A population of CD4hiCD38hi T cells correlates with disease severity in patients with acute malaria

Abstract Objective CD4+ T cells are critical mediators of immunity to Plasmodium spp. infection, but their characteristics during malarial episodes and immunopathology in naturally infected adults are poorly defined. Flow cytometric analysis of PBMCs from patients with either P. falciparum or P. knowlesi malaria revealed a pronounced population of CD4+ T cells co‐expressing very high levels of CD4 and CD38 we have termed CD4hiCD38hi T cells. We set out to gain insight into the function of these novel cells. Methods CD4+ T cells from 18 patients with P. falciparum or P. knowlesi malaria were assessed by flow cytometry and sorted into populations of CD4hiCD38hi or CD4norm T cells. Gene expression in the sorted populations was assessed by qPCR and NanoString. Results CD4hiCD38hi T cells expressed high levels of CD4 mRNA and canonical type 1 regulatory T‐cell (TR1) genes including IL10, IFNG, LAG3 and HAVCR2 (TIM3), and other genes with relevance to cell migration and immunomodulation. These cells increased in proportion to malaria disease severity and were absent after parasite clearance with antimalarials. Conclusion In naturally infected adults with acute malaria, a prominent population of type 1 regulatory T cells arises that can be defined by high co‐expression of CD4 and CD38 (CD4hiCD38hi) and that correlates with disease severity in patients with falciparum malaria. This study provides fundamental insights into T‐cell biology, including the first evidence that CD4 expression is modulated at the mRNA level. These findings have important implications for understanding the balance between immunity and immunopathology during malaria.


INTRODUCTION
Malaria remains a significant threat to global health with 228 million cases and 405 000 deaths estimated in 2018 1 . The causative agent, the Plasmodium spp. parasite, has a complex multistage life cycle that has co-evolved with the human host. The human immune response to the parasite is complex and engages innate and adaptive cellular and humoral immune components, and the parasite has developed multiple strategies to avoid the host immune response in order to persist. Managing the balance between controlling a rapidly multiplying and potentially fatal pathogen and avoiding immunopathology is difficult (reviewed in Coban et al. 2,3 ). Because of the complexity of this parasite-host relationship, immune correlates of protection remain unknown despite many decades of intensive research, and the development of an effective vaccine has proved elusive.
The blood stages of the Plasmodium spp. parasite life cycle are responsible for the clinical symptoms of malaria. Animal studies have demonstrated a critical role for CD4 + T cells in the control of the blood stage of Plasmodium infection (reviewed in Kurup et al. 4 ), and immunisation studies utilising controlled human malaria infection (CHMI) of malaria-na€ ıve human volunteers have associated protection from reinfection with CD4 + T-cell responses 5,6 . Furthermore, mounting evidence in humans and animal models suggests that CD4 + T cells perform a critical cytokine-mediated balancing act in order to mediate immunity without pathology, in particular, via production of pro-inflammatory cytokines including IFN-c and TNF-a, and the immunosuppressive cytokine IL-10 (reviewed in Kumar et al. 7 ).
In CHMI studies, we have shown that the control of the parasite in malaria-na€ ıve volunteers following the infection with P. falciparuminfected red blood cells (pRBCs) is associated with the expansion of a population of CD4 + T cells coexpressing the activation marker CD38 8 . However, in naturally exposed individuals from Uganda, an increased frequency of CD38-expressing CD4 + T cells predicts treatment failure 9 , and in another study, CD38 + CD4 + T-cell frequency correlated positively with HIV viral load, which was further accentuated in patients co-infected with malaria 10 .
