Corresponding Author Jean-Charles Sanchez, Biomedical Proteomics Research Group, Structural Biology and Bioinformatics Department, Medical University Centre, 1 rue Michel Servet, Geneva, Switzerland. E-mail: firstname.lastname@example.org
Objectives A critical step before treatment of human African trypanosomiasis (HAT) is the correct staging of the disease. As late stage is established when trypanosomes cross the blood–brain barrier and invade the central nervous system, we hypothesized that matrix metalloproteinases and cell adhesion molecules could indicate, alone or in combination, the disease progression from the first to the second stage of HAT.
Methods We measured the levels of MMP-2, MMP-9, ICAM-1, VCAM-1 and E-selectin in the cerebrospinal fluid (CSF) of 63 Trypanosoma brucei gambiense-infected patients (15 stage 1 and 48 stage 2). Staging was based on counting of white blood cells (WBC) and/or parasite detection in CSF. Concentrations were obtained either by ELISA or multiplex bead suspension assays, and results were compared with three known HAT staging markers (CXCL10, CXCL8 and H-FABP).
Results ICAM-1 and MMP-9 accurately discriminated between stage 1 and stage 2 patients with HAT with 95% sensitivity (SE) for 100% specificity (SP), which was better than CXCL10 (93% SE for 100% SP), one of the most promising known markers. Combination of ICAM-1 and MMP-9 with H-FABP provided a panel that resulted in 100% of SE and SP for staging HAT.
Conclusions ICAM-1 and MMP-9, alone or in combination, appeared as powerful CSF staging markers of HAT. Final validation of all newly discovered staging markers on a large multi-centric cohort including both forms of the disease as well as patients with others infections should be performed.
Objectifs: Une étape cruciale avant le traitement de la trypanosomiase humaine africaine (THA) est la détermination correcte du stade de la maladie. Comme le stade tardif est établie lorsque les trypanosomes traversent la barrière hémato-méningée et envahissent le système nerveux central, nous avons émis l’hypothèse que les métalloprotéinases matricielles et les molécules d’adhésion cellulaires pourraient indiquer, seuls ou en combinaison, la progression de la maladie du stade 1 vers le stade 2 de la THA.
Méthodes: Nous avons mesuré les taux de MMP-2, MMP-9, ICAM-1, VCAM-1 et E-sélectine dans le liquide céphalo-rachidien (LCR) de 63 patients infectés par Trypanosoma brucei gambiense (15 au stade 1 et 48 au stade 2). La classification a été basée sur le comptage des globules blancs (GB) et/ou la détection des parasites dans le LCR. Les concentrations ont été obtenues soit par ELISA ou par tests multiplex bead suspension et les résultats ont été comparés avec trois marqueurs connus de la classification de la THA (CXCL10, CXCL8 et H-FABP).
Résultats: ICAM-1 et MMP-9 ont distingué avec précision les patients THA des stades 1 et 2 avec une sensibilité de 95% et une spécificité de 100%, ce qui était meilleur qu’avec CXCL10 (sensibilité 93%, spécificité 100%), l’un des marqueurs les plus prometteurs connus. La combinaison de ICAM-1 et MMP-9 avec H-FABP a fourni un tableau menant à 100% de sensibilité et de spécificité pour la classification des stades de la THA.
Conclusions: ICAM-1 et MMP-9, seuls ou en combinaison, apparaissent comme puissants marqueurs LCR des stades de la THA. La validation finale de tous les nouveaux marqueurs découverts, dans une large cohorte multicentrique, incluant les deux formes de la maladie ainsi que des patients atteints d’autres infections devrait être effectuée.
Objetivos: Un paso crítico antes del tratamiento de la tripanosomiasis humana africana (THA) es averiguar correctamente el estadio de la enfermedad. Puesto que la etapa tardía corresponde al cruce de los tripanosomas a través de la barrera hematoencefálica y su invasión del sistema nervioso central, hemos planteado la hipótesis de que las metaloproteinasas matriciales (MMP) y las moléculas de adhesión celular podrían indicar, solas o en combinación, la progresión de la enfermedad del primer a segundo estadio de THA.
