Background

Malaria, caused by infection with parasites of the genus Plasmodium, remains a significant public health concern [1]. Despite a vaccine in pilot implementation trials [2] and widespread application of control measures [3], the disease is still responsible for a huge burden of mortality and morbidity worldwide and a concerning increase in incidence has been seen in previously well-controlled areas [3].

With repeated exposure to infection, individuals in malaria-endemic regions develop naturally acquired immunity (NAI), first to the most severe clinical forms, such as cerebral malaria and then more slowly to infection itself [1]. Although the role of antibodies in controlling parasite density, symptomatology and severity of disease is well established [4, 5], less is known about mechanism in terms of the role of the innate and cellular immune responses [6]. Increased understanding of the immune response to malaria, in particular those that mediate NAI, could aid identification of diagnostic and prognostic markers, inform vaccine development and assist with the identification of treatment strategies to modify the immunological mechanisms mediating severe pathology [1].

Transcriptomics, which allows the expression of thousands of genes to be assessed in parallel for a single RNA sample, is an exciting, expanding area of research with vast potential application in the field of infection [7]. Facilitating a systems biology approach, gene expression data from high-throughput technologies (such as microarrays [8] and next generation sequencing enabling RNA sequencing for bulk cell populations and at single-cell resolution [9, 10]) can allow greater understanding of individuals’ response to infection. To date, expression data have been used to dissect mechanisms of vaccine immunogenicity [11], inform the design of new vaccines [12, 13], predict response to infection and outcome [14, 15], characterize and improve understanding of sepsis [16], and offer a novel approach to the diagnosis of infectious pathogens [17,18,19] together with RNA expression in the pathogen [20].

Given the limited understanding of the mechanisms of NAI to malaria from traditional immunological studies, a systems approach characterizing the gene expression patterns associated with infection could provide novel and valuable insights [21, 22]. Transcriptional profiling of the immune response to malaria in humans to date has sought to identify markers to aid diagnosis [23], to understand the pathogenicity of severe disease [24] and dissect the mechanisms of NAI [25, 26]. However, interpreting this body of work is difficult given considerable variation in study design, definition of disease, patient selection and methodology employed.

This review outlines a comprehensive analysis of all GEP studies of the human immune response to malaria with two aims: (i) to understand the application of this technology to date, in particular how these studies have informed understanding of NAI; and (ii) to determine the extent of variability in methodology between studies to allow informed comparison of data and interpretation of results.

Methods

A search of Gene Expression Omnibus (GEO) [27] for datasets including the search terms; ‘plasmodium’ or ‘malaria’ or ‘sporozoite’ or ‘merozoite’ or ‘gametocyte’ and ‘Homo sapiens’ was performed on 10th September 2019. Each of these datasets were examined and those not relating to the human immune response to malaria infection or using the Homo sapiens platform excluded. Of note, datasets of gene expression changes in relation to malaria vaccines were excluded.

Results

Studies identified

The search identified 30 GEO datasets. Seven of these datasets were excluded, as published analyses were unavailable. Twenty-three datasets and 25 related publications were therefore included in the final review (Table 1 and Additional file 1: Figure S1). All datasets related to Plasmodium falciparum infection except two that related to Plasmodium vivax infection (Table 1). The majority of datasets included samples from individuals infected with malaria ‘naturally’ in the field (n = 13, 57%), however some related to controlled human malaria infection (CHMI) studies (n = 6, 26%), or cells stimulated with Plasmodium in vitro (n = 6, 26%). Studies included samples from individuals with a wide range of ages (from 2 months—varying ages of adulthood) with differing degrees of prior exposure and, therefore, NAI to malaria. Samples were often collected as part of wider immuno-epidemiological studies or vaccine trials, leading to variation in study design and sampling intervals.