Although CD38 expression has primarily been used as a marker of CD4 + T-cell activation, the CD38 molecule itself performs a multitude of functions and is ubiquitously expressed in different tissues. In broad terms, CD38 has three known functions in T cells: (1) it is a multifunctional ectoenzyme with activities that include catalysis of extracellular nicotinamide adenine dinucleotide (NAD), thereby regulating extracellular levels of NAD and modulating intracellular Ca 2+ levels (reviewed in Malavasi et al. 11 ); (2) it acts as a cellular receptor, associating with the T-cell receptor (TCR) on lipid rafts on T cells, and transduces activation signals by phosphorylating tyrosine residues on CD3, ZAP-70 and LAT 11 ; and (3) it acts as a ligand, capable of binding CD31 on endothelial cells and participating in lymphocyte homing and inducing signalling through CD31 on CD31-expressing cells 11 . These functions have particular relevance in the case of Plasmodium infection: infected RBCs have very high levels of NAD 12 , and lysis of infected RBC releases NAD into the extracellular milieu, which acts as an immune 'danger signal' promoting inflammation and activating immune cells including granulocytes 13 . CD38 binding CD31 on macrophages negatively regulates TLR4 signalling in those cells, which is particularly relevant as the major Plasmodium pathogenassociated molecular pattern molecules (PAMPs) of glycosylphosphatidylinositol (GPI) and haemozoin (bound to fibrinogen) are known to signal via TLR4 and induce the release of proinflammatory cytokines 14,15 . Taken together, these observations suggest that CD38 + -expressing CD4 + T cells may play a critical role in immunomodulation during Plasmodium infection rather than merely being a marker of 'activated' cells.
Herein, to better understand the role of CD38 + CD4 + T cells during malaria and in particular during acute malaria, we assessed the genotypic and phenotypic characteristics of the CD38 + CD4 + T-cell population present in the peripheral blood of adults presenting to health facilities in Indonesia or Malaysia with acute P. falciparum or P. knowlesi malaria. Unexpectedly, in most patients, we observed by flow cytometry a prominent population of CD4 + T cells coexpressing high levels of CD4 and CD38 (CD4 hi CD38 hi ). Assessment of mRNA expression in FACS-sorted CD4 hi CD38 hi T cells confirmed increased CD4 gene expression relative to non-CD4 hi CD38 hi T cells and a prominent signature of genes associated with regulatory T-cell function including LAG3, HAVCR2 (TIM3), IL10, together with other genes that suggest CD4 hi CD38 hi cells serve a specialised immunoinhibitory role during acute malaria.

CD4 hi CD38 hi T cells arise during acute malaria
We previously identified a significant positive correlation between the frequency of CD4 + T cells expressing normal levels of CD4 and increased levels of CD38 (CD4 + CD38 hi cells) and control of parasite growth during the first seven days of primary Plasmodium blood-stage infection in malaria-na€ ıve volunteers, using a controlled human malaria infection (CHMI) model 8 . However, the function and characteristics of CD38expressing CD4 + T cells in individuals naturally exposed to malaria, and during an acute malarial infection, are unknown. To determine whether a population of CD4 + CD38 hi cells is present in individuals undergoing secondary exposure to Plasmodium, similar to individuals undergoing primary exposure 8 , we performed flow cytometric assessment of CD38 expression on CD4 + T cells from PBMCs collected from Malaysian and Indonesian adults who presented to hospitals with acute uncomplicated P. falciparum malaria or P. knowlesi malaria (before the commencement of drug treatment: acute, day 0) (subject details in Table 1). In most cases of acute falciparum malaria or knowlesi malaria, we observed a distinctive and unexpected population of CD4 + T cells that co-expressed high levels of CD38 and CD4, which we have termed CD4 hi CD38 hi cells (Figure 1a). In paired patient samples taken in convalescence, 28 days following the successful drug treatment, the CD4 hi CD38 hi population was no longer present ( Figure 1a). These CD4 hi CD38 hi T cells appear to be an activated subset of conventional peripheral TCR-ab CD4 + T cells, since they express CD3 and TCR-ab, but do not express markers of NK cells (CD56, NKp46), NKT cells (TCR-Va24) and cd T cells (TCR-cd) and have lost expression of CD45RA (Figure 1b).
