not in animals treated with the vehicle or 30 mg kg -1 benznidazole (Fig 3A and 3B). These antigens might be considered as sensitive to treatment. In contrast, IBAG257 reactivity showed a decline of only borderline significance in both the 100 mg kg -1 benznidazole and the vehicletreated subgroups (Fig 3C). Therefore, we consider this antigen as non-sensitive to treatment, and highly reactive to the continued presence of antibodies. IBAG108 and IBAG131 were reactive, but no difference in longitudinal profile between subgroups could be observed.

IBAG
For three other antigens (IBAG36, IBAG37 and IBAG101), significant reactivity only became apparent 135 days post-infection in the vehicle control group, followed by a slow but steady increase until 300 days post-infection (Fig 3D-3F). In contrast, in the mice treated with 100 mg kg -1 benznidazole, there was no detectable immune response against these antigens, while in mice that had been treated at 30 mg kg -1 , a small increase was observed towards the end of the experimental period (Fig 3D-3F). These data indicate that IBAG36, IBAG37 and IBAG101 represent specific markers for late stage chronic infection and immune reactivity to these antigens can be reduced or eliminated by benznidazole treatment in a dose-dependent manner. Approximately 20% of mice recognised antigen IBAG112 by day 100 post-infection, but by day 300, this reactivity was on a downward trajectory. For the other antigens (IBAG35, IBAG95, IBAG97, IBAG99, IBAG112, IBAG20, IBAG134, and IBAG110), no significant reactivity was recorded at the different time points tested throughout infection (Table 1 and Fig 2C).

Correlation between bioluminescence intensities and individual serology biomarkers
As bioluminescence imaging is an accepted reference method for monitoring parasite burden in mice, we next assessed the linear correlation coefficient for each reactive antigen and bioluminescence data after the initiation of benznidazole treatment. Overall, treatment resulted in a reduction in the bioluminescence-inferred parasite burden close to background levels (Figs 4 and S1). However, by 300 days post-infection, the experimental end-point, all mice treated at 30 mg kg -1 had relapsed (n = 15) (S1 and S3 Figs). In contrast, mice treated at 100 mg kg -1 were bioluminescence-negative by in vivo imaging at day 300 (Figs 4, S2 and S3). When these mice were further assessed by ex vivo imaging, small bioluminescent foci were detectable in the carcass of 4 animals, with the other 11 mice being bioluminescence-negative (Figs 4B and S2). This latter group was designated as cured. We did not use immunosuppression to promote relapse in this study, because of the long post-treatment period (almost 200 days), and because of concerns that it would perturb the immunoassay readout. Table 2 shows the linear correlation coefficients for individual biomarkers and bioluminescence data from each mouse. Strikingly, the analysis revealed a very high linear correlation between antigen reactivities of IBAG39 and the log-transformed bioluminescence variable in mice treated with 100 mg kg -1 benznidazole, indicating that IBAG39 is a potential therapy monitoring biomarker (Fig 5).