Métodos: Hemos medido los niveles de MMP-2, MMP-9, ICAM-1, VCAM-1 y la Selectina E en líquido cefalorraquídeo (LCR) de 63 pacientes infectados con Trypanosoma brucei gambiense (15 estadio 1 y 48 estadio 2). El estadio se basó en el conteo que células blancas (CB) y/o la detección de parásitos en LCR. Las concentraciones se determinaron mediante ELISA o técnicas multiplex con partículas en suspensión, y los resultados se compararon con tres marcadores con estadios conocidos de THA (CXCL10, CXCL8 y H-FABP).
Resultados: ICAM-1 y MMP-9 discriminaron de forma precisa entre el estadio 1 y el estadio 2 en pacientes con THA con una sensibilidad (SE) del 95% para una especificidad (SP) del 100%, lo cual fue mejor que CXCL10 (93% SE para un 100% SP), uno de los marcadores conocidos más prometedores. Una combinación de ICAM-1 y MMP-9 con H-FABP proveyó un panel que dio un 100% de SE y SP para los estadios de THA.
Conclusiones: ICAM-1 y MMP-9, por si solos o en combinación, son marcadores poderosos para los estadios de THA en LCR. Debería realizarse la validación final de todos los marcadores de estadio recientemente descubiertos en un gran ensayo multicéntrico de cohortes que incluya ambas formas de la enfermedad así como pacientes con otras infecciones.
Human African trypanosomiasis (HAT), or sleeping sickness, is a tropical disease occurring in sub-Saharan Africa, caused by the protozoan parasite Trypanosoma brucei (T. b.). Two forms of the disease affect humans, a chronic form caused by the Trypanosoma brucei gambiense subspecies, and an acute form because of Trypanosoma brucei rhodesiense (Brun et al. 2010). The disease evolves from a first haemolymphatic stage (stage 1, S1) to a second meningo-encephalitic stage (stage 2, S2), after parasites cross the blood–brain barrier (BBB) and invade the central nervous system (CNS) (Kristensson et al. 2010). If untreated, disease outcome is always fatal (Kristensson et al. 2010). The treatment differs between the two stages and is more toxic and complicated in the case of second (late)-stage disease. After diagnosis, patients therefore undergo a lumbar puncture for staging of sleeping sickness, which relies on cerebrospinal fluid (CSF) examination (Chappuis et al. 2005). Correct staging of patients with HAT remains a critical issue (Kennedy 2008). According to WHO guidelines, the late stage is defined by the detection in the CSF of more than five white blood cells (WBC)/μl and/or the presence of trypanosomes (WHO 1998). However, higher thresholds for the WBC count as well as existence of a possible intermediate stage have been suggested (Rodgers 2009). Trypanosome detection in CSF defines the second stage, but direct finding suffers from sensitivity (SE) (Chappuis et al. 2005). In this context, discovery of new highly sensitive and specific staging markers is essential to replace or complement current staging methods.