To investigate the functional relevance of these cells, we evaluated a second cohort of P. falciparum-infected Indonesian adults, to correlate the frequency of CD4 hi CD38 hi cells with disease severity (subject details in Table 2). The frequency of CD4 hi CD38 hi T cells increased with disease severity (Figure 1c). While CD4 hi CD38 hi T-cell frequency was not significantly increased in asymptomatic (parasitaemic) individuals (AS) compared with local healthy (parasite-free) controls (HC), it was significantly increased in patients with uncomplicated malaria (UM; P = 0.0381), and further increased in patients with severe malaria (SM; P = 0.0117) (Figure 1c). Although in parasitaemic patients CD4 hi CD38 hi T-cell frequency increased with increasing parasitaemia or plasma HRP2, a marker for parasite biomass (both r = 0.54; P = 0.0006), there was no relationship between CD4 hi CD38 hi T-cell CD4 expression in CD4 hi CD38 hi cells is modulated at the mRNA level Upregulation of the CD4 co-receptor on CD4 + T cells has not been described previously in human malaria or in experimental mouse malaria. To confirm that the apparent CD4 high population was not an aberration of the flow cytometric analysis, we FACS-sorted populations of CD4 hi CD38 hi cells and CD4 + T cells expressing normal levels of CD4 and intermediate/low levels of CD38 (CD4 norm ) from PBMCs collected from adult patients with acute uncomplicated P. falciparum malaria (gating methodology as above, shown in Figure 1b), and then stimulated the cells for 3 h with PMA/ ionomycin and assessed mRNA expression in the cells by real-time qPCR (Figure 2a). Expression of CD4 mRNA was significantly increased in the CD4 hi CD38 hi population when compared to the CD4 norm cells in the acutely infected samples, but not in the convalescent samples ( Figure 2b). This suggests that the increased CD4 expression by CD4 high T cells is programmed into the cells at the RNA level.
Investigating the relative expression of several other genes, we observed that IFN-c mRNA was expressed equally by CD4 hi CD38 hi and CD4 norm T cells at the acute stage, but at the convalescent stage the CD4 hi CD38 hi T cells produced significantly less IFNG mRNA than the CD4 norm cells. This pattern of reduced expression of IFNG mRNA in the convalescent samples mirrors our previously published observations for CD4 + CD38 high cells from healthy volunteers (malaria-na€ ıve Australian volunteers) and for CD4 + CD38 high cells from CHMI volunteers at day 7 of infection 8 . Unlike IFNG, CD4 hi CD38 hi cells produced significantly more IL10 mRNA at the acute stage, but this was lost at the convalescent stage ( Figure 2b). Both CD4 hi CD38 hi and CD4 norm cells expressed low levels of FOXP3 (Figure 2b). This pattern of co-expression of IFN-c and IL-10 together with an absence of FOXP3 expression has been observed in CD4 + T cells from children living in malaria-endemic regions who have suffered recent malarial episodes [16][17][18] and is associated with a type 1 regulatory T-cell phenotype (TR1) (reviewed in Roncarolo et al. 19,20 ). Those studies reported CD4 + (CD4 norm ) T cells only and did not identify CD4 hi cell populations.
Given the increased frequency of CD4 hi CD38 hi T cells in patients with uncomplicated malaria and severe malaria, compared with asymptomatic but parasitaemic individuals and healthy uninfected controls, we next assessed whether CD4 hi CD38 hi T cells recognise Plasmodium antigens. PBMCs collected from adult patients with acute uncomplicated P. falciparum malaria were incubated for 24 h with an extract made from P. falciparum-parasitised RBCs (pRBCs) or a control extract made from non-infected RBCs (nRBCs). Immediately following this in vitro stimulation, the CD4 hi CD38 hi and CD4 norm cells were sorted by flow cytometry (as above) and placed into RLT lysis buffer for mRNA processing without further stimulation. CD4 hi CD38 hi cells exposed to pRBC responded with an increase in IFNG mRNA that was not observed for CD4 norm cells or when the control extract was used ( Figure 2c). IL-10 levels were elevated in CD4 hi CD38 hi cells from P. falciparum patients regardless of antigen exposure.

