Background
Cerebral malaria (CM) and severe malarial anemia (SMA) are the main drivers of morbidity and mortality due to
Plasmodium falciparum. CM is characterized by coma and has a mortality rate of 13–15% [
1,
2]. In CM, adhesion of infected erythrocytes (IEs) to other uninfected erythrocytes (UEs) (rosette formation) and sequestration of IEs, leukocytes, and platelets to the blood-brain barrier (BBB) endothelium, combined with an imbalanced immune response and endothelium activation are thought to lead to BBB dysfunction and adverse clinical outcomes [
3‐
7]. Within CM, malarial retinopathy has been proposed to distinguish “true” CM (retinopathy positive, RP) from coma due to other causes, with incidental
P. falciparum parasitemia (retinopathy negative, RN) [
8]. However, it has been suggested that RN CM may be part of the clinical spectrum of CM [
9]. Assessment of parasite gene expression could help determine whether parasite virulence factors expressed in RP CM are also associated with RN CM.
SMA is characterized by severe anemia and inflicts a substantial burden in sub-Saharan Africa, causing 20% of
P. falciparum hospitalizations [
10]. In settings where there is easier access to blood transfusions, the mortality from SMA is lower than that from CM (1–5%) [
1,
11]. In SMA, destruction of IEs and UEs, dyserythropoiesis, and suppression of erythropoiesis are considered important contributors to severe anemia. Little is known about how parasite virulence factors contribute to the development of these different clinical manifestations of severe malaria.
P. falciparum erythrocyte membrane protein 1 (PfEMP1) is considered a key virulence factor in malaria, as it binds to various host receptors on the endothelium or UEs (rosetting) to sequester infected erythrocytes from circulation and destruction in the spleen [
12‐
17]. PfEMP1 is a target of antibody-mediated immunity [
18], and in response, PfEMP1 molecules have diversified extensively. Despite this extensive sequence variation, PfEMP1 function is conserved, and PfEMP1 molecules have a highly ordered domain composition, kept in check by highly ordered organization and mechanism of recombination of the encoding
var genes [
19‐
21]. Thus, each haploid parasite genome carries 50–60 polymorphic
var genes [
14,
22,
23], divided by chromosomal location and direction of transcription into groups A, B, and C. The extracellular portion of PfEMP1 varies in organization and length but comprises a combination of Duffy binding-like domains (DBLα-ζ) and cysteine-rich interdomain regions (CIDRα-δ) [
20,
21]. The N-terminal domain composition of PfEMP1 is conserved and linked to the genetic control of
var groups. Group A
var genes encode PfEMP1 with CIDRα1 domains shown to bind endothelial protein C receptor (EPCR) [
24] or a set of more diverse CIDRβ/γ/δ domains of unknown function, but potentially associated with rosetting [
16]. Groups B and C
var genes encode cluster of differentiation 36 (CD36)-binding PfEMP1 [
25]. One exception to this rule is the so-called conserved tandem arrangements known as domain cassette 8 (DC8) PfEMP1 [
21], which is a group A-like EPCR-binding PfEMP1, recombined into a group B
var gene location.
Consensus from previous studies of
var gene expression in patients shows that expression of group A and DC8
var genes is associated with severe malaria [
26‐
32]. Specifically, group A and DC8 PfEMP1 that bind EPCR have been suggested to play a key role in severe malaria, through their ability to support IE binding to various microvasculature beds [
33,
34] and through reducing the production and cytoprotective effects of activated protein C, due to functional impairment of EPCR upon PfEMP1 engagement [
35‐
37]. As a result, the extent of PfEMP1-EPCR binding could determine the amount of sequestration, coagulation defects, endothelial activation, and permeability, which in turn could define the outcomes of severe malaria. In line with this, EPCR-binding PfEMP1 transcript levels were recently associated with increased disease severity, from asymptomatic infections to both SMA and CM, in Tanzanian children [
32]. More studies are needed to confirm these findings. In particular, the importance of EPCR-binding PfEMP1 in RP vs. RN CM is not well understood.