Discussion
This study addresses one of the key challenges in Chagas disease control, which is the lack of serological tools and reliable biomarkers to monitor parasitological cure. The current strategy, for investigational settings, is to use PCR-based methods for assessing Chagas disease drug efficacy. However, because of the highly focal and low-level nature of chronic infections [22,27], long-term follow-up is required [16,17], and the potential for "false-cure" outcomes remains high. This presents major problems in clinical trials, and in future, will complicate the roll-out of new therapies. The recently developed multiplex serology assay system has shown potential as a means of monitoring T. cruzi persistence and identifying biomarkers that are predictive of successful treatment [29][30][31]. To further explore the utility of this approach, we applied the multiplex methodology to a highly sensitive experimental infection model, based on bioluminescence, where the limit of detection by ex vivo imaging is less than 20 parasites [22]. The aim was to determine if procedures such as this could be incorporated into the drug development pipeline, as an additional approach to identifying drug candidates that eliminate an infection. This experimental system also has the advantage of a standardized infection timeline and a method for designating cure that is more reliable than PCR-based methodologies [34].
During the acute stage, T. cruzi infection induces a strong polyclonal B cell/antibody response which can protect against virulent infection [35]. In the chronic stage however, the role of the humoral immune system is less certain [36]. The difficulties in addressing this reflect the diversity and complexity of the antibody response, even within an in-bred mouse population infected by a single T. cruzi clone (Fig 2). As shown, responses to some parasite antigens are rapid and almost universal (IBAG38, IBAG39, IBAG108 and IBAG257), whereas with others (eg. IBAG36, IBAG37 and IBAG101), induction is slow, but progressive. One of the most reactive antigens identified in the multiplex assay, IBAG257, corresponds to Tc24, a flagellar-localised Ca 2+ binding protein that is highly conserved among different T. cruzi strains and is expressed throughout the development cycle [37,38]. It has been widely tested as a vaccine candidate [39][40][41], and its use in serodiagnosis has been reported [42,43]. Here, we found that the reactivity to this antigen using the multiplex assay was rapid, high and ubiquitous (n = 45) (Fig 2), but did not display discriminatory power to identify treated mice. A Table 2. Linear correlation coefficients of each antigen for each infected mouse treated with 100 mg kg -1 benznidazole. Correlation coefficients were calculated, for log10 (total flux) and biomarkers, for each mouse separately (n = 6 observations post-treatment). Statistically significant correlation coefficients are marked in red. Red: linear correlation coefficients > 0.8 (p-value < 0.05). similar rapid and robust response was observed to IBAG39 (SAPA [44,45]) and IBAG38, with all 45 mice in the current study (Fig 2), and all 25 mice in the pilot experiment, generating reactive antibodies. However, it was possible to discriminate between treated and non-treated mice on the basis of the reactivity towards these antigens. In response to 100 mg kg -1 benznidazole treatment, there was a slow but steady decline in the intensity of IBAG38 and IBAG39 antibody reactivity post-dosing, compared to a continued increase in the non-drug treated mice (Fig 3A and 3B), a profile that was highly correlated (for IBAG39) with the bioluminescence flux (Fig 5). The response was intermediate in the case of mice treated at 30 mg kg -1 . Non-treated mice could also be discriminated by their response to IBAG36, IBAG37 and IBAG101. In the responding mice, the net intensity increased progressively from day 100 postinfection onwards, whereas there was no significant response in those mice that had been treated with 100 mg kg -1 (Fig 3D-3F). IBAG108 proved to be highly reactive, with a rapid and long-lasting response in >90% of the mice tested (Fig 2C), but showed no difference between treatment subgroups. Therefore, in the context of the multiplex system, IBAG38, IBAG39, IBAG108 and IBAG257 represent a highly sensitive and powerful combination for diagnosis of both acute and chronic T. cruzi infection, with IBAG39 having potential as a therapy-monitoring biomarker in mice. The multiplex system approach offers significant value for biomarker identification and validation due to its multi-parametric nature, with information collected simultaneously for several different antigens. Each antigen provides data on the development of reactivity and the probable antibody load in individual animals. The combination of independent parameters, enhances the analysis and provides a rapid, reliable and robust indication of signal decay linked to waning antibody levels. This approach, combined with recently developed prediction and dilution concepts based on mathematical models, is especially relevant for rapid evaluation of the dynamics of the antibody-mediated immune response [31,46]. Treatment success is difficult to measure using the current criterion for parasitological cure because T. cruzi-specific antibodies persist for long periods after successful anti-parasitic treatment, and long-term follow-up is needed to monitor seroconversion.

Linear correlation coefficients
Detailed validation of the Multi-Cruzi immunoassay and selected clinically proven biomarkers is important as a demonstration of translational value-its usefulness for in vivo basic research and diagnostic applications for patients. One limitation of the current study concerns the restriction to the BALB/c mouse infection model and the T. cruzi CL Brener strain. Another could be the reported immune modulation associated with benznidazole treatment [47]. It will therefore be worthwhile to verify whether reduction of parasite load is also associated with a decrease of the specific serological markers when using other T. cruzi lines, mouse strains, and therapeutic agents. Finally, it would be useful to extend our study to other laboratory animals, such as naturally infected non-human primates, which can serve as a model for evaluating the effect of drug candidates on infection with this parasite [48], and for informing strategies applicable to a clinical context.
In summary, the Multi-cruzi assay represents a highly promising research and discovery tool that can be used in combination with currently used diagnostic techniques (PCR and bioluminescence) to perform serological profiling studies and to monitor drug and possibly vaccine efficacy in preclinical models. Furthermore, its early adoption during the discovery process, and as a tool for clinical trial evaluation and regulatory approval, could help to drive acceptance of new treatments for Chagas disease.