Changes in Inflammatory Markers in Patients Treated for Buruli Ulcer and Their Ability to Predict Paradoxical Reactions

Abstract

Mycobacterium ulcerans causes Buruli ulcer, the third most frequent mycobacterial disease after tuberculosis and leprosy. Transient clinical deteriorations, known as paradoxical reactions (PRs), occur in some patients during or after antibiotic treatment. We investigated the clinical and biological features of PRs in a prospective cohort of 41 patients with Buruli ulcer from Benin. Neutrophil counts decreased from baseline to day 90, and interleukin 6 (IL-6), granulocyte colony-stimulating factor, and vascular endothelial growth factor were the cytokines displaying a significant monthly decrease relative to baseline. PRs occurred in 10 (24%) patients. The baseline biological and clinical characteristics of the patients presenting with PRs did not differ significantly from those of the other patients. However, the patients with PRs had significantly higher IL-6 and tumor necrosis factor alpha (TNF-α) concentrations on days 30, 60, and 90 after the start of antibiotic treatment. The absence of a decrease in IL-6 and TNF-α levels during treatment should alert clinicians to the possibility of PR onset.

Mycobacterium ulcerans causes Buruli ulcer (BU), a disabling neglected tropical disease mostly affecting the skin and soft tissues. BU cases are reported in 33 countries, mostly in West and Central Africa. BU treatment is generally based on an 8-week course of antibiotic treatment with a combination of rifampicin and streptomycin, or clarithromycin [1, 2]. This antibiotic regimen is highly effective, although some patients also require surgery to prevent or limit the severity of functional and aesthetic sequelae [3, 4]. As also observed in infections with Mycobacterium tuberculosis and Mycobacterium leprae, some patients with BU experience a worsening of the disease, known as a paradoxical reaction (PR), during or after antibiotic treatment [3, 5–7]. It is estimated that about 20% of patients treated for BU develop PRs, mostly in the 6–10 weeks after the start of antibiotic treatment [8]. The risk factors and mechanisms underlying this phenomenon remain poorly understood. It has been suggested that it may occur more frequently in older patients, in patients with plaques or an edematous form of the disease [5], and in patients treated with an aminoglycoside [3]. The management of BU-related PRs is largely symptomatic. Longer antibiotic regimens have been tested but did not prove useful. Systemic corticosteroid treatment may be beneficial in severe cases [9] but with a significant risk of adverse events, especially in cases of prolonged use and in settings in which resources are limited [10]. Identifying strategies for preventing BU-related PRs and improving their management is, therefore, an important research goal.

We conducted a prospective cohort study on consecutive patients treated for BU at a single center in Pobè, Benin. We performed a longitudinal analysis of the clinical and biological data for these patients, with the aim of identifying features predictive of, or associated with, the occurrence of a PR.

PATIENTS, MATERIALS, AND METHODS

Study Design and Patient Selection

This prospective study was conducted at the Centre de diagnostic et de traitement de la lèpre et de l’ulcère de Buruli (CDTLUB) in Pobè, Benin. Patients with polymerase chain reaction (PCR)–confirmed BU diagnosed between February 2012 and November 2013 were included in the study once they had given informed consent. Microbial diagnosis was performed by quantitative PCR (qPCR) targeting M ulcerans DNA on serosity collected by needle aspiration, swabs of the ulcer borders, or biopsy specimens of the primary lesion [11]. Patients with human immunodeficiency virus, hepatitis B virus, or hepatitis C virus infections were excluded from the study. We also excluded patients <5 years of age, pregnant women, and patients who did not attend follow-up visits. Two groups of subjects without BU were used as controls: “non-BU patients” were subjects consulting for another condition at CDTLUB at the same time as Buruli patients; and “healthy controls” were healthy blood donors sampled in France.

Patient Management

All patients managed for BU at the CDTLUB were initially hospitalized. Antibiotic treatment with a combination of rifampicin and either streptomycin or clarithromycin was administered daily for 56 days by nurses from the center, in accordance with the 2012 World Health Organization (WHO) recommendations. For patients with extensive lesions, surgical debridement and skin grafting were performed if necessary, in accordance with the decision of a college of plastic surgeons, at least 30 days after the beginning of antibiotic therapy. Patients with large lesions, living a long way from the CDTLUB, or with lesions that required daily dressing changes, remained hospitalized. The other patients were rapidly discharged home but were asked to attend the center daily for clinical assessment and treatment delivery.