In the current study, we used qRT-PCR primers with coverage and high specificity [
32] for group A and DC8
var genes to assess differential gene expression in parasites from Ugandan children with CM vs. SMA, in children with CM with vs. without retinopathy, and in children with CM who died vs. those who survived. The primers used in this study have been recently designed [
32] based on the analysis of 226
var genomes as compared to only 7 used by the previous studies in the field [
31,
38]. As a result, these primers provide the best coverage to date, and the current study presents the first time they are used to study the association of
var types with CM discriminated by retinopathy.
Methods
Study design
This prospective cohort study with the overall goal of understanding the effects of severe malaria on neurodevelopment was conducted at Mulago National Referral and Teaching Hospital in Kampala, Uganda in 2008–2015 and enrolled children with CM, children with SMA, and community children (CC). The study was reviewed and approved by the Ugandan National Council for Science and Technology (UNCST), the Makerere University School of Medicine Research and Ethics Committee, and the University of Minnesota Institutional Review Board. Written informed consent was obtained from parents or guardians of study participants.
Children between 18 months and 12 years of age, meeting the World Health Organization definition for CM or SMA, were recruited from the Acute Care Unit at Mulago Hospital as previously described [
1]. CM was defined as (1) coma (Blantyre coma score [BCS] ≤ 2), (2)
P. falciparum on blood smear, and (3) no other known cause of coma. SMA was defined as presence of
P. falciparum on blood smear in children with hemoglobin
< 5 g/dL. Exclusion criteria for children with SMA included any impairment of consciousness or having > 1 seizure. Children with severe malaria were managed according to the Ugandan Ministry of Health treatment guidelines at the time, which included quinine treatment [
1].
CC were recruited from the nuclear family, extended family, or household compound area of children with CM or SMA. Eligible CC were aged 18 months to 12 years and currently healthy. A blood smear was taken from these children at the time of enrollment, and those who had any density of P. falciparum on the smear are indicated here as asymptomatic parasitemic (AP). Exclusion criteria for all children included (1) known chronic illness requiring medical care, (2) known developmental delay, or (3) prior history of coma, head trauma, cerebral palsy, or hospitalization for malnutrition. A total of 269 children with CM, 232 children with SMA, and 217 CC were enrolled in the study. Of the 217 CC, 32 had asymptomatic parasitemia.
Sample collection and RNA isolation
Whole blood was collected at enrollment in PAXgene Blood RNA preservative solution (PreAnalytiX, Hombrechtikon, Switzerland) in a ratio of 2.76 mL of additive per mL of blood. The samples were stored long term at –80 °C. RNA was isolated using the PAXgene Blood RNA Kit (PreAnalytiX, Hombrechtikon, Switzerland).
Primer design
Primers were designed and optimized and previously described [
32]. Briefly, the primers used in this study were designed based on full-length DBL and CIDR domain encoding sequences from seven
P. falciparum genomes and 226 Illumina whole genome sequenced
P. falciparum field isolates [
32]. Primer sequences, coverage, and specificity are depicted in Additional file
1: Figure S1.
Quantification of var transcript levels by qRT-PCR
Total RNA was treated with DNase I (Invitrogen, Carlsbad, CA, USA). Complementary DNA (cDNA) was synthesized using random hexamers and the SuperScript® III First-Strand Synthesis System (Invitrogen, Carlsbad, CA, USA) according to manufacturer’s instructions. qRT-PCR was performed in 20-μL reactions using KiCqStart® SYBR® Green qPCR ReadyMix™ (Sigma-Aldrich, St. Louis, MO, USA) with the 7500 Real Time PCR System (Applied Biosystems, Foster City, CA, USA). Amplification was performed following the previously published conditions [
31], and data was collected at the final elongation step. No reverse transcriptase and no template controls for both housekeeping genes were included in the plates to rule out DNA contamination in the RNA samples and any nucleic acid contamination in reagents, respectively. Gene expression was normalized to the average of two housekeeping genes: seryl-tRNA synthetase and fructose-bisphosphate aldolase (ΔC
t var_primer = C
t var_primer − C
t average_control primers). ΔC
t var_primer was transformed into arbitrary transcript units using T
u = 2
(5−ΔCt). Only samples that had a C
t average_control below 25 were included in the analysis. Melting temperature analysis was performed for each target, and only samples with
T
m within 1.7 °C of median
T
m were analyzed. If only primer dimers or non-specific larger targets were detected, T
u for that target was assigned as 1.