CD4 hi CD38 hi cells express specialised regulatory T-cell genes
In order to gain a more in-depth insight into the function and phenotype of CD4 hi CD38 hi T cells, we next utilised the NanoString TM technology platform to assess the differential mRNA expression of 130 genes associated with T-cell function in CD4 hi CD38 hi and CD4 norm cells in acute malaria. CD4 hi CD38 hi and CD4 norm cells were Gene expression assessment by qPCR of FACS-sorted CD4 norm (norm) and CD4 hi CD38 hi (high) T cells from patients with acute P. falciparum malaria (n = 6, log 2 -normalised data, paired t-test). (c) P. falciparum antigen-specific responses measured in FACS-sorted CD4 norm (blue squares) and CD4 hi CD38 hi (red circles) T cells from patients with acute P. falciparum malaria (Pf inf') or healthy Sabah controls (HC) by qPCR after 24-h co-incubation with extract made with P. falciparum-infected RBC (pRBC) or normal RBC (nRBC) (n = 3 Pf inf' and n = 3 HC, log 2 -normalised data, paired t-test).
sorted from PBMCs collected from acutely infected individuals (Table 1; methodology as above in Figure 2a) with P. falciparum infection (6 patients from Papua and 6 from Sabah) or P. knowlesi infection (6 patients from Sabah). Of the 130 genes, 78 were reproducibly detected across the groups and were included in further analysis (Supplementary table 1). We first assessed differences between the P. falciparum (n = 12) and P. knowlesi (n = 6) samples with a principal component analysis (PCA) using the Genomics Data Miner (GMine) 21 , and also by comparing the fold change in gene expression between the CD4 hi CD38 hi and CD4 norm cells by ANOVA ( Figure 3a). Since our analysis revealed no significant differences between the P. falciparum and P. knowlesi data sets (FDR < 0.05), we combined those data and performed a PCA comparing differential gene expression between CD4 hi CD38 hi and CD4 norm cells, as well as a paired t-test of CD4 hi CD38 hi and CD4 norm cells from each subject (n = 18) ( Figure 3b). In total, 49 genes were significantly differentially expressed between the CD4 hi CD38 hi and CD4 norm cells (FDR < 0.05, Supplementary table 1 and Figure 3b and c).
Next, we performed a linear discriminant effect size analysis (LEfSe) 22 using GMine to identify features most likely to explain the differences between the CD4 hi CD38 hi and CD4 norm populations (using all 78 detected genes) (Figure 3d); and a Random Forest analysis to identify the most relevant features of the cohort (Figure 3d) (GMine). Both analyses identified expression of a distinct set of genes associated with T reg function in the CD4 hi CD38 hi cells, including genes strongly associated with TR1 cells 19 : IL10, LAG3, HAVCR2 (TIM3), PDCD1 (PD1), IL21, ICOS, MAF and PRF1. Furthermore, another canonical TR1 cytokine, TGFB1, was very highly expressed in both the CD4 hi CD38 hi and the CD4 norm subsets (but not differentially expressed) (Figure 3c, lower panel).

Insight into the ontogeny of CD4 hi CD38 hi cells
We utilised the Ingenuity Pathway Analysis platform and the data generated in the NanoString study (above) to predict upstream regulators of the CD4 hi CD38 hi cells. Using this method, the cytokine IL-27 was identified as a potential inducer of these cells (Figure 4a and b). To test whether IL-27 could indeed induce the CD4 hi CD38 hi cell population in vitro, we cultured PBMCs from healthy malaria-na€ ıve (Brisbane) volunteers with P. falciparum-parasitised RBC extract for 14 days +/À recombinant human IL-27. Cells cultured in the presence of IL-27 displayed significantly increased co-expression of both CD4 and CD38 by flow cytometry (Figure 4c and d), and sorted bulk CD4 + T-cell populations from these cultures displayed enhanced CD4, IFNG, IL10 and LAG3 mRNA expression (Figure 4e).

