Cytokine Network in Adults with Falciparum Malaria and HIV-1: Increased IL-8 and IP-10 Levels Are Associated with Disease Severity

Background Co-infection with malaria and HIV increases the severity and mortality of both diseases, but the cytokine responses related to this co-infection are only partially characterised. The aim of this study was to explore cytokine responses in relation to severity and mortality in malaria patients with and without HIV co-infection. Methods This was a prospective cross-sectional study. Clinical data and blood samples were collected from adults in Mozambique. Plasma was analysed for 21 classical pro- and anti-inflammatory cytokines, including interleukins, interferons, and chemokines. Results We included 212 in-patients with fever and/or suspected malaria and 56 healthy controls. Falciparum malaria was diagnosed in 131 patients, of whom 70 were co-infected with HIV-1. The malaria patients had marked increases in their cytokine responses compared with the healthy controls. Some of these changes, particularly interleukin 8 (IL-8) and interferon-γ-inducing protein 10 (IP-10) were strongly associated with falciparum malaria and disease severity. Both these chemokines were markedly increased in patients with falciparum malaria as compared with healthy controls, and raised levels of IL-8 and IP-10 were associated with increased disease severity, even after adjusting for relevant confounders. For IL-8, particularly high levels were found in malaria patients that were co-infected with HIV and in those who died during hospitalization. Interpretations Our findings underscore the complex role of inflammation during infection with P. falciparum, and suggest a potential pathogenic role for IL-8 and IP-10. However, the correlations do not necessarily mean any causal relationship, and further both clinical and mechanistic research is necessary to elucidate the role of cytokines in pathogenesis and protection during falciparum malaria.


Introduction
Human Immunodeficiency Virus (HIV) 1 and Plasmodium falciparum malaria are two of the major lethal infectious diseases in sub-Saharan Africa, with 1.47 million and 1.13 million deaths, respectively, in 2010 [1,2]. Co-infection with malaria and HIV presents challenges in both diagnosis and therapy and results in increased severity and mortality of both infections [3][4][5].
Infection with P.falciparum is often associated with the activation of several inflammatory pathways, resulting in systemic and local inflammatory responses that involve the production of a wide range of cytokines,-from several cell types including leukocyte subsets, endothelial cells and various tissues cells such as macrophages. This inflammatory response during malaria infection could activate anti-microbial responses in macrophages, T cells, granulocytes and endothelial cells and thereby be beneficial for the host [6]. On the other hand, an overwhelming inflammatory response could be detrimental for the host by contributing to tissue destruction and excessive systemic inflammation that could contribute to morbidity and mortality during falciparum malaria. Rodent studies indicate that the malaria parasites themselves produce their own cytokines, which interfere with the host's immune response [7], but characterisation of the different actors in the activated cytokine network during clinical falciparum malaria is not fulfilled.
Although the immunological hallmark of HIV infection is a loss of CD4 + T cells and the development of severe immunodeficiency, it is also characterised by a state of chronic inflammation [8], and this non-resolving inflammation will further contribute to immunodeficiency through mechanisms such as immune exhaustion [9,10]. This persistent low-grade immune activation will affect immune responses to co-infection with other microbes, including P. falciparum.
However, in what way co-infection with HIV influences cytokine responses during falciparum malaria is still only partly known. In addition, although there are numerous studies on cytokine response during infection with P. falciparum, less is known about the more detailed cytokine network and the relative importance of different mediators with regard to disease severity and mortality in malaria patients.
Mozambique has one of the highest global incidences of co-infection with HIV-1 and falciparum malaria, with a population HIV-1 prevalence of 11.5% that is 22.5% in Maputo (adults 15-49 years, 2009) and a malaria parasite prevalence of 27.7% in small children in Southern Mozambique (2007) [11,12]. To characterise inflammatory responses during infection with P. falciparum with and without coinfection with HIV-1, we analysed a wide range of cytokines and related inflammatory mediators in blood samples obtained at admission from adult patients admitted to Central Hospital of Maputo in Mozambique with fever and/ or suspected malaria. The levels of the different mediators were related to disease severity, mortality, and co-infection with HIV.