Relations between chemokine levels in CSF and HAT progression have been demonstrated (Lejon et al. 2002; Courtioux et al. 2009) and recently, CXC motif chemokine 10 (CXCL10), alone or in combination with interleukin-8 (CXCL8) and heart-type fatty acid–binding protein (H-FABP), has been highlighted as a powerful CSF staging biomarker (Amin et al. 2009; Hainard et al. 2009). Moreover, it has been suggested that trypanosomes may cross the BBB using similar mechanisms as leucocyte infiltration into the CNS (Kristensson et al. 2010). Cell adhesion molecules (CAMs) and matrix metalloproteinases (MMPs) are known to be involved in BBB permeability (Rosenberg 2009) and leucocyte extravasation (Ransohoff et al. 2003). Studies have demonstrated the up-regulation of intercellular adhesion molecule 1 (ICAM-1), vascular cell adhesion molecule 1 (VCAM-1) and E-selectin (endothelial leucocyte adhesion molecule 1) associated with parasite infiltration into the CNS (Mulenga et al. 2001; Girard et al. 2005). The crucial role of matrix metalloproteinase-2 (MMP-2) and matrix metalloproteinase-9 (MMP-9) for WBC infiltration into the brain parenchyma has also been shown (Agrawal et al. 2006). In this context, we hypothesized that MMPs and CAMs could be used, alone or in combination, as HAT staging biomarkers. We measured the level of MMP-2, MMP-9, ICAM-1, VCAM-1 and E-selectin in the CSF of T. b. gambiense-infected patients and evaluated their performances in the context of CNS invasion by the parasite. Levels of three other biomarkers (CXCL10, CXCL8 and H-FABP) previously shown to be efficient for staging were also measured and used for comparison.
Materials and methods
Samples originated from a prospective study on 360 T. b. gambiense-infected patients (THARSAT), conducted between 2005 and 2008 at Dipumba hospital in Mbuji-Mayi (Kasai Oriental province, Democratic Republic of the Congo). Details of the THARSAT study design and results are reported elsewhere (Mumba Ngoyi et al. 2010). The study was approved by the Ministry of Health, Kinshasa, DRC, and by the Ethical Committee of the University of Antwerp, Belgium. Written informed consent was obtained from the patients or their responsible before enrolment. In this study, a total of 63 CSF samples comprising 15 stage 1 (S1) and 48 stage 2 (S2) patients, taken before treatment, were tested (Table 1). The patients were chosen sequentially and were also classified according to their neurological condition (Table 1). The S1 and S2 groups did not differ significantly in terms of age and sex (chi-square or Mann–Whitney U test). Three categories of neurological signs were defined: absence (no neurological signs), moderate (at least one major neurological sign but no generalized tremors) and severe (at least two major neurological signs including generalized tremors). Major neurological signs were defined as: daytime somnolence, sensory and gait disturbances, presence of primitive reflexes (Babinski’s sign, palmo-mental reflex, perioral reflex), modified tendon reflexes (exaggeration or abolition), abnormal movements such as tremor (fine, diffuse and generalized). Disease staging was determined after CSF examination using the following criteria: patients with WBC ≤5/μl and no trypanosomes in CSF were classified as S1; patients with WBC >5/μl and/or trypanosomes in CSF corresponded to S2.
Table 1. Characteristics of the selected population of patients with human African trypanosomiasis
*Chi-square test: ns.
‡Mann–Whitney: P < 0.001.
§Chi-square test: P = 0.001.
Age (years) median (range)†
WBC/μl median (range)‡
Detected trypanosomes in cerebrospinal fluid (n)‡
Matrix metalloproteinases (MMPs), cell adhesion molecules (CAMs) and chemokines measurements
The concentrations of MMP-2, MMP-9, ICAM-1, VCAM-1, E-selectin, CXCL10 and CXCL8 were measured using commercially available multiplex bead suspension assay (mBSA) kits (R&D Systems, UK; Bio-Rad, Hercules, CA, USA) according to manufacturer’s instruction. Samples were diluted 1:4 with the corresponding provided buffer and assayed in duplicate. The concentration of each target was automatically calculated by the Bio-Plex Manager v4.1 software using corresponding standard curves (5-PL regression) obtained with recombinant protein standards.
The concentration of H-FABP was measured using a commercially available ELISA (Hycult Biotechnology, Uden, the Netherlands) following the protocol described elsewhere (Hainard et al. 2009).
Combinations of all evaluated proteins were performed to determine whether a panel could be more efficient for staging than the molecules alone. Optimized cut-off values were obtained by modified iterative permutation–response calculations (rule-induction-like, RIL) as described in a previous study (Hainard et al. 2009).