CD4 hi CD38 hi cells in mice
We were curious as to why no analogue of the CD4 hi CD38 hi cell population has been yet reported in mice, given the very large number of immunological studies performed utilising rodent malaria models including those focused on the role of CD4 + T cells. Indeed, in our laboratory, despite many thousands of flow cytometry analyses conducted using a flow cytometric method for monitoring Plasmodium infection 23 , we have never observed CD4 high cells in the blood of mice during infection. To provide a foundation for dissecting the role of CD4 hi CD38 hi cells in host-pathogen immunity, we revisited our P. yoelii murine malaria model with a different approach. To identify CD4 + T cells specifically responding to infection, we utilised a mouse model where cells are engineered to express the fluorescent protein YFP upon the production of IFN-c 24 . Using this model, mice were infected with non-lethal P. yoelii 17XNL-parasitised RBC via intravenous injection to establish a blood-stage infection. On days 1, 3, 5, 8 and 11 of infection, cells were recovered from several organs including the spleen, blood, bone marrow, liver-draining lymph nodes (DLNs) and inguinal lymph nodes (Ing LNs). Using this more-focused approach, IFNc-producing CD4 + T cells were evident in the spleen at day 3 of infection ( Figure 5a), but by day 8, the YFP + CD4 + T cells were diminished, and by day 11, the remaining YFP + CD4 + cells in the spleen were also CD4 high (Figure 5a and b). This pattern of IFN-c expression coincides with a peak in plasma IFN-c routinely observed in mouse models [25][26][27] , and in our own experiment in C57BL/ 6 mice (Figure 5c) in which plasma IFN-c peaked around day 3 and IL-10 around day 11. This timing coincides with the onset of acute parasitaemia, which can reach very high levels in the P. yoelii mouse model (Figure 5d). These CD4 high cells were apparent only in the spleen and  bone marrow, and not in blood, liver DLN or inguinal LN (Figure 5e). The CD4 high cells from the spleen at day 11 were also CD38 high , but the levels of CD38 were not high enough to make this population as prominent as observed in the human studies (Figure 5f). Finally, to assess gene expression in the CD4 high murine cells, CD4 high and CD4 norm cells were FACS-sorted from spleens collected at day 11 of infection and processed directly for analysis by real-time qPCR without further stimulation. As in human CD4 hi CD38 hi cells, murine CD4 hi cells had increased CD4 mRNA and contained within this population were the IL10-producing cells with increased LAG3 and an absence of FOXP3 (Figure 5g). Thus, the population identified in the P. yoelii murine model shares core similarities with the CD4 hi CD38 hi population in humans.

DISCUSSION
We have identified a population of CD4 + T cells in the blood of acute malaria patients infected with P. falciparum or P. knowlesi that co-express high levels of the CD4 co-receptor and the ectoenzyme CD38. These CD4 hi CD38 hi T cells appear to be an activated subset of conventional peripheral TCR-ab CD4 + T cells, since they express CD3 and TCR-ab but do not express markers of NK cells (CD56, NKp46), NKT cells (TCR-Va24) or cd T cells (TCR-cd) and have lost expression of CD45RA. Because of reductions in absolute CD4 T-cell numbers in the blood during acute malaria, NK cells are relatively more abundant 28 ; hence, we excluded these based on the expression of CD56. These CD4 hi CD38 hi cells also expressed a set of genes strongly associated with immunomodulatory function, including IL10, LAG3 and TIM3 (HAVCR2), and were present in the peripheral blood of patients in malariaendemic areas during acute symptomatic infection but were absent following the drug treatment and convalescence.
CD4 + T cells play a critical role in the control of the blood stage of Plasmodium infection, both in humans and in animal models (reviewed in Kurup et al. 4,29 ); however, a clear role also exists for CD4 + T cells in suppressing immunopathology. The appearance of IL-10-expressing CD4 + T cells during Plasmodium infection in humans [16][17][18]30,31 and mice 32,33 has been well documented, and early reports associated cerebral malaria in children with high circulating IL-10 levels 34-36 : one study found IFN-c/IL-10-co-producing cells in children with severe malaria 17 .
CD4 hi CD38 hi cells appear to provide a more comprehensive set of markers for TR1 regulatory T cells as they strongly express canonical TR1 genes including IL10, IFNG, TGFB1, LAG3, TIM3, IL21, ICOS, PD1 (PDCD1), MAF, and PRF1 and do not express FOXP3 (reviewed in Roncarolo et al. 19,20 ). CD4 hi CD38 hi cells strongly express CXCL8 (IL-8), which is a pro-inflammatory cytokine not previously associated with TR1 cells but shown to be produced by human FOXP3 + T regs 37 and upregulated by CD4 + T cells very early during primary Plasmodium infection 38 .