Clinical and Routine Biological Surveillance

Clinical assessments of lesion size and the occurrence of treatment-related adverse events were performed daily for all patients. Routine biological surveillance was performed monthly and comprised liver function tests, blood cell counts, and creatinine determinations. Automated blood cell counts were performed on site. After the end of antibiotic treatment, all patients underwent clinical assessments by a physician from the CDTLUB at least once per month, for a year.

Definition of Paradoxical Reaction

A PR was defined as the occurrence of a new ulcerative lesion and/or the enlargement of a preexisting lesion despite initial improvement (clinically assessed in patients with optimal observance of the antibiotic regimen), as previously described [3].

Cytokine Quantification

Blood samples were collected into ethylenediaminetetraacetic acid tubes on the first day of antibiotic treatment (baseline), and then 30, 60, and 90 days later. Serum samples were frozen at −20°C and sent on dry ice to Angers, France, for further analysis. Changes in the immune response during and after treatment were assessed by quantifying 27 soluble immune mediators (cytokines and chemokines) with the Bio-Plex Pro Human Cytokine 27-plex Assay (Bio-Rad, Hercules, California). In brief, the samples were diluted before incubation with specific antibody-coated fluorescent beads according to the manufacturer's recommendations. After washing, 50 beads were analyzed with the Luminex 200 analyzer and Bio-Plex Manager software version 6 (Bio-Rad), and the analytes concentrations of the samples were estimated through the serial dilution of cytokine standards. The following cytokines and chemokines were analyzed: fibroblast growth factor (FGF), granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), interferon gamma (IFN-γ), interferon-γ–induced protein 10 (IP-10), eotaxin, monocyte chemoattractant protein 1, macrophage inflammatory protein 1 alpha and beta subunits (MIP-1α and MIP-1β, respectively), platelet-derived growth factor beta subunit (PDGF-β), tumor necrosis factor alpha subunit (TNF-α), vascular endothelial growth factor (VEGF), interleukin 1 beta subunit (IL-1β), the alpha subunit of the interleukin 1 receptor (IL-1rα), and interleukins 2 (IL-2), 4 (IL-4), 5 (IL-5), 6 (IL-6), 7 (IL-7), 8 (IL-8), 9 (IL-9), 10 (IL-10), 12, (IL-12), 13 (IL-13), and 17 (IL-17). Blood samples of control subjects were analyzed the same way.

Ethical Considerations

The study protocol was approved by the national ethics committee (approval number 002/14 February 2013) and the Ministère de la Santé of Benin (approval number 2893/MS/DC/SGM/DRF/SA). All patients or their legal representative gave written informed consent.

Statistical Analysis

Quantitative variables are expressed as the median (interquartile range [IQR]). Categorical variables are expressed as numbers (percentages). Nonparametric Mann–Whitney and paired, nonparametric Wilcoxon tests were performed; P < .05, P < .01, P < .001, and P < .0001 were considered statistically significant. Data were analyzed and figures were drawn with EpiInfo V7.2.8 (Centers for Diseases Control and Prevention) and GraphPad Prism 9 software.

RESULTS

Description of the Cohort

Baseline Characteristics

Forty-four patients were included in the study, 3 of whom were lost to follow-up after the first visit and were excluded from all analyses (Figure 1). The median age of the patients was 20 (IQR, 7.8–25) years; 27 (65%) were <15 years of age, and 20 (49%) were male (Table 1). Ulcerative BU was the most frequent form (n = 32 [78%]), and most patients had advanced disease (WHO classification category III in 25 patients [61%]). The most frequent localization was the lower limbs. Two patients had osteomyelitis attributed to M ulcerans.

Figure 1.

Study flowchart. Abbreviation: PR, paradoxical reaction.

Table 1.