Data and statistical analysis
Statistical analysis were performed using pasw v18 and GraphPad Prism v4.03 software. Correlations and differences between groups were tested with non-parametric tests (i.e. Mann–Whitney U test, Kruskal–Wallis test, Dunn’s post hoc test and Spearman test). Marker’s SE was evaluated at 100% of specificity (SP). Overall accuracy was further assessed using the Youden index (Y = maximum [SE + SP − 1]) or partial area under ROC curve (pAUC) between 90% and 100% of SP (McClish 1989).
HAT CSF staging
As presented in Table 2 Part A, all the CAMs and the MMPs were able to discriminate the two stages of the disease with a high significance (P < 0.0001, Mann–Whitney U test) except E-selectin (P value = 0.0152). E-selectin and MMP-2 showed pAUC below 80% (respectively at 73% and 78%), whereas VCAM-1, ICAM-1 and MMP-9 presented higher pAUC ranging from 93% to 95% (Figure 1). The five molecules were able to identify S2 patients with SE ranging from 48% (E-selectin) up to 90% for ICAM-1 and MMP-9 (Figure 1) for 100% SP. As the number of WBC is the current staging reference, we evaluated the correlation between this parameter and the analyte levels. All these molecules significantly correlated with the number of WBC, with ICAM-1 showing the highest correlation (Spearman rho coefficient of 0.794). The three already known markers, CXCL10, CXCL8 and H-FABP, were also able to significantly discriminate between the two stages (Table 2 part B) with pAUC, respectively, at 84%, 89% and 93%. These three mole-cules were also significantly correlated with the number of WBC.
Table 2. Detailed results for the cell adhesion molecules and matrix metalloproteinases (Part A) as well as three other known human African trypanosomiasis markers (Part B) tested in respect with the stage of the disease
Stage 1 median (range)
Stage 2 median (range)
Correlation with WBC†
Partial area under ROC curve‡ (%)
% SE (95% CI)§
SE, sensitivity; SP, specificity.
Median values (and ranges) are expressed in pg/ml.
*P value according to the Mann–Whitney U test.
†Spearman rho value [**significant correlation at the 0.01 level (two-tailed)].
‡Partial AUC between 90% and 100% of SP.
§SE was set for a SP of 100%.
3085.3 (175.7–22 643.0)
8493.1 (604.6–39 166.5)
9100.5 (1466.4–23 567.9)
50 625.6 (8894.3–195 040.3)
15 354.8 (4010.0–38 832.9)
39 798.7 (10 565.9–108 146.1)
11 699.1 (24.3–65 596.9)
759.8 (234.1–16 680.7)
Presence of trypanosomes in the CSF
To evaluate the relationship between the presence of parasites in the CSF and the concentration of the molecules, patients were classified into two groups: without (T−) and with (T+) detected trypanosomes in CSF (Figure 2 and Table S1). Concentrations of all MMPs and CAMs were significantly increased in T+ patients (P < 0.01, Mann–Whitney U test). MMP-2 presented the lowest pAUC at 65%, whereas the other molecules (ICAM-1, E-selectin, VCAM-1 and MMP-9) had pAUC either at 72% or 73%. MMP-2, E-selectin, MMP-9 and VCAM-1 produced Youden indices ranging from 0.16 to 0.64. ICAM-1 showed the highest one with a value at 0.84. WBC count was also considered as an independent variable to compare its performance relatively to the evaluated analytes. The number of WBC was significantly increased in association with the presence of parasites in CSF (P < 0.0001, Mann–Whitney U test) yielding a pAUC of 59% and a Youden index of 0.53. CXCL10, CXCL8 and H-FABP presented pAUC at, respectively, 79%, 65% and 55% and Youden indices ranging from 0.16 to 0.74 (Table S1).