Study area and participants
The Central Hospital of Maputo is a public teaching hospital for Maputo's 1.2 million citizens and a national referral hospital for Mozambique's population of 22 million. From 8 January 2011 to 31 March 2011 and from 7 November 2011 to 14 March 2012, we conducted a prospective cohort study in all patients consecutively admitted to the Medical Emergency Department in the Central Hospital in Maputo every week from Sunday to Friday, as previously described [13]. Most of the patients and most of the healthy controls came from the more peripheral suburbs of Maputo city, including marshland. Malaria is mesoendemic and most frequent from November to April, with P. falciparum accounting for 95-100% of the malaria cases [12,13]. The study was designed and performed according to the Helsinki Declaration, as adopted by the 59th WMA General Assembly, Seoul, Republic of Korea, October 2008, and was approved by The National Ethical Committee at the Ministry of Health in Mozambique and the Regional Ethical Committee in Eastern Norway. A signed consent or fingerprint was obtained from each patient or next of kin and from the healthy controls. Management of the patients was carried out according to the local standard of care by the hospital's doctors and staff.

Inclusion criteria
All non-pregnant adults $18 years with an axillary temperature $38.0˚C and/or suspected malaria were included, provided written consent. As controls, we included health care workers at the hospital, friends, and family members who reported a subjective feeling of wellbeing and had a healthy appearance, as evaluated by the researchers. Female healthy controls were not included if suspected or confirmed pregnant. We defined ''Suspected malaria'' as a history of fever, chills, headache, mental confusion, vomiting and/or diarrhoea, dyspnea, myalgia, and/or general malaise, provided that there were no other symptoms, findings on clinical examination, or additional diagnostic tests indicating other infections. Additional diagnostic tests and exams were basic laboratory tests (e.g., Hemoglobin (Hb), WBC, differential count, ESR, AST, ALT, ALP, bilirubin, L DH, creatinine); other blood tests (e.g., bacteriological and fungal culture and antigen/antibodies for HIV-1, HIV-2, CMV, EBV, hepatitis B and C); urine analysis (stix, micro, culture); Cerebrospinal fluid (CSF) analysis (erythrocytes, WBC, differential count, protein, glucose, chloride, syphilis and cryptococcal tests, bacteriological and fungal culture); sputum analysis for M. Tuberculosis with microscopy (AAFB) and culture; stool analysis (e.g., microscopy for ova and cysts, bacteriological culture); and cytological/histological and different radiological exams, if indicated. ''Malaria positive'' (confirmed malaria) was defined as a patient with a positive malaria PCR test. One patient who died with a positive malaria PCR test was categorised as a non-malarial death, because he had a history and clinical findings more compatible with tuberculous meningoencephalitis as the cause of death. No malaria PCR was performed for two patients: they had a positive HRP2-antigen test and were slide-positive with parasitemia 3+ and 5+, respectively, and were defined as malaria positive. HIV-infected patients were defined by a positive HIV serological test and/or a positive HIV PCR test. Malaria severity was categorised according to the number of criteria fulfilled for severe malaria, as defined by the WHO and adjusted to what could be measured locally, as previously reported [13,14]. The HIV disease severity was categorised according to the WHO HIV clinical staging, I-IV [15].

Clinical data and blood sampling
We recorded a predefined set of clinical data from the patients' files, which was performed as part of the routine clinical examination on admission and on follow-up during the hospital stay. The details are published elsewhere [13]. Blood samples from the patients and controls were collected from a pre-alcohol-cleaned peripheral vein and drawn into pyrogenic-free tubes with EDTA as a anticoagulant (plasma) or with no additive (serum).
The EDTA tubes were immediately put on melting ice and centrifuged within 30 minutes at 2000xg for 20 minutes to obtain platelet-poor plasma. The serum tubes were allowed to clot at room temperature prior to being centrifuged at 1000xg for 10 minutes. The serum and plasma samples were stored in multiple aliquots at 220˚C for 24 hours before being placed at 280˚C. The samples were thawed only once.

Laboratory analyses
According to the hospital routine and in line with standard procedures in the laboratory of the Central Hospital of Maputo HIV-test (Determine, Abbot Japan Co., Ltd. and Unigold, Trinity Biotech plc, Bray, Ireland), HRP-2 Rapid Diagnostic Test for malaria (RDT) (2010-2011 First Response Malaria antigen P. falciparum, Premium Medical Corporation Ltd., Daman, India; 2011-2012 ICT Malaria P.f., ICT Diagnostics Cape Town, South Africa), malaria thick blood slides (Giemsa 20% for 5 minutes), and other routine laboratory tests were performed. Parasitemia was categorised in a thick smear as 1+ when 1-10 parasites/100 fields, 2+ when 11-100 parasites/100 fields, 3+ when 1-10 parasites/ field, 4+ when 11-100/field, and 5+ when .100 parasites/field [16]. In addition, blood samples were taken separately for malaria and HIV PCR analyses, as previously described [13]. An HCG-urine pregnancy test was performed for the female patients of fertile age (Quick Vue, Quidel Corp., San Diego, CA, USA).