Clinical Characteristics of the Cohort, Diagnosis, and Treatment

Characteristic	All Patients	No-PR Group	PR Group	P Value
	(N = 41)	(n = 31)	(n = 10)
Clinical characteristics
Demographics
Age, y, median (IQR)	20.3 (7.8–25.0)	20.7 (8.0–32.5)	19.1 (7.8–22.0)	.81
Children (<15 y old)	27 (65)	20 (65)	7 (70)	1.00
Sex (male)	20 (49)	15 (48)	5 (50)	1.00
Type of lesion at baseline
Edema without ulceration	9 (22)	5 (16)	4 (40)	.18
Ulceration	32 (78)	26 (84)	6 (60)
Osteomyelitis	2 (5)	0 (0)	2 (20)
Site of lesion
Upper limb	15 (37)	11 (35)	4 (40)	1.00
Lower limb	28 (68)	20 (65)	8 (80)	.46
Both	2 (5)	0 (0)	2 (20)
WHO classification
II	16 (39)	14 (45)	2 (20)	.26
III	25 (61)	17 (55)	8 (80)
Diagnosis and treatment
Specimen
Biopsy	1 (2)	1 (3)	0 (0)
Swab	28 (68)	23 (74)	5 (50)	.24
Serosity	12 (30)	7 (23)	5 (50)	.12
Direct microscopy (ZN staining)
Positive	23 (56)	20 (65)	3 (30)	.07
AFB, count/100 fields, median (IQR)	157 (33–425)	106 (32–500)	233 (167–342)	.62
Negative	10 (24)	6 (19)	4 (40)	.22
Not done	8 (20)	5 (16)	3 (30)	.38
Bacterial load on PCR, Nb bacteria/ml, median (IQR)	18 000 (2300–100 000)	18 000 (2150–155 000)	36 500 (3600–83 000)	.84
Antibiotic regimen
Streptomycin + rifampicin	33 (81)	26 (84)	7 (70)	.33
Clarithromycin + rifampicin	7 (17)	4 (13)	3 (30)	.37
Both (sequential treatment)	1 (2)	1 (3)	<0 (0)
Need for surgery
Debridement	33 (80)	25 (81)	8 (80)	1.00
Skin graft	25 (61)	17 (55)	8 (80)	.26

Data are presented as No. (%) unless otherwise indicated.

Abbreviations: AFB, acid-fast bacilli; IQR, interquartile range; PCR, polymerase chain reaction; PR, paradoxical reaction; WHO, World Health Organization; ZN, Ziehl–Neelsen.

Management

Most patients received streptomycin plus rifampicin (Table 1). Surgery was performed in 80% of cases (33 patients), with several interventions in some cases, including skin grafting in 25 patients (61%). One patient (female) was switched from a streptomycin regimen to a clarithromycin regimen because of streptomycin intolerance. Complete healing was achieved within 1–6 months except for the 2 patients with the largest ulcerative lesions, who underwent 1 or more large surgical debridement, for whom complete healing was longer and functional impairment could not be prevented.

Paradoxical Reactions

Ten of the 41 patients (24%) included in this cohort had a PR during or after antibiotic treatment. The median time from the start of treatment to PR was 75 (IQR, 63–98) days (range, 30–100 days). PR was diagnosed on the basis of an enlargement of a preexisting lesion in 5 patients, whereas the other 5 patients presented a new ulcerative lesion. PR was diagnosed later (albeit not significantly so) in patients with a new lesion than in patients with an enlargement of a preexisting lesion (median time from the start of treatment to PR diagnosis: 90 [IQR:90–100] vs 58 [IQR:30–70] days, respectively; P = .08, Mann–Whitney test).

Change in Biological Markers During BU Treatment in the Whole Cohort

White Blood Cell Counts

In the whole cohort, neutrophil counts decreased significantly after the start of treatment, from 6.2 ± 5.1 × 10⁹/L (mean± SD) at baseline to 3.5 ± 1.5 × 10⁹/L on day 30, and 2.8 ± 1.5 × 10⁹/L on day 60 (P < .001) (Figure 2). By contrast, eosinophil counts increased significantly, from 0.46 ± 0.54 × 10⁹/L at baseline to 0.68 ± 0.62 × 10⁹/L on day 30 and 0.70 ± 0.65 × 10⁹/L on day 60 (Figure 2). Lymphocyte and monocyte counts did not vary significantly over the study period. On days 30 and day 60, neutrophil counts remained higher in patients with edematous forms (5.7 × 10⁹/L ± 2.4 and 3.6 × 10⁹/L ± 1.4, respectively) than in patients with ulcerative forms (2.9 × 10⁹/L ± 1.5 and 2.5 × 10⁹/L ± 1.2, respectively) (P = .001 and P = .02, respectively, Mann–Whitney tests). No other significant differences between the ulcerative and nonulcerative BU subgroups were observed for other leukocyte populations on days 30, 60, and 90.

