Background
Plasmodium vivax infections cause substantial morbidity with an estimated 8.5 million infections each year and can result in severe disease [
1]. Anaemia may occur during primary or relapse infections caused by
P. vivax and may be due, in part, to bone marrow (BM) dysfunction as evidenced by post-mortem examinations of malaria patients [
2]. This dysfunction leads to dyserythropoiesis and insufficient erythropoietic output to compensate for the loss of red blood cells (RBCs) from parasitism as well as immune-mediated removal of uninfected RBCs [
3‐
6]. The underlying mechanisms and characteristics of BM dysfunction in humans remain poorly understood [
2,
7,
8] due to the difficulties and ethical restrictions in obtaining longitudinal bone marrow samples from patients. Animal models can overcome this obstacle and be used to investigate mechanisms related to the development and recovery of BM dysfunction during malaria.
Nonhuman primate (NHP) macaque models are optimal for anaemia studies in that they show haematopoietic responses and erythropoietic processes that are very similar to those of humans, and BM aspirates can be collected at multiple time points during longitudinal malaria studies [
9,
10]. Although rodent models of malaria demonstrate overarching similarities in the development of anaemia, differences in the haematopoietic responses between rodents and humans impose limitations on the extrapolation of some conclusions [
11,
12]. For example, there are notable differences in the transcriptional programmes that underlie erythropoiesis in mice compared to humans and different physiological mechanisms to deal with anaemia [
13].
The
Macaca mulatta–
Plasmodium cynomolgi model system recapitulates critical aspects of
P. vivax infection in patients, including BM dysfunction and insufficient erythropoiesis during acute infections [
10,
14]. Importantly,
P. cynomolgi also produces hypnozoites in the liver of rhesus macaques, enabling the study of relapse infections that occur with
P. vivax in patients [
9,
10,
15]. There has been extensive study of malaria-induced anaemia in
Plasmodium falciparum, in particular in the study of severe malarial anaemia. Comparatively, much less is known about the pathogenesis of anaemia during
P. vivax infection, even though it is increasingly recognized that
P. vivax also causes substantial anaemia [
16]. In fact,
P. vivax causes a greater removal of uninfected red blood cells than
P. falciparum [
4]. Furthermore, the effect of relapse infections on the bone marrow in comparison to an initial infection has not been thoroughly characterized even though relapses are speculated to potentially drive anaemia in areas of high endemicity with frequent relapses [
17].
Here, a malaria systems biology study is presented that used BM samples from a cohort of
Macaca mulatta infected for about 100 days with
P. cynomolgi M/B strain [
9] as a model of vivax malaria. The main goal was to perform integrative analyses and identify potentially dysfunctional BM mechanisms that may contribute to anaemia during acute vivax malaria and relapses. Multiple data types were generated, analysed, and integrated (i.e. transcriptome, immune profiling, and clinical data). In this paper, the analysis of the bone marrow transcriptome during the infection is presented, focusing on large-scale changes via pathway analysis and highlighting the inflammation in the marrow during acute infection. Next, the cell types that may be responsible for these changes in the marrow are explored, and the upregulation of immune pathways identified via transcriptional profiling is validated by measuring systemic cytokine levels. Finally, the consequence of these changes on erythroid progenitors is explored, with evidence suggesting the possibility that the function of GATA1/GATA2, which are known master regulators of erythroid differentiation, may be disrupted. Collectively, this is the first systems biology study using NHP models of
P. vivax infection to study BM dysfunction and anaemia.
Discussion
The development of malarial anaemia is multi-factorial, and it is clear that the loss and sustained reduction of RBCs during an infection is not due to parasitism alone. Bone marrow suppression and the removal of uninfected RBCs contribute to the development of anaemia. In the rhesus macaque—
P. cynomolgi model, both processes are active, based on haemoglobin and reticulocyte kinetics described here and in Joyner et al. [
9]. Interestingly, insufficient erythropoiesis appears to contribute to the development of anaemia early during a blood-stage infection since appropriate compensation for anaemia is observed after the peak of parasitaemia or after the administration of blood-stage treatment (Fig.