Laboratory testing
Peripheral blood smears were assessed for
Plasmodium species by microscopy with Giemsa staining using standard protocols. Blood culture was performed with the Bactec 9050 Blood Culture System (Becton Dickinson, Franklin Lakes, NJ, USA). Blood culture samples negative by this method were further cultured on blood agar or chocolate agar to further rule out bacterial infection. PfHRP-2 quantification was performed using the Malaria Ag CELISA (Cellabs, Brookvale, Australia). Sequestered parasite biomass was calculated as previously described [
39]. Plasma soluble intercellular adhesion molecule-1 (sICAM-1), vascular cellular adhesion molecule-1 (sVCAM-1), and soluble P-Selectin and E-Selectin were measured by magnetic cytometric bead assay in plasma diluted 1:300 (R&D Systems, Minneapolis, MN, USA) according to manufacturer’s instructions with a BioPlex-200 system (Bio-Rad, Hercules, CA, USA). Plasma angiopoietin-2 (Ang-2) and von Willebrand factor (VWF) levels were quantified using the human angiopoietin-2 DuoSet ELISA kit (R&D Systems, Minneapolis, MN, USA) and the REAADS von Willebrand Factor activity ELISA kit (Corgenix, Broomfield, CO), respectively. Soluble EPCR levels in plasma were quantified using the Asserachrom® sEPCR immunoassay (Stago Group, Gennevilliers, France) according to manufacturer’s instructions.
Malarial retinopathy diagnosis
Children were assessed for malarial retinopathy by indirect ophthalmoscopy. Ophthalmoscopy was done by medical officers in all CM patients on admission, and repeated every 24 h while they remained comatose. Before each examination, the pupils were dilated with sequential instillation of cyclopentolate 1% and tropicamide 1%. Using a binocular indirect ophthalmoscope, an eye exam was performed 30–60 min later. The medical officers were trained by an ophthalmologist experienced in the evaluation of malarial retinopathy. The study investigators and ophthalmologist then continued training and assessing the study medical officers for accuracy in this assessment and recording of the ophthalmoscopic finding. Children with retinopathy on any exam were classified as RP.
Statistical analysis
Data was analyzed using Stata/SE 12.1 (StataCorp, College Station, TX, USA). Transcript abundance of var genes was compared between disease groups using the Mann-Whitney U test. Clinical and laboratory findings for children in the different disease groups were compared using the chi-squared test for categorical data and t tests for continuous measures. Associations between var types and parasite biomass, sequestered parasite load, and markers of endothelial activation and anemia were determined by Spearman’s correlation and adjusted for multiple comparisons by a Bonferroni correction. Tu for group A-EPCR binders was determined as the sum of [CIDRa1.4, CIDRa1.5a, CIDRa1.5b, CIDRa1.6b, and CIDRa1.7]Tu-4; Tu for group B-EPCR binders was determined as the sum of [CIDRa1.1, CIDRa1.8a, and CIDRa1.8b]Tu-2; Tu of CIDRα1 EPCR binders was calculated as the sum of [CIDRa1.1-CIDRa1.8b]Tu-7.