Our cell population is also distinct from the CD38expressing CD4 + T-cell subset identified previously as a hypoproliferative IL-13-secreting CD4 + T-cell subset that retained expression of na€ ıve cell surface markers following the TCR stimulation and secreted IL-13, and IFN-c and TNF 39 . Note that the CD4 hi CD38 hi population described in our study may or may not be a homogeneous population.
Our pathway analysis ( Figure 4) suggested IL-27 as an upstream regulator of the CD4 hi CD38 hi phenotype. In vitro studies with mice have identified IL-27 as an inducer of IL-10 production in CD4 + T cells 40,41 and confirmed a role for IL-27 receptor signalling 41 , and similar observations were made with in vitro cultures of human cells 42 . In the context of experimental rodent malaria, IL-27 (acting via the IL-27 receptor) was shown to induce the generation of IL-10-expressing CD4 + T cells that were critical to limit immunopathology 32,43 . It remains possible that other cytokines may also induce the CD4 hi CD38 hi phenotype; indeed, our pathway analysis ( Figure 4) also identified IL-12 and IL-21 as potential upstream regulators and others have shown that IL-6 and TGF-b could induce IL-10 production by CD4 + T cells in mice 41 . The mechanisms of CD4 hi CD38 hi cell generation in vivo remain to be elucidated, but we think it likely that IL-27 levels in vivo are likely to be higher around areas of antigen presentation and more likely to affect cells responding to antigen. This may explain the fairly uniform upregulation of CD4 and CD38 we observed in our in vitro cultures. Possibly adding support to the notion that IL-27 is driving the CD4 hi CD38 hi phenotype in our patients with acute malaria is a recent paper by Oterdal et al. who found that IL-27 levels are elevated in the plasma of adults living in malariaendemic areas with acute falciparum malaria 44  A striking feature of the CD4 hi CD38 hi cell is its very high expression of the CD4 co-receptor. The canonical role of the CD4 co-receptor is to bind a region on the MHC-II molecule and restrict signalling through the CD4 + T-cell receptor to antigens bound only on MHC-II molecules 45 . However, the CD4 co-receptor serves multiple other roles including acting as a receptor for the chemotactic cytokine IL-16 46 , and facilitating adhesion to cells expressing MHC-II independently of TCR engagement (reviewed in Glatzova et al. 47 ). The latter is particularly relevant in malaria as MHC-II expression is dramatically increased on murine brain microvascular endothelium during experimental cerebral malaria 48 , and on human endothelium in response to IFN-c 49 .  Another vital role of the CD4 co-receptor is to enhance signalling through the T-cell receptor during engagement with the peptide-MHC-II complex 50 . An analogous effect has been observed in CD8 + T cells, by us 51 and others 52 , where the sensitivity of the T-cell receptor was shown to be modulated by increasing or decreasing the expression levels of the CD8 coreceptor. This can increase the range of specificity of the T cell, thereby allowing a T cell that is typically restricted to a very specific peptide to be able to respond in a cognate fashion to a broader range of targets. In the study reported herein, CD4 hi CD38 hi cells responded to parasitised RBC by strongly expressing IFN-c and IL-10, but did not respond to uninfected RBC controls, suggesting that TCR signalling is functioning in these cells and that they are not exhausted 53 . However, we cannot rule out the possibility that some cells responding to antigen may have started out as CD4 norm prior to the in vitro assay.
Downregulation of the CD4 co-receptor has been reported in one study in mice 54 , and several other studies have demonstrated increased CD4 expression on CD4 + T cells following the very strong or chronic activation, such as in a mixed lymphocyte reaction 55 or an alloreactive transplant setting 56,57 . However, increased CD4 expression in the context of natural hostpathogen interactions, as shown herein for our novel CD4 hi CD38 hi cell population, has not been previously reported. Importantly, we show that this increase in CD4 expression is modulated at the mRNA level, providing the first evidence of this fundamental insight into T-cell biology. This is significant because CD4 levels could be transiently modulated by cycling of the co-receptor between the cell surface and cytoplasm 58 ; however, increased CD4 gene expression together with the required energy expenditure emphasises the level of commitment to the cell phenotype, suggesting that increased CD4 expression is important to the function of these cells.