As presented in Figure 3, patients were grouped according to their reported neurological signs (absent, moderate or severe). All the CAMs and MMPs, except E-selectin, presented a significant increase in concentration associated with the severity of the neurological signs (Table S1). However, MMP-9, ICAM-1 and VCAM-1 showed lower P values (<0.0005, Kruskal–Wallis test) than MMP-2 and E-selectin. VCAM-1, ICAM-1 and MMP-9 were able to significantly discriminate between absence and moderate as well as between absence and severe neurological signs (P value, respectively, <0.05 and <0.01, Dunn’s post hoc test). E-selectin presented non-significant P values and MMP-2 only discriminated between absence and moderate neurological signs (P < 0.05, Dunn’s post hoc test). CXCL10, CXCL8 and H-FABP significantly distinguished between the different neurological conditions (P < 0.005, Kruskal–Wallis test). These three molecules discriminated between absence and moderate as well as between absence and severe neurological signs, but CXCL8 was the only one able to discriminate between moderate and severe neurological signs (P < 0.05; Dunn’s post hoc test). WBC count presented significant differences between the three groups (P < 0.005, Kruskal–Wallis test), discriminating between absence and moderate and between absence and severe neurological signs with P values, respectively, <0.05 and <0.01 (Dunn’s post hoc tests).
All the evaluated markers were combined using the RIL method explained earlier. This approach resulted in the identification of a three molecules panel able to detect S2 patients with 100% of both SE and SP. This panel was characterized by MMP-9, ICAM-1 and H-FABP with cut-off values, respectively, at 424.7, 600.9 and 391.4 pg/ml. A positive test response (i.e. identification of a S2 patient) was obtained as soon as two molecules were above their cut-off values (Table 3). This panel more accurately discriminated between S1 and S2 than the one made of CXCL10, CXCL8 and H-FABP (Table 3), previously described (Hainard et al. 2009).
Table 3. Detailed results for the three molecules panels
No. of negative test
No. of positive test
Mann–Whitney U test, P-value
% pAUC* (ROC curve)
Cut-off [no. of positive mol.]
% SE (95% CI)†
SE, sensitivity; SP, specificity; pAUC, partial area under ROC curve.
*Partial AUC between 90% and 100% of SP.
†SE was set for a SP of 100%.
MMP-9 ICAM-1 H-FABP
CXCL10 CXCL8 H-FABP
This study has demonstrated that ICAM-1 and MMP-9 have high staging performance, with a SE of 90% for 100% SP. Furthermore, analysis of a combination of molecules allowed the selection of a panel comprising of MMP-9, ICAM-1 and H-FABP that detected S2 patients with 100% SE and SP.
The brain is protected from the free entry of molecules and pathogens by the BBB, mainly composed of endothelial cells. Second stage of HAT occurs when parasites cross the BBB, inducing a neuroinflammatory process associated with leucocyte infiltration into the CNS (Enanga et al. 2002; Grab & Kennedy 2008). CAMs play an important role in leucocyte migration through the BBB during inflammatory processes (Radi et al. 2001; Man et al. 2007) and are modulated in many CNS infections (Brown et al. 2000; Jaber et al. 2009). In this study, the measured concentration of ICAM-1, VCAM-1 and E-selectin in CSF showed a significant increase according to disease progression. E-selectin is synthesized and expressed by endothelial cells during acute or chronic inflammatory processes, and it is a mediator of leucocyte-endothelial interactions (Man et al. 2007). The increased level of E-selectin in the CSF of S2 patients is a further evidence of its probable association with the inflammatory process; additionally, the significant correlation between its concentration and WBC was coherent with a potential involvement of E-selectin in leucocyte trafficking (Ransohoff et al. 2003). ICAM-1 and VCAM-1 belong to the immunoglobulin superfamily of adhesion molecules and are principally expressed by endothelial cells (Petruzzelli et al. 1999). They are also implicated in leucocyte trafficking (Luster et al. 2005) and in vitro studies demonstrated that the presence of T. b. gambiense parasites activates endothelial cells (the major cell population of the BBB) and enhances the expression of these two proteins (Girard et al. 2005). The increased concentration of ICAM-1 and VCAM-1 observed in the CSF of late-stage patients, and the strong correlation with the number of WBC is a clear indication of their involvement in disease progression. Higher concentrations of ICAM-1 and VCAM-1 were also observed for patients with detected trypanosomes in CSF (T+), suggesting a possible up-regulation associated with the passage of trypanosomes across the BBB (Mulenga et al. 2001).