Discussion
In the present study, we show that children with severe malaria have higher levels of both EPCR-binding group A and DC8 PfEMP1 transcripts than children with asymptomatic parasitemia, that transcript levels of EPCR-binding PfEMP1 are higher in children with CM than SMA, that children with both CM and SMA have higher levels of EPCR-binding group A PfEMP1 transcripts than children with CM alone, and that PfEMP1 transcript levels in RN or PfHRP-2-low CM fall between those in RP CM and those in SMA. Together the findings suggest that not only the presence, but more importantly, the transcript level and therefore the extent of EPCR binding by PfEMP1 may be important in determining the clinical manifestation of SM. A particularly important and novel finding in the current study is the progressive increase in EPCR-binding PfEMP1 expression through the stages of malaria infection and disease, from asymptomatic parasitemia, in which there is very little expression, through SMA to RN CM and RP CM. The finding that EPCR-binding PfEMP1 expression in RN CM falls between that of SMA and RP CM, and far above that in AP, suggests that
P. falciparum plays a role in the disease process of many children with RN CM, and that RN CM represents a milder disease, a finding consistent with a recent study of clinical manifestations of RP vs. RN CM in this cohort [
9].
The findings regarding EPCR-binding PfEMP1 expression are largely consistent with conclusions drawn from two recent studies in Tanzania, showing that CIDRα1 was the only common domain encoded by most prominently expressed
var transcripts in CM and SMA patients [
43], and that higher levels of EPCR-binding PfEMP1 transcripts were associated with increasing symptoms of severity in patients suffering uncomplicated malaria vs. SMA or CM [
32]. However, in contrast to the present study, the latter study [
32] found no difference in transcript levels of EPCR-binding PfEMP1 between Tanzanian children with CM and SMA, despite application of the same primer set in both studies. The current study had a larger CM group with higher mortality than in the Tanzanian study, and it did not have any mortality in the SMA group. The larger sample size and greater disease severity and mortality in children with CM than SMA in the present study as compared to the Tanzanian study may explain why higher PfEMP1 transcript levels in children with CM as compared to SMA were seen in the present study but not the Tanzanian study [
32]. Likewise, expression of other PfEMP1 traits or variants may also account for differences observed between the two populations. In this study, we did not quantify CD36-binding PfEMP1; thus, we cannot infer on the total expression levels of all
vars, or the proportion of transcripts encoding EPCR-binding PfEMP1 between CM and SMA.
We have found only one other study to date that examines PfEMP1 transcript levels in RP vs. RN CM. In this cohort of Kenyan children, the authors did not find any significant difference in group A, DC8, and CIDRα1.4 transcript levels between RP and RN CM [
38], although a higher proportional expression of group A and DC8 compared to group B and C
var genes was found in patients with RP compared to those with RN [
38]. The present study, which uses primers with a better coverage, found only higher levels reported by the CIDRa1.4/6a primers in RP compared to RN CM (Fig.
3). The subset of group A
var genes targeted by these primers include the so-called domain cassette 13 PfEMP1, which has been shown to often bind both EPCR and ICAM1 [
44,
45] and to provide higher binding levels to endothelial cells [
44]. Moreover, this PfEMP1 subset has been shown to be more frequently expressed, albeit at lower levels, in patients with CM compared to those with SMA [
32,
45]. It is therefore possible that dual EPCR- and ICAM1-binding PfEMP1 account for the higher transcript levels reported by the CIDRa1.4/6a primers in the present study between CM/SMA vs. CM, and RP vs. RN CM patients. Further studies are required to elucidate this hypothesis. We did not assess proportional expression, because transcript levels are not absolute values, and no study captures 100% of
var diversity in a patient, so proportional values can be strongly influenced by outlier values.