Table 1 Summary of gene expression datasets investigating the human immunological response to malaria infection
Full size table

Review of methodological approaches

Significant heterogeneity in the datasets was found in terms of study design, sample type, platform used and method of analysis (Tables 1, 2 and Fig. 1), making direct comparison of results between studies difficult. Most datasets were generated from whole blood samples (n = 11, 48%), however some used PBMCs (n = 3, 13%) or individual tissue or cells types (n = 8, 35%) (Table 1). For the majority of studies, expression profiling was performed by array (n = 16, 70%), with others using high throughput sequencing (n = 6, 26%) or RT-qPCR [28] (n = 1, 4%) (Table 1). There was heterogeneity in data generation between studies with variation in methods used for normalization of data and adjustment for co-variables (Table 2). Thresholds for significance varied considerably and not all studies applied corrections for multiple testing. Choice of database used for gene ontology analysis also varied and there was variable, often incomplete reporting of analysis methods used (Table 2).

Table 2 Comparison of methodological approaches for analysis of gene expression data
Full size table
Fig. 1
figure 1

Comparison of key methodological variables between datasets or publications. a Antigenic stimulation; CHMI controlled human malaria infection, ‘field’ infection naturally by mosquito bite, ‘in-vitro’ in vitro stimulation by sporozoites or infected red blood cells. Some datasets employed more than one method of antigenic stimulation. b Tissue type analysed; PBMC peripheral blood mononuclear cells. c Expression profiling method: HTS high throughput sequencing. d Manipulation of data, go gene ontology

Full size image

Transcriptional insights into the immune response to malaria infection

Seven datasets provided insight into the transcriptional changes associated with NAI to malaria (Table 3) [24,25,26, 28,29,30,31]. However, given the difficulty in defining or quantifying NAI for an individual, studies varied in their approach, choosing to examine GEPs in settings of varying history of prior exposure to malaria [25, 26, 28, 29], symptomatology during infection [25] or severity of disease [24, 32]. All studies examining NAI included small numbers of subjects and all deployed different experimental designs (Table 3).

Table 3 Gene expression studies informing understanding of naturally acquired immunity to malaria infection
Full size table

The findings from a number of studies supported a dampening of the innate pro-inflammatory immune response as a mechanism underpinning NAI [24,25,26, 33] although this finding was not observed in all studies [28, 29, 31].

One study by Franklin et al. provided evidence of ‘pro-inflammatory priming’ of the innate immune system in acute malaria infection [34]. Comparison of GEP in Brazilian adults presenting with uncomplicated malaria with paired convalescent samples showed an increase in expression in genes involved in TLR signalling pathways supporting a role for TLR hyper responsiveness in the pathology of malaria infection [34, 35].

Quin et al. sought to use RNA sequencing to elucidate the mechanism driving lower infection rates, lower parasite densities and fewer symptomatic cases of P. falciparum in the population of Fulani compared to other sympatric ethnic groups [33]. Comparison of the GEP of monocytes from infected and uninfected Fulani and Mossi adults showed a marked difference, with a significantly greater number of differentially expressed (DE) genes in infected Fulani compared to infected Mossi participants (1239 versus 3 DE genes respectively). Pathway analysis showed that infected Fulani, but not infected Mossi, individuals demonstrated a marked reduction in expression of inflammasome pathway components, suggesting a blunting of the innate pro-inflammatory immune response post-infection could explain the differences in susceptibility.

Another study sought to examine the genetic basis of gene expression variation in malaria [36]. Idaghdour et al. compared GEP in children diagnosed with uncomplicated malaria (n = 94) in Benin with age matched controls (n = 64) [36] and performed a genome wide association test of transcript abundance. Testing for genotype-by-infection interactions demonstrated the existence of genome wide significant interactions and other genes subject to interaction effects beneath genome-wide significance but still likely to have important roles in modulating the course of infection. These interactions affected the complement system, antigen processing and presentation and T cell activation [36].

In work to identify a transcriptional signature to distinguish acute malaria from other febrile illnesses, Griffiths et al. compared the GEP of twenty-two Kenyan children admitted with febrile illnesses (fifteen of which had malaria infection alone) with six convalescent samples collected 2 weeks post discharge [23]. Two main GEPs relating to neutrophil and erythroid activity were shown to differentiate acutely ill and convalescent children, with significantly higher expression of genes in the neutrophil-related gene region in subjects with bacterial infections and significantly higher expression of genes related to lymphocyte and T cell activation in subjects with malaria. The authors also identified two gene profiles whose expression intensity correlated with host parasitaemia.