1). Furthermore, there were few changes in the BM transcriptome after the peak of parasitaemia, suggesting BM is no longer dysregulated (Fig.
2). Thus, it is clear that there are multiple phases in the development of malarial anaemia in macaques infected with
P. cynomolgi. In each phase, there are different pathological processes that may predominate, and future studies should aim to explore the molecular mechanisms underlying each phase, particularly the removal of uninfected red blood cells.
Predictions from field studies have suggested that relapses are responsible for most
P. vivax blood-stage infections and potentially most clinical illness [
38‐
40]. Therefore, the initial hypothesis of this study was that both
P. cynomolgi acute primary infections and relapses would cause alterations in the BM transcriptomes since anaemia would be an expected complication in both cases. However, unlike the acute infections, relapses did not result in significant changes in the BM transcriptomes. Although unexpected, this finding is in agreement with the conclusions in Joyner et al., which demonstrated that
P. cynomolgi relapses are not associated with the development of anaemia, even when peripheral parasitaemia is detectable [
9].
The lack of anaemia and changes in the BM transcriptome during
P. cynomolgi relapses is likely due to the significant decrease in parasite burden during relapses in comparison to the initial infection (Fig.
1). This decrease is likely due to the development of immunity that prevents the parasite from reaching levels similar to those during an initial exposure. Lower levels of parasite replication may allow the BM to maintain its normal compensatory functions during relapses, unlike acute infection where insufficient erythropoietic output is linked to higher parasite densities [
9]. Indeed, there is evidence from malaria chemotherapy studies that relapses caused by
P. vivax also do not necessarily result in disease, and thus, this phenomenon does not seem to simply be attributable to nonhuman primates [
41‐
43]. Taken together, this evidence suggests that there is a threshold of parasite burden that disrupts bone marrow function and erythropoietic output prior to the development of effective anti-parasite immunity.
During acute infection, many biological pathways and processes in the BM were upregulated, including pathways related to cellular metabolism and transcription. Such pathways are likely representative of the ongoing host response against the parasite both in the periphery and potentially in the BM since both
P. vivax and
P. falciparum iRBCs can readily be found in the BM [
44,
45]. Many of the upregulated pathways in the BM were related to inflammation and largely composed of cytokine signalling pathways. This agrees with previously published work demonstrating that cytokines, including IL-10 and IL-27, are important in BM responses during rodent malaria and/or in malaria patients [
32,
46]. The current data suggest that these cytokines may also be involved in the BM of NHPs with malaria.
Here, Type I and Type II IFN signalling pathways were identified as potentially key cytokines involved in malarial anaemia. However, despite an enrichment in transcriptional pathways involved in Type I IFN signalling in the BM, IFNα protein levels in the plasma were not increased (Fig.
3h). (
ifnα gene expression in the BM was not reliably quantified.) In stark contrast, the Type II IFN signature was accompanied by an increase in
ifnγ gene expression in the BM and IFNγ concentration in the plasma, and the erythroid progenitor population was negatively correlated with IFNγ (Figs.
2i,
3g,
4). These data suggest that Type II IFNs could negatively impact the erythroid progenitor population through direct or indirect mechanisms during acute infection. Indeed, IFNγ can directly cause apoptosis of erythroid progenitors in vitro and it has been previously implicated in malarial anaemia in rodent malaria models [
12,
47,
48].
Type I IFNs can work in concert with IFNγ to suppress erythropoiesis [
49‐
51], and therefore it is possible that Type I IFNs may initiate the disruption of erythropoiesis earlier during infection. The reasons why Type I IFN transcripts were too low to be quantified could be because they are more tightly regulated than IFNγ and the sampling regimen missed the point at which these proteins are synthesized and released into the plasma. Alternatively, these proteins may act predominantly at local levels and may have only been present in the BM. There also may be other intermediary molecules that are unmeasured here but may have a role in the observed phenomena. Regardless, recent studies demonstrate Type I IFNs to be important during early blood-stage malaria, and the results presented here suggest these cytokines should be explored further in relation to malarial anaemia [
52‐
54].