The present study provides two important additional pieces of information: RN CM transcript levels fall between those of two forms of severe malaria, RP CM and SMA, and transcript levels are similar in children above and below a proposed PfHRP-2 cutoff level that would indicate “true” CM. Together the findings provide evidence suggesting that
P. falciparum sequestration via PfEMP1 plays a role in the development of RN CM. The finding that PfEMP1 expression did not differ between those with levels above and below a suggested cutoff for PfHRP-2 levels to define “true” CM [
42] also suggests that PfHRP-2 levels may be less useful than hoped in distinguishing “true” CM from coma due to other causes with incidental parasitemia. Assessment of
var transcript levels in the field is unlikely to ever be a practical diagnostic tool, but it could be very useful in future research studies of CM for attributing coma to
P. falciparum or another cause. Retinopathy could also be occurring at levels not detectable by standard funduscopic exam, and our study medical officers may have occasionally missed retinal findings that would be seen by an ophthalmologist, but having received training and validation of testing mid-study from highly experienced ophthalmologists, they likely represent a “gold standard” for field ophthalmoscopy testing. Newer technologies for assessing retinopathy with camera and/or radiologic imaging may provide better understanding of the extent to which “subclinical” retinopathy is occurring, but these methods are also likely to remain limited to research.
Interestingly, transcript levels reported by the DBLa1ALL primers and the summarized levels from primer sets specific to genes encoding CIDRα1 domains were lower in children who died, despite their having higher PfHRP-2 levels as compared to survivors. This remained true when analysis was restricted to RP CM, confirming that death was most likely caused by
P. falciparum infection. A similar trend towards lower
var transcript abundance in children with CM who died was observed in one previous study [
31] but not in a more recent study [
32]. These inconsistencies may reflect the complexities of the disease at the end of life complicated by the limited number of samples for children who died. DBLα1ALL and CIDRα1 transcript levels were particularly low for around half of the children who died. In these children, a transcript level above baseline was picked up by the DBLa2/1.1/2/4/7 primers, suggesting that either rare group A or CD36-binding PfEMP1 was expressed, and possibly associated with death in these children. Possible biological reasons for an altered
var profile compared to that for surviving SM patients include that a particularly adverse host response to infection, unrelated to or even allowing diverse PfEMP1 phenotypes, led to death. Even though we found the same results for RP children, it cannot be completely ruled out that another co-infection that increases the risk of mortality in SM, such as bacteremia [
46], could be contributing to death in those children with CM who have low group A and CIDRa1 transcript levels. However, in the present study we did not find an association between PfEMP1 transcript levels and the presence of bacteremia in children with CM. Deeper characterization of the
var transcripts in these patients, as well as thorough testing for other co-infections, may offer clues as to the reason for the unexpected finding of lower group A and DC8
var transcript levels in children with CM who died.
The present study does not provide clear information on the clinical relevance of rosetting. While most rosetting PfEMP1 types are group A and carry DBLα1.5/6/8 domains [
40,
41], it is still unclear if such domains, or a specific subset of these, consistently confer rosetting. In future studies, we plan to assess transcript levels of groups B and C CD36-binding PfEMP1, which have shown to be similar [
28] or higher [
47] in AP as compared to uncomplicated malaria or SM in prior studies. We did not enroll children with uncomplicated malaria in this study, and assessment of PfEMP1 transcript levels in this group, who represent another important comparison group of malaria without severe manifestations, will be important for future studies. However, the AP group in this study had no history of prior SM, and only one experienced SM over the 2 years of follow-up, despite presumably similar malaria exposure (since they lived in the same extended households as children with SM). As a result, AP represents an important comparison group, since parasites from patients with uncomplicated malaria could still express some of the domains associated with SM, even though the children present with uncomplicated malaria, because they are protected from development of SM by early treatment.
Only results reported by DBLa2/1.1/2/4/7 primers were weakly but not significantly associated with endothelial activation (specifically increased sVCAM-1 levels) in SM. This suggests that at this stage of the disease, pathways that lead to sequestered parasite load and endothelial activation are more complex than simply PfEMP1 binding to host receptors. In the current study, none of the EPCR-binding PfEMP1 transcript levels were associated with plasma levels of sEPCR, suggesting that binding of PfEMP1 to EPCR might prevent shedding of EPCR in an inflammatory context. This potential mechanism would be interesting to explore in vitro with parasite strains that bind specifically to EPCR.