Statistical analysis
For the cytokine data, we report the means, medians, and 25 th and 75 th percentiles. Differences in categorical variables such as sex and HIV-status were tested by chi-square tests. Differences in cytokine distributions between the groups of study participants were tested by Mann-Whitney tests. Spearman rank correlation was used for calculating correlations between cytokines in all malaria patients (S1 Figure). Because this was an exploratory study with many interdependent markers, we used no statistical corrections for multiple comparisons. For multivariate analyses, due to high correlations among many of the cytokines, ordinary multivariable logistic regression was not suitable. Instead we used Pelora, a method constructed for supervised clustering in situations with strong correlations among predictors (here i.e. cytokines), and many predictors compared to the number of individuals (Pelora), which is recommended for such cytokine analyses [17,18], i.e. exactly the challenges we have in cytokine data [17]. Pelora finds groups of cytokines characterizing the property we are searching for, as which group of cytokines are characteristic for the difference between all malaria patients compared to healthy controls, for HIV positive malaria patients compared to the HIV negative malaria patients and for those with severe malaria compared to the patients with uncomplicated malaria. The ability of a cytokine group to distinguish between patient groups, or patients versus controls, is given as the ''predictive ability'', measured by the area under the operator characteristic (ROC) curve [19], and where the closer to 1,-the better. Each cytokine in a group may have a positive or a negative effect on the patient group membership probability [17,18]. According to the recommendations for this method, we included clusters into the models until there was a levelling off in the increase in predictive ability, and we adjusted for the effect of age, sex and where relevant HIV status. Prior to Pelora, the cytokine values were log transformed to normalize their skewed distribution, and the 1-2 most extreme values of the most skewed cytokines were adjusted down to 3SD from the mean.
For analyses involving variables for which there were missing observations for certain patients, these subjects were excluded; the number of included patients is indicated in the tables. In the calculation of malaria severity, missing observations were reckoned as ''normal'' in cases in which the investigation in question, e.g., bilirubin, was performed only on the clinical suspicion of specific organ involvement. Most of the statistical analyses were performed with SPSS-21 (IBM Corporation 1, New Orchard Road, Armonk, New York 10504-1722, USA, 914-499-1900), but the Pelora analysis and the correlations plots were performed in R version 3.0. (http://www.r-project.org) using the R-packages ''supclust''and ''corrplot'', respectively. There is a validation of all statistical methods used in S1 File.

Characteristics of study participants
A total of 212 non-pregnant adults with fever and/or suspected malaria and 56 healthy controls were included. Of the 212 patients, 131 (62%) had malaria, and 70 of these (53%) were co-infected with HIV-1 (Fig. 1). All malaria patients had P. falciparum; two also had double infection, one with P. vivax and the other with P. malariae. Four of the healthy controls had HIV infection, including one with P.ovale in addition. Those were excluded from further analysis rendering 52 healthy controls.
Among the malaria patients and healthy controls, the median ages were 38 years (range 18-79 years) and 26 years (range 18-56 years), respectively: 45% and 41% were women, and 98.5% and 96% were ethnic Mozambicans. There was a significant difference between the malaria patients and healthy controls in relation to age (p50.003), but not in relation to sex (p50.567). The malaria patients with and without HIV co-infection showed no significant differences in age, sex, and duration of symptoms; 66% of the patients (85/129) had severe malaria defined as one or more severity criteria (missing observations for two patients). The malaria patients with HIV co-infection in general had more severe disease, including inhospital mortality, as recently reported [13]. Of the malaria patients with HIV coinfection, 59% (41/70) had severe HIV infection, with HIV WHO stage 3 or 4. The HIV-infected patients had a median HIV viral load at 1.8 x 10 4 (range 0-8.3 x 10 6 ) HIV RNA copies/mL plasma (n562, missing blood sample analysis for eight patients). Unfortunately, data on the CD4 T-cell counts were lacking in all except 11 patients who had a median of 206 CD4 + T cells/mL (range 14-632).

Overall cytokine response
The cytokine responses were extremely variable both inter-and intra-individually, with levels ranging from below the lower detection limit to a 1000-fold increase above this limit. Within the malaria patient group, several of the cytokines were strongly inter-correlated, irrespective of co-infection with HIV, illustrating the marked changes in and close connections between cytokines during malaria infection (S1 Figure).

