Human African trypanosomiasis (HAT), commonly known as sleeping sickness, is a neglected tropical disease caused by the Trypanosoma brucei parasite and transmitted to humans through the bite of the tsetse fly 1 . Two morphologically identical subspecies of parasites are responsible for the disease: Trypanosoma brucei gambiense and T. b. rhodesiense 2 . In both cases, the disease progresses from a haemolymphatic first stage (S1), to a meningo-encephalitic second stage (S2). The latter reflects invasion of the central nervous system (CNS) by the parasites across the blood–brain barrier (BBB) with severe neurological complications, which can ultimately lead to coma and death, when untreated 3 . The two forms of HAT differ in their clinical presentations and geographic distribution. The gambiense form is widespread in Central and Western Africa and is commonly considered to be a chronic infection, which slowly progresses from the first to the second stage. The rhodesiense form of sleeping sickness, that affects communities in Eastern Africa, is a more aggressive illness, which rapidly progresses to the meningo-encephalitic stage 3 and accounts for less than 5% of all HAT cases 4 . Contrary to T. b. gambiense, for which a relatively safe drug combination has recently been introduced for treatment of S2 patients 4 5 6 , treatment of S2 T. b. rhodesiense patients still relies on melarsoprol 7 8 9 . Melarsoprol has been reported to cause reactive encephalopathies in 8% of T. b. rhodesiense treated patients, which are fatal in 57% of them 8 . As a drug to safely treat both stage 1 and stage 2 patients is yet to be identified, and as S2 treatment is associated with severe side effects and toxicity 8 , stage determination remains a key step in the management of patients suffering from T. b. rhodesiense HAT.

Staging is based on the examination of the cerebrospinal fluid (CSF) by microscopy. According to WHO, patients having ≤ 5 white blood cells (WBC) per microliter of CSF and absence of parasites are considered to be in the first stage of the disease, while patients having more than 5 WBC/μL and/or presence of parasites in the CSF are considered as S2 10 . These methods suffer from limited specificity and reproducibility of the counting of WBC and lack of sensitivity in finding of parasites in CSF 11 12 (Dieudonné Mumba Ngoyi, personal communication).

The discovery of surrogate markers to complement or replace the counting of WBC in the staging of HAT is highly desired 11 13 14 . Many studies have focused on the staging in T. b. gambiense HAT 13 15 16 17 18 19 20 , while less attention has been paid to T. b. rhodesiense, with a paucity of data on staging markers 13 21 22 . Some pro- and anti-inflammatory factors have been shown to be associated with the late stage of T. b. rhodesiense sleeping sickness, including IL-10, IL-6, CXCL10 and neopterin 13 21 22 .

The aim of the present study was to investigate eight immune-related factors, shown to be powerful markers for stratification of T. b. gambiense HAT patients 13 15 16 17 18 19 20 23 , as staging markers for T. b. rhodesiense sleeping sickness.

Stage determination in T. b. rhodesiense sleeping sickness patients is a critical step in ensuring that the appropriate treatment is used 14 . An imperfect gold standard for staging and the lack of a safe S2 drug highlight the need for new tools for staging this form of disease 27 28 . In the present study we investigated, on a small population of patients suffering from T. b. rhodesiense HAT, a number of molecules (MMP-9, CXCL10, CXCL13, IgM, neopterin, ICAM-1, VCAM-1 and B2MG) known to be over-expressed in the CSF of late stage T. b. gambiense patients 23 . Since melarsoprol is still the only treatment for S2 rhodesiense patients, we evaluated their staging ability as highly specific markers, to try to limit unnecessary exposure of patients to this toxic drug. IgM, MMP-9 and CXCL13 were shown to be the most accurate discriminators between early and late stage disease (pAUC ≥ 85%) and showed the same accuracy as WBC in distinguishing between patients having parasites in their CSF from those without. Furthermore, combination of the molecules into panels of three markers (IgM-CXCL13-CXCL10 or MMP-9-CXCL13-CXCL10) significantly increased the staging accuracy, leading to the correct classification of all S1 patients and 62 out of 71 S2 patients.

All the markers investigated here are known to be involved in the immune response elicited by the presence of the parasite in the host. Interestingly, a different behavior of the 8 molecules was observed in T. b. gambiense patients, which may reflect the differences in immunopathogenesis 29 and clinical presentation 3 30 of the two forms of HAT. It has already been proposed, for example, that different activation pathways of macrophages and astrocytes may take place in the two forms of HAT 22 . Such differences may be responsible of the less accurate staging ability of neopterin on T. b. rhodesiense patients, compared to its very high staging power on T. b. gambiense patients.

The role of IgM, the best individual marker in the present study, in disease progression has been extensively studied. An increased CSF concentration of IgM of intrathecal origin was shown to be a good indicator of brain involvement in HAT 15 , leading to the development of a rapid latex agglutination test (Latex/IgM) for stage determination in the field 31 . However, when assessed under field conditions, this assay did not represent an advantage compared to counting of WBC 31 . Furthermore, when used for evaluation of the outcome after treatment, IgM levels were not an optimal indicator of recovery due to their slow normalization 32 . Studies in animal models have shown that HAT meningo-encephalitis is characterized by an increased number of leukocytes in the CNS 33 . CXCL13, also known as BCA-1, is a chemokine mainly produced by dendritic cells 34 , which specifically attracts B and T lymphocytes to the site of inflammation 35 . Its over-expression in CSF has been associated with increased WBC and intrathecal production of immunoglobulins in many pathological conditions 36 37 , including late stage T. b. gambiense HAT 19 . On the other hand, MMP-9 (matrix-metalloproteinase 9), an enzyme involved in tissue homeostasis and remodeling 38 39 , has been extensively studied in a number of pathologies affecting the CNS 39 40 41 42 , in addition to T. b. gambiense HAT 17 . Due to its ability to degrade β-dystroglycan, this protein has been proposed to be involved in the passage of leukocytes through the glia limitans to reach the brain parenchyma 43 . However, the temporal relationship between the events leading to CNS invasion and the appearance of various signs and symptoms of nervous system dysfunction needs to be investigated further.