Only two datasets included gene expression changes following P. vivax infection [28, 30, 37]. Rojas-Penas et al. interrogated GEP changes in malaria naïve (MN) and malaria-exposed (ME) Columbian volunteers following infection with P. vivax in a CHMI setting [28]. Significant GEP changes were consistent with time-point rather than prior malaria exposure, with a decline in innate immune signalling and neutrophil number (in contrast to strong up regulation of the same genes reported by Igadour et al. [36]) and an increase in interferon induction seen at diagnosis. No significant GEP changes were noted at other time points, including those relating to the liver stage of infection. Further analysis of this dataset by Vallejo et al. using network co-expression analysis showed that while P. vivax infection induced strong inflammatory responses in all participants, the inflammatory response was attenuated with pathways associated with antigen processing and presentation less enriched in those with prior exposure to P. vivax, suggesting a more ‘tolerogenic’ immune response in these individuals [30].

In contrast to this work, Rothen et al. found that transcriptional changes post-CHMI via intradermal injection of cryopreserved P. falciparum sporozoites were most pronounced on day 5 after inoculation, during the clinically silent liver stage rather than during the blood-stage of infection [38].

Transcriptomic studies in specific cell types

Whilst the majority of studies examined the immune response from whole blood or PBMCs, some examined transcriptomic changes in other cell types or tissues [26, 33, 39,40,41,42,43,44,45]. For example, the work of Muehlenbachs et al. with placental tissue highlighted a previously unappreciated role for B cells in chronic placental malaria [39]; whilst Sullivan et al. compared GEPs of classical and ‘atypical’ memory B cells obtained from Ugandan children showing the latter demonstrated down-regulation of B cell receptor signalling and apoptosis [43].

Discussion

GEP is a powerful tool to analyse the immune response to infection. As this review demonstrates, the application of these studies for malaria are wide-ranging, from attempts to dissect the mechanisms of NAI to improving understanding of the interaction between host genotype and infection outcome. However, as a field in its relative infancy, studies are often hypothesis generating with extremely small sample sizes. There is a lack of standardization ranging from methodological (such as sample type, RNA extraction, platform and analysis) to phenotype (including precision in disease context and immune status). This variation means interpreting published data and comparison between studies is challenging. Some of this is unavoidable, however, much could be addressed, for example by implementing standardization in blood sampling, methodological protocols for data generation and analysis with robust significance testing and approaches to confounders, use of ontologies (for example human phenotype and gene ontologies) and expert curation and annotation of data on deposition [46,47,48,49].

GEP studies are well placed to examine the mechanisms of NAI and have already helped highlight the role of the innate and early adaptive immune responses [24,25,26]. However, work has been limited by the lack of an in vitro correlate or universally accepted definition of NAI, meaning identifying the immune status of individuals or quantification of immunity is problematic [6, 50]. In field studies where the timing of infection and parasite burden and dynamics are unknown, and potentially hugely variable between individuals, only limited information can be reliably extrapolated from any GEP changes seen. Most studies assess gene expression from peripheral blood or its components, which does not provide reliable information regarding the transcriptional changes in key organs such as spleen, liver, and bone marrow. In addition, when subjects are recruited at presentation with disease, no baseline comparator data are available to use as a control. Even if a clear difference in GEP were to be reported between individuals with and without NAI, it would be near impossible to distinguish GEP changes associated with parasitaemia from those mediating immunity.

However, there is much potential for the future use of GEP studies, particularly in CHMI studies [51, 52] where the parasite burden can be pre-defined and dynamics of infection closely monitored using highly sensitive qPCR. As these studies are increasingly performed in endemic settings [53,54,55], there will be growing opportunity to use GEP to understand detailed time-course changes in immune response, particularly at the skin, liver and pre-symptomatic blood-stage, which to date have been difficult to study in human subjects infected in the field.

Conclusion

GEP in malaria is a potentially powerful tool, but to date studies have been hypothesis generating with small sample sizes and widely varying methodology. As CHMI studies are increasingly performed in endemic settings, there will be growing opportunity to use GEP to understand detailed time-course changes in host response and understand in greater detail the mechanisms of NAI.