Cytokine levels in malaria patients compared with healthy controls
In general, the malaria patients had a pronounced and broad cytokine response compared with the healthy controls (Table 1). We observed significantly higher levels of IL-1ra, IL-6, IL-8, IL-9, IL-10, eotaxin, IP-10, MCP-1, and MIP-1b. Pelora first indicated an increased IP-10 response as the most important cytokine for differentiating between malaria patients and healthy controls, with a predictive ability of 0.973. The second most important group of cytokines were increased Cytokine levels in malaria patients with or without HIV coinfection.
In the malaria patients with HIV co-infection compared with those without, there was a significant increase in the levels of IL-8, eotaxin, and MIP-1a. (Table 2). Pelora extracted high IL-8 and eotaxin and low TNF, with a predictive ability of 0.697. When analysing HIV-infected malaria patients with (n513) and without (n533) antiretroviral therapy (ART) we saw no significant effects on cytokine levels (data not shown). However, most of the patients on ART had less than three months treatment and the data must therefore be interpreted with caution.

Cytokine levels related to disease severity
For patients with severe malaria (n584), as defined according to the adjusted WHO criteria [13,14], there were significantly increased levels of IL-8, IP-10 and MIP-1b compared with the patients with uncomplicated malaria (n545) ( Table 3). Severe compared to uncomplicated malaria was associated with significantly increased mortality rate with p50.017, with missing data for three patients. The Pelora analysis extracted increased IL-8 and IP-10 and reduced MCP-1 levels, with a predictive ability of 0.756 (Table 3).
Categorising the study population into healthy controls and patients with uncomplicated and severe malaria, there was a steady increase with increasing severity in those cytokines that were associated with severity, i.e., IL-8 and IP-10 ( Fig. 2).