The markers investigated in the present study were combined into panels in order to increase their accuracy in stage determination. The utility of this approach to achieve a better diagnostic accuracy has already been shown 18 44 . Using this method, we highlighted highly specific combinations comprising CXCL13, CXCL10 and MMP-9 or IgM. Interestingly, CXCL10 was present in both panels. This molecule was not efficient in staging T. b. rhodesiense patients when considered individually. However, when combined to CXCL13 and MMP-9 or IgM, it helped in reaching a significantly increased staging accuracy. This chemokine, which specifically attracts T lymphocytes to the site of inflammation 45 , was reported to be produced by activated astrocytes in trypanosome-infected mice 46 . The activation of astrocytes and macrophages are early events in stage 2 infection 47 48 49 , suggesting that CXCL10 may represent an early indicator of CNS involvement in HAT.

Interestingly, the markers did not show the same staging performances when assessed on patients classified according to their geographic origin (i.e. Malawi or Uganda). Although the low number of Malawian S1 patients (n=3) certainly represents a bias and may be responsible for the differences observed, this result could reflect the variable clinical presentation of rhodesiense disease observed in different foci 27 . This may suggest that potentially different markers will be needed to stage T. b. rhodesiense patients according to their geographical origin and the parasite strain.

The present study has a number of limitations that should be considered. First, the data presented resulted from analyses on a small number of patients. This is a common problem associated to the investigation of this form of HAT. Collecting samples from T. b. rhodesiense patients is considerably difficult, not only due to the lower incidence of this disease compared to the gambiense form, but also as a consequence of a less effective active screening, since the CATT test can only detect T. b. gambiense cases 50 . To further evaluate the staging properties of the markers, larger cohorts of patients should be investigated. Moreover, due to the reported differences between T. b. rhodesiense HAT among foci, the results presented here should be validated in a more controlled set of patients (i.e. in which the same parasitological examinations were performed).

Another drawback could be represented by the choice of selecting highly specific markers, with the consequence of compromising the sensitivity. Management of T. b. rhodesiense patients is far from being optimal, thus both choices of high specificity or sensitivity would be associated either to a risk of missing the diagnosis of late stage patients, or to the exposure of S1 patients to a highly toxic stage 2 drug, respectively. However, it should be emphasized that a new staging biomarker for rhodesiense HAT would be combined with the detection of parasites in CSF, which would increase the sensitivity, and with clinical evaluation of the neurological status of the patients.

The absence of information on neurological signs exhibited by patients in the present study prevented an efficient assessment of the association between the levels of the markers and the signs of CNS involvement. This aspect is particularly important in the light of a recent publication on T. b. rhodesiense HAT reporting the poor association between disease progression, the levels of a number of cytokines and patients’ neurological manifestations 51 .

The 8 markers investigated here behaved differently when assessed on T. b. gambiense or T. b. rhodesiense samples, underlining the differences between the two forms of disease, and suggesting that potentially new rhodesiense-specific markers could be discovered.

Despite the high staging accuracy shown by the combinations of markers described in the present study (i.e. CXCL10-CXCL13-MMP-9 and CXCL10-CXCL13-IgM), their translation into a rapid field diagnostic test could be difficult, due to a potential increase in the costs of production, suggesting that deeper investigations should be performed. The individual staging power of the molecules should be assessed on a larger cohort of T. b. rhodesiense HAT patients, including CSF samples collected during the post-therapeutic follow-up, and the possibility of their translation into a point-of-care test for stage determination in the field should be evaluated. Furthermore, their study in animal models, as already done for IL-10 52 , could help in the further characterization of the role of these markers in disease progression.

The results presented in this work on T. b. rhodesiense sleeping sickness highlight the potential utility of IgM, MMP-9 and CXCL13, alone or combined with CXCL10 in staging patients. We believe that this work has paved the way for further investigations on the role of these markers in detecting the meningo-encephalitic stage of T. b. rhodesiense HAT, and therefore making a more accurate choice of treatment.

HAT: Human African trypanosomiasis; WBC: White blood cells; CSF: Cerebrospinal fluid; ROC: Receiver operating characteristic; AUC: Area under the ROC curve; pAUC: Partial AUC; S1: Stage 1; S2: Stage2; CNS: Central nervous system; WHO: World health organisation; NECT: Nifurtimox-eflornithine combination therapy; SP: Specificity; SE: Sensitivity; CI: Confidence interval.

Joseph Ndung’u is an employee of the Foundation for Innovative New Diagnostics (FIND). Veerle Lejon, Philippe Büscher and Sanjeev Krishna were consultants for FIND at the time of the study. All other authors declare that they have no competing interests.

NTi, AH, JCS, VL, JMN conceived and designed the experiments. NTi and AH performed the experiments. NTi, AH, JCS, NTu and XR analyzed the data. EM, JCE, DMN, PB, VL and KK collected samples. VL, SK, JMN, PB, KK helped in data interpretation. All authors have either participated in writing or reviewing of the manuscript.