Figure 2.

White blood cell (WBC) counts in the whole cohort from baseline to day 90 (horizontal lines represent medians, boxes are drawn from quartiles 1 to 3, and whiskers indicate the minimum to the maximum). Statistical analyses were performed with paired, nonparametric Wilcoxon test (Benjamini–Hochberg false discovery rate method). *P < .05, **P < .01, ****P < .0001.

Quantification of Systemic Soluble Immune Mediators

IL-6, G-CSF, and VEGF were the soluble immune mediators for which median levels were significantly lower at all time points (Figure 3) than at baseline (day 0) levels. The levels of IL-4, IL-10, IL-12, and IL-1rα were significantly lower on days 60 and 90 than at baseline (Figure 3). Median levels of FGF, IP-10, IL-13, and IL-17 on day 90 and of TNFα, IL-1β, IL-7, IFN-γ, and PDGF-β on day 60 were significantly lower than those measured at baseline (Figure 3). The levels of eotaxin, GM-CSF, IL-2, IL-5, IL-8, IL-9, MIP-1α, and MIP-1β were not significantly different from baseline values at any of the time points considered.

Figure 3.

Change in biomarker levels for the whole cohort from baseline to day 90 (horizontal lines represent medians, boxes are drawn from quartiles 1 to 3, and whiskers indicate the minimum to the maximum). Statistical analyses were performed with paired, nonparametric Wilcoxon test (Benjamini–Hochberg false discovery rate method). *P < .05, **P < .01, ****P < .0001. Abbreviations: FGF, fibroblast growth factor; G-CSF, granulocyte colony-stimulating factor; IFN-γ, interferon gamma; IL, interleukin; IP-10, interferon-γ–induced protein 10; PDGF-β, platelet-derived growth factor beta; TNF-α, tumor necrosis factor alpha; VEGF, vascular endothelial growth factor.

Comparison of Patients With and Without Paradoxical Reactions

Clinical Findings

No demographic or clinical differences were found at baseline between patients who went on to develop a PR and those who did not (Table 1). There was a nonsignificant trend toward a lower bacterial load, as assessed by determining the proportion of acid-fast bacterium–positive samples in patients who underwent direct microscopic examination, or by qPCR, in patients who developed a PR than in patients who did not (Table 1).

Biological Findings

At baseline (Figure 4A), no significant differences were observed in the comparative analyses of leukocyte (neutrophils, eosinophils, lymphocytes and monocytes) counts between the PR and no-PR groups. Eosinophilia (eosinophil counts >0.5 × 10⁹/L) was observed in both groups, at baseline for 9 of the 31 patients (29%) of the no-PR group and 1 of 10 (10%) patients of the PR group, with no significant difference between groups (Figure 4A). Monocyte counts increased by 20% in the PR group and decreased by 15% in the no-PR group by day 60, with a significant statistical difference (P < .01) (Figure 4B).

Figure 4.

A, Leukocyte counts from baseline to day 90, plotted for the paradoxical reaction (PR) and no-PR groups. B, Changes in monocyte counts from baseline (start of antibiotic treatment) to day 90 in the PR and no-PR groups. Individual ratios of monocyte counts from baseline for the PR and no-PR groups (horizontal lines represent medians, boxes are drawn from quartiles 1 to 3, and whiskers indicate the minimum to the maximum). Statistical analyses were performed with unpaired, nonparametric Mann–Whitney tests. **P < .01.