Matrix metalloproteinases are a family of proteases produced by a variety of cells, such as neurons, astrocytes and microglia (Van den Steen et al. 2006). A subgroup of MMPs, the gelatinases, made of MMP-2 (gelatinase A) and MMP-9 (gelatinase B), play an important role in neuroinflammation. They are involved in BBB permeability by attacking the extracellular matrix (Rosenberg 2009), as demonstrated in other CNS disorders, such as cerebral malaria (Van den Steen et al. 2006), meningitis (Tsai et al. 2008) or multiple sclerosis (Liuzzi et al. 2002). In our study, MMP-2 and MMP-9 were significantly elevated in the CSF of patients with second-stage HAT and were highly correlated with the number of WBC in CSF, confirming their role in BBB dysfunction associated with leucocyte penetration into the CNS (Rosenberg 2009). Indeed, it has been demonstrated that MMP-2 and MMP-9 create a localized temporary opening of the glia limitans (part of the BBB) by selective cleavage of the β-dystroglycan subunit anchoring the astrocyte end-feets to the parenchymal membrane (Agrawal et al. 2006). A clear association between these two MMPs and the presence of parasites in CSF has also been observed, supporting a probable association with trypanosome infiltration into the CNS. The observed modulation of MMP-9 according to the severity of the neurological signs suggested a potential role in neuronal injury as described by Leib et al. (2000) in a rat model of meningitis. MMP-9 performed better than MMP-2 as a marker for staging HAT. Because MMP-9 is secreted during inflammatory processes, whereas MMP-2 is constitutively expressed in CSF (Liuzzi et al. 2002), greater differences in concentrations could be observed for MMP-9, resulting in better staging performances. The roles of MMPs and CAMs in the pathogenesis of HAT remain to be clarified.
CXCL10 was recently reported as a powerful CSF staging biomarker (Amin et al. 2009; Hainard et al. 2009), which was confirmed by the present study, although the accuracy of ICAM-1 and MMP-9 for staging HAT was better. These two molecules showed also better accuracies than the WBC count regarding to the presence of parasites in the CSF and the progression of the neurological signs.
In an effort to further improve SE and SP for the detection of S2 patients, we applied a marker combination strategy. The two best performing molecules, ICAM-1 and MMP-9, associated with H-FABP, resulted in a panel able to detect all the S2 patients and to rule out all the S1 (i.e. 100% of both SE and SP). Interestingly, the brain damage marker (H-FABP), which was part of the panel reported in our previous study, was not a good marker on its own, but clearly improved the accuracy of staging when used in combination with ICAM-1 and MMP-9.
This work therefore highlighted MMP-9 and ICAM-1 as powerful staging markers for T. b. gambiense HAT and demonstrated the utility, alone or in combination, of MMPs as well as CAMs to indicate the second stage of HAT. Nevertheless, deeper investigations of more patients are needed to validate the performance of these biomarkers. The influence of other concomitant diseases, such as malaria and HIV, should also be evaluated, as well as staging ability for patients infected with the T. b. rhodesiense parasite. All these molecules were also evaluated on plasma samples from a small cohort of patients with HAT (n = 30; 15 S1 and 15 S2), without showing any significant differences between the two stages of the disease. Thus, a major drawback remains the continuous need of lumbar puncture for staging.
The authors thank Catherine Fouda, Noémie Roze-Fumaux and Nadia Walter for technical assistance. This work was supported with funds from the Foundation for Innovative New Diagnostics (FIND). The THARSAT study and D. Mumba Ngoyi received financial support from the Belgian Directorate General for International Cooperation.