This study’s results also suggested the possibility that monocytes were largely responsible for changes in the BM transcriptome during acute infection (Figs.
3,
4). It has previously been shown that macrophages and monocytes in the BM produce inflammatory cytokines in response to parasite byproducts such as haemozoin. It has been hypothesized that this process drives the dyserythropoiesis observed during malarial anaemia [
55]. Although most of these conclusions were drawn from in vitro experiments, the analysis here supports a model where monocytes in the BM negatively influence the erythroid lineage in vivo via cytokine production during acute malaria [
55]. It is interesting to speculate that these signatures may be coming from BM-resident monocytes and/or macrophages, but more work will be required to explore this possibility in future studies.
Although intermediate and non-classical monocytes are thought to be important for parasite control in the periphery [
56], this analysis from the BM implicated these monocytes as potentially being responsible for the decrease in the erythroid progenitors. WGCNA analysis associated these monocyte subsets with an increase of pro-inflammatory cytokines such as IFNγ, MIP1-α/β and TNFα in the plasma, which would not necessarily be expected of these monocytes, as these subsets are not considered to be classically pro-inflammatory. Indeed, many of these cytokines are known to negatively impact erythroid progenitors, directly or indirectly [
12,
55]. It is interesting to speculate that these monocytes may play a dual role in controlling parasite growth through cytokine production, etc. in the peripheral blood but also contribute to the development of anaemia through these same processes. Future work could assess this directly using this NHP model of
P. vivax infection.
Previous evidence has suggested that the disruption of erythropoiesis may be due to dysregulation of transcription factors that control erythropoiesis [
57]. In this study, it was shown that GATA1 and GATA2, two master regulators of erythropoiesis, may not function appropriately during acute malaria. Genes regulated by GATA1 and GATA2 were downregulated during acute infection, but upregulated whenever appropriate erythropoietic output was restored. Indeed, some of the cytokines (e.g. TNFα, IFNγ) that were upregulated are known to antagonize GATA1 and, thus, disrupt terminal erythroid differentiation [
58,
59], providing a potential mechanism for what was observed. Although both GATA1 and GATA2 were identified, GATA1 may be more central in the process, based on
gata2 gene expression not being upregulated when erythropoietic output was restored (Fig.
5g). Future studies should examine the factors that influence the function of GATA1/GATA2 during malaria since it is clear that a variety of intermediate molecules besides those described here could also affect each protein’s function [
12,
60,
61].
Although this study provides the basis for future investigations related to malarial anaemia using NHPs, there are limitations. First, the number of NHPs presented in this study is smaller than initially designed because one macaque from the cohort developed severe disease during the acute infection period and required euthanasia [
9]. This decrease in the cohort size may limit the generalizability of the results, but other studies using in vitro systems, rodent malaria models, and samples from malaria patients have arrived at conclusions similar to those in this study [
12,
55]. This suggests that studies with NHPs, which in comparison to studies using patient samples or rodent models will be restricted to small sample size, have the potential to produce generalizable results.
A second limitation of this study is that the transcriptomic profiling of the BM by RNA-Seq was performed on BM MNCs. This sample type is composed of multiple cellular lineages, and thus it is difficult to pinpoint the specific cell type in the BM that was responsible for changes in the transcriptome. The changes identified here are thus necessarily correlative rather than describing definitive and direct mechanisms in specific cell types. Although such uncertainty is in some ways a weakness, this approach also has its advantages. Most importantly, it enabled an unbiased survey of the major changes in the BM during acute malaria and relapse infections in macaques. In contrast, an initial focus on a singular cell type would likely have missed some critical transcriptional changes in the BM, and thus the insights derived from those changes. Future studies can target specific cellular lineages to better understand the role of each individual cell type in the BM during infection.
Authors’ contributions
MRG, TJL, and MS conceived and designed the experiments; CJJ, MC, SL, and SS conducted the experiments; YT and CJJ performed data analysis and prepared figures; CJJ and YT wrote the paper; MS, TJL, and MRG provided experimental oversight, analytical guidance, and edited the paper; SBP, MVN, JDB, and JCK managed the data and deposited it into republic repositories. All authors read and approved the final manuscript.