The levels of soluble immune mediators and markers did not differ between the PR and no-PR groups at baseline (Figure 5A and Supplementary Figure 1), but for half of the cytokines and chemokines, levels were higher in the PR group than in the no-PR group for at least 1 time point (day 30, 60, or 90). TNF-α and IL-6 concentrations were significantly higher in the PR group than in the no-PR group on days 30, 60, and 90 (Figure 5A). The individual ratios were higher for IL-6 and TNF-α, but the difference was not statistically significant (Supplementary Figure 2). Similarly, IL-1β concentrations were significantly higher in the PR group than in the no-PR group on days 30 and 90, and IL-4 concentrations were significantly higher in the PR group than in the no-PR group on days 60 and 90 (Figure 5A). IL-7, IL-9, IL-10, IL-12, IL-17, FGF, and G-CSF concentrations were significantly higher in the PR group only on day 90 (Figure 5). The concentrations of eotaxin, VEGF, GM-CSF, IL-1rα, IL-2, IL-5, IL-8, IL-13, IFN-γ, MIP-1α, MIP-1β, and PDGF-β did not differ significantly between the PR and no-PR groups at any of the time points considered. These changes in the concentrations of soluble immune mediators over time, in the PR and no-PR groups, are summarized in Figure 5B.

Figure 5.

A, Changes in the levels of immune biomarkers between baseline and day 90 in the PR and no-PR groups. Statistical analyses were performed with unpaired, nonparametric Mann–Whitney tests. *P < .05, **P < .01. B, Heatmap of normalized values for each soluble immune marker, with a comparison of median values at baseline and on days 30, 60, and 90. ▴ indicates significant difference for day 30 relative to day 0; ♦ indicates significant difference for day 60 relative to day 0; ★ indicates significant difference for day 90 relative to day 0. Abbreviations: FGF, fibroblast growth factor; G-CSF, granulocyte colony-stimulating factor; IFN-γ, interferon gamma; IL, interleukin; IP-10, interferon-γ–induced protein 10; MIP-1, macrophage inflammatory protein 1; PDGF-β, platelet-derived growth factor beta; PR, paradoxical reaction; TNF-α, tumor necrosis factor alpha; VEGF, vascular endothelial growth factor.

Comparison of Systemic Soluble Factors Between BU Patients and Controls

The levels of systemic soluble factors in BU patients at baseline were compared to those measured in 2 control groups (see Methods). For 10 of the soluble factors, including IL-6 and TNF-α, there was no significant difference between the 3 groups (Figure 6). In contrast, levels of IL-17 and IP-10 were significantly higher in patients with BU than in healthy blood donors (Figure 6).

Figure 6.

Levels of interleukin 6 and tumor necrosis factor alpha between patients with Buruli ulcer (BU) and control groups. Results for BU patients (n = 31) were obtained at day 0, before antibiotic treatment. Non-BU patients (n = 60) are patients managed for another condition contemporaneously in the same hospital. Healthy controls (n = 19) are healthy blood donors sampled in France. Comparisons were performed with Kruskal–Wallis test with Dunn test for multiple comparisons. *P < .05, ***P < .001. Abbreviations: BU, Buruli ulcer; FGF, fibroblast growth factor; G-CSF, granulocyte colony-stimulating factor; IL, interleukin; IP-10, interferon-γ–induced protein 10; TNF-α, tumor necrosis factor alpha.

DISCUSSION

In this observational cohort study, 24% of the 41 consecutive BU patients studied had a PR, a proportion close to that observed in other cohorts from West Africa (53 of 241 patients [22%] [12]; 47 of 354 patients [13%] [5]) and Australia (32 of 156 [21%] [3]). These paradoxical reactions occurred with a median of 10 weeks after treatment initiation, versus 6 and 10 weeks in the previously published West African cohorts [5, 12] and 5.6 weeks in the Australian cohort [3]. There was a trend toward a higher risk of PR in patients with an edematous or disseminated (WHO category III) form of BU, although no significant difference could be found at baseline, probably due to the small number of patients included in our study. Consistent with published findings, our results indicate that clinicians managing BU patients should be vigilant concerning the risk of PR after the first month of antibiotic treatment, particularly in patients with edematous or severe forms of the disease. However, the occurrence of PR was not restricted to a single clinical form of BU, and we therefore looked for biological predictors of this complication.