Discussion
In the present study, marked changes in the cytokine response during infection with P. falciparum were observed in malaria patients compared with healthy controls. Some of these changes, particularly the IL-8 and IP-10 levels, were strongly associated with falciparum malaria and disease severity. Both chemokines were markedly increased in patients with falciparum malaria as compared with healthy controls, and raised levels of IL-8 and IP-10 were together with low MCP-1 levels, associated with increased disease severity, even after adjusting for relevant confounders. For IL-8, particularly high levels were also found in malaria patients that were co-infected with HIV and in those who died during hospitalization.  Previous studies on the impact of cytokines on clinical outcome in malaria have identified TNF, IFN-c, IL-1b, IL-6, IL-10, IL-12, transforming growth factor b and IP-10 as important mediators, but data on IL-8in this setting are more limited [20][21][22][23][24]. A Zambian study of falciparum malaria patients found no increased levels of IL-8, but the sample size was small, and few patients had severe malaria (8/30) [25]. In the present study, there was a strong and independent association of IL-8 with disease severity in adult falciparum malaria,. In support of our findings, in an older study from Thailand, IL-8 levels on admission were correlated with malaria parasitemia and disease severity, but the highest IL-8 levels were observed 14 days after admission, a time when viable parasites were no longer found [26]. More recently, an increased frequency of the genotype IL-8-251T/A, which leads to high IL-8 expression, has been associated with susceptibility to severe falciparum malaria in a meso-endemic area in India [27]. In children with malaria, high plasma IL-8 and IP-10 has been associated with increased severity and high cerebrospinal fluid IL-8 and IP-10 with mortality [28][29][30]. Nonetheless, clinical presentation, immune response, and mortality are often quite different in children compared with adults [31].
Several cells with relevance to severe falciparum malaria, including tissue macrophages and endothelial cells, can release IL-8 when activated. It is tempting to hypothesize that hemozoin and other parasite-related molecules can induce IL-8 release in these cells. [32]. Moreover, P. falciparum secretes a functional histamine-releasing factor homolog that induces IL-8 release from eosinophils [33]. The role of IL-8 with relevance for falciparum malaria include its ability to promote production of reactive oxygen species in monocytes and macrophages, to induce endothelial cell activation and to promote T cell chemotaxis and CD8 + T cell cytotoxicity [34]. IL-8 is a potent activator and chemo-attractant of neutrophils, and based on studies in murine models, it has also been suggested that neutrophils could be involved in sequestration and inflammation during severe malaria [35]. Moreover, a study of Nigerian children found that neutrophil extracellular traps (NETs) may induce pathology when they are infected with falciparum-malaria, but the same mechanism could potentially also be beneficial in adults by ''trapping'' the malaria parasites [36]. At the present, the role of neutrophils in the pathogenesis of clinical malaria is unclear. [35]. The function of the IL-8 receptor on erythrocytes has remained unclear to date, we can not exclude a role for IL-8 in erythrocyte sequestering during falciparum malaria. [37][38][39]. However, in spite of the hypothesis above, the role of IL-8 in malaria pathogenesis and protection remains unclear.
Increased levels of IP-10 have been associated with increased severity and mortality with cerebral malaria and with high viral load in HIV-infection [20,40]. In the present study, we confirmed these findings by showing a strong and independent association between plasma levels of IP-10 and disease severity in patients with falciparum malaria. Also, in the Pelora analyses high IP-10 showed the strongest association with falciparum malaria as compared with plasma levels of healthy controls. In addition to high plasma levels of IL-8 and IP-10, low plasma levels of MCP-1 were significant associated with disease severity during falciparum malaria. We have no explanation for this finding, but interestingly, MCP-1 was suggested to be beneficial in relation to falciparum malaria in Fulani children in Mali [29,41], and the role of MCP-1 in malaria is still unclear.
Besides IL-8, MCP-1 and IP-10, several of the cytokines that were associated with disease severity during falciparum infection belong to the CC or CXC chemokine family, i.e., eotaxin, MIP-1a, and MIP-1b. All of these are chemoattractants for different types of leukocytes, including some with important roles in malaria pathogenesis, as the eosinophil granulocytes. Thus, increased levels of those chemokines have been associated with organ failure, such as in malariaassociated acute respiratory distress syndrome (ARDS) and severe malaria, and occasionally with mortality [20,29,32,40,42,43].
We have previously shown that co-infection with HIV is significantly associated with increased malaria severity and death [13]. Patients co-infected with falciparum malaria and HIV had higher levels of IL-8, eotaxin and MIP-1a and lower levels of TNF compared with the patients without HIV. Eotaxin has previously been associated with anaemia in HIV-infected children with falciparum malaria [43]. IL-8 has been implicated in inflammatory responses during HIV infection as well as malaria [44]. Notably, although TNF has an established role in the pathogenesis of HIV infection and malaria, we found low rather than high levels of TNF, particularly in co-infected individuals. We have no clear explanation for this finding but it could potentially reflect that other markers such as soluble TNF receptors could be more reliable markers of TNF activity than TNF itself. It could potentially also reflect that when measuring a wide range of cytokines, trying to reflect the cytokine network, other conclusions may be obtained than when analysing one or a few cytokines. Nonetheless, this finding may underscore the complexity of the cytokine response during falciparum malaria.
The present study has some limitations. First, although most patients were severely ill, relatively few died, which limits conclusions concerning cytokine levels and mortality. Second, even if the WHO HIV clinical staging is considered an acceptable disease severity marker, the lack of information on CD4 and CD8 T cell counts is still an important limitation, in relation to the influence of HIV coinfection on the cytokine response during falciparum malaria. Third, since malaria severity tends to increase in older age groups, at least in non-endemic areas [45], the uneven age distribution may have ''exaggerated'' the differences in cytokine response between the patients and the healthy controls. Ideally, we should have had a selection process of the healthy controls achieving similar age groups. Fourth, assuming that tests usually done only on clinical suspicion of specific organ involvement were normal if not done, is not optimal, but still a reality in a low resource setting. Fifth, a large number of statistical tests were performed without adjusting for multiple comparisons. However, we consider it less probable that the main findings could be due purely to statistical coincidences because there was such a ''systematic'', significant trend in the IL-8 and IP-10 observations. In addition, the use of a statistical method (Pelora), which is particularly suited for the actual cytokine results, limits the influence of multiple comparisons.

Conclusions
In the present prospectively designed study with consecutive inclusion of falciparum malaria patients with and without HIV co-infection, we found marked changes in the cytokine network during malaria infection. In particular, high plasma levels of IP-10 and IL-8 were associated with falciparum malaria and disease severity in infected individuals, and for IL-8, particularly high levels were found in those that were co-infected with HIV. While several previous studies have appointed IP-10, IL-8 has been less described in relation to severe malaria.
Our findings underscore the complex role of inflammation during infection with P. falciparum. However, the correlations do not necessarily mean any causal relationship, and further clinical and mechanistic research is necessary to elucidate the role of cytokines in pathogenesis and protection during falciparum malaria.
Supporting Information S1 Figure. The correlation between the different cytokines in all malaria patients. Pairwise Spearman correlation is given between the cytokines in all malaria patients. The darker blue, the higher correlation as seen by the correlation scale at the right. The cytokines are sorted according to groupings of highly correlated cytokines, determined by the angular order of the eigenvector. doi:10.1371/journal.pone.0114480.s001 (TIFF) S1 File. Validation of the statistical methods presented in the manuscript. doi:10.1371/journal.pone.0114480.s002 (DOCX)