In 1 West African cohort, a higher baseline bacterial load in lesion samples, as assessed by microscopy, qPCR, or culture, was associated with a higher risk of PR [5]. Our results did not confirm these findings: Microscopic evaluation at baseline was slightly less frequently positive in patients who subsequently experienced PR, and we found no difference between the 2 groups in terms of qPCR findings. However, it was not possible to draw any firm conclusions on this point, given the small number of patients studied and the variability of the sampling method. Indeed, fine-needle aspiration was more frequently performed for diagnosis in patients who went on to develop a PR. This sampling method, which is preferred for patients with an edematous form of the disease, is intrinsically less sensitive than biopsy or swabs [11, 13]. The nature of the sampling method is thus an important parameter that should be taken into account whenever bacterial load is considered in BU.

We evaluated the utility of several blood markers for predicting the occurrence of PR in patients treated for BU. We were unable to identify any baseline biological signature associated with PR. However, the pattern of change in IL-6 and TNF-α levels differed between the PR and no-PR groups. The median concentration of these 2 cytokines decreased over time in the patients in the no-PR group, whereas it remained stable in patients in the PR group. Importantly, this difference between the 2 groups was observed from 1 month after the start of antibiotic treatment, which is before the clinical diagnosis of PR in most cases. These biomarkers should, therefore, be evaluated further, to assess their ability to detect PR early, thereby making early treatment possible.

The determination of these cytokines is costly and there is therefore a need to explore cheaper alternatives. Unfortunately, we did not determine levels of C-reactive protein (CRP), the concentration of which is closely related to IL-6 concentration in various conditions [14–16], and which is widely used as a marker of inflammation. We observed no significant differences in blood cell counts between the 2 groups at baseline. However, monocyte count tended to have increased by days 30 and 60 in the PR group, whereas it decreased in the non-PR group. Overall, our results suggest that patients who experience a PR during treatment for BU are characterized by an inflammatory phenotype that tends to persist during the first few months of treatment.

We performed a longitudinal assessment of leukocyte counts and soluble immune mediator concentrations during treatment for BU. We show that the concentrations of 16 proinflammatory mediators decreased during treatment, whereas no increases in concentration were observed. Neutrophil counts tended to decrease and eosinophil counts tended to increase, as observed in other bacterial infections [17]. Furthermore, we compared the level of systemic soluble factors between BU patients at baseline with those measured in 2 control groups, and we did not see any immunosuppressive systemic signature in BU patients. This result conflicts with the findings of previous studies suggesting that BU has a systemic immunosuppressive signature [18, 19]. Nevertheless, it is noted that the CRP value was measured in 1 of these studies [19], and the results showed a clear decrease in CRP between day 0 and day 90, which is coherent with our study. Furthermore, the main virulence factor of M ulcerans, the macrolide toxin mycolactone, is considered an immunomodulatory molecule, essential to M ulcerans colonization and immune-escape at the skin tissue level [20–23]. Our results do not necessarily question the immunomodulatory role of mycolactone. Indeed, mycolactone immune effects could be restricted to the site of infection, whereas our study focused on systemic data.

Our study has several major limitations, the most important of which is the small number of patients included. In addition, we did not determine the concentration of 1 important inflammatory marker. Nevertheless, our results provide important information about changes in the concentrations of inflammatory markers in patients with treated BU, and the biological predictors of PR.

CONCLUSIONS

This study confirms that paradoxical reactions are frequent in patients treated for BU. An analysis of several immune and inflammatory markers revealed an inflammatory profile at baseline. Patients who went on to develop a PR were characterized by the persistence of this inflammatory profile, which generally resolved within 1 month in the other patients. IL-6 and TNF-α concentrations and changes in leukocyte count should be further evaluated to assess their ability to predict paradoxical reactions.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. M. P. and V. D. analyzed data and wrote the manuscript. E. M. collected data, performed experiments, and wrote the manuscript. M. R.-S. analyzed data. L. G., A. C., A. A., and C. J. collected data. S. B. performed experiments. L. M. conceptualized and obtained funds for the project.

Financial support. This work was supported by the Centre Hospitalier Universitaire CHU Angers (to L. M.); Fondation Raoul Follereau (to L. M.); Inserm (to L. M.); équipe FRM Fondation pour la Recherche Médicale (to E. M.); and Université Angers (to E. M.). Funding to pay the Open Access publication charges for this article was provided by Centre Hospitalier Universitaire Angers and Université Angers.