The unusual characteristics of CD4 hi CD38 hi cells are well suited to a specialised T regulatory cell arising in response to Plasmodium infection. Firstly, in residents of two distinct malariaendemic areas (Sabah and Papua), CD4 hi CD38 hi cells were present in relatively large numbers during acute malaria but were absent in convalescence following the successful drug clearance of parasites, and CD4 hi CD38 hi cells could recognise Plasmodium-infected RBC in vitro, suggesting that CD4 hi CD38 hi cells are responding directly to Plasmodium infection. Secondly, CD4 hi CD38 hi cells express a unique set of genes with particular relevance to Plasmodium infection, including those mentioned above and CXCR6 whose ligand (CXCL16) is constitutively expressed in cerebrospinal fluid 59 , on endothelial cells in the brain 60 and in the placenta during malaria 61 , and may induce migration and localisation of CD4 hi CD38 hi to these tissues. CD4 would promote localisation to areas of increased MHC-II expression during infection, and CD38 is a ligand for CD31 (platelet endothelial cell adhesion molecule, PECAM-1) whose expression on endothelial cells is upregulated during malaria 62 , which may in turn promote CD4 hi CD38 hi cell localisation. Consistent with this, it was notable that the frequency of CD4 hi CD38 hi cells in severe malaria was strongly associated with measures of endothelial activation. CD38 is also a critical ectoenzyme catalysing extracellular nicotinamide adenine dinucleotide (NAD), thereby regulating extracellular levels of NAD released by lysis of infected RBC 12 , reducing levels of proinflammatory NAD 13 and providing the substrate ADPR that is further catalysed to generate the strongly immune-inhibiting molecule adenosine 63 . Another role for CD38 is to bind CD31 on macrophages and inhibit TLR4 signalling, which is relevant since the major Plasmodium PAMPs of glycosylphosphatidylinositol (GPI) and haemozoin (bound to fibrinogen) are known to signal via TLR4 and induce the release of pro-inflammatory cytokines 14,15 .
Taken together, our data suggest that CD4 hi CD38 hi cells are a specialised TR1 cell that arise during acute infection. We propose a role in limiting immunopathology. In our study, CD4 hi CD38 hi cells were increased in proportion to malaria disease severity (Figure 1c), and there are phenotypic analogies between the CD4 hi CD38 hi T cells identified herein and IL-10-expressing CD4 + T cells that were shown to limit immunopathology in a rodent malaria model 32 . Whether these cells also inhibit parasite clearance is not known. Studies in rodents have shown that blocking T regulatory cell function can enhance parasite clearance 33,64 , and Jagannathan et al. found that the CD4 + T cells co-expressing IFN-c and IL-10 may associate with an increased risk of future malaria in children 17 . In a previous study, we found that control of the parasite following the infection of malaria-na€ ıve volunteers with P. falciparum-infected RBC was associated with an increased frequency of CD4 + T cells co-expressing CD38, but these did not have increased expression of CD4, were poor producers of IFN-c 8 and appeared to be analogous to mouse CD4 T cells activated in vitro 65 and markedly different from those we describe herein during clinical disease. Whether the CD38-expressing CD4 + T cells associated with treatment failure in clinical malaria in Uganda had increased expression of CD4 was not reported. While there was no relationship between CD4 hi CD38 hi cells and parasite biomass evident in the current study, parasite clearance could not be assessed, and the role of this cell population in parasite clearance requires prospective study.
In conclusion, we describe herein a population of CD4 + T cells expressing high levels of CD4 and CD38 (CD4 hi CD38 hi cells) in adult patients acutely infected with either P. falciparum or P. knowlesi. Assessment of differential mRNA expression of 130 genes associated with T-cell function revealed that these CD4 hi CD38 hi cells had a TR1 phenotype, with high expression of LAG3, IL10, TIM3 and numerous other genes associated with modulatory activity. We further provide the first naturally occurring example of increased CD4 expression on CD4 + T cells, and the first evidence that changes in CD4 co-receptor expression levels are modulated at the mRNA level. These findings give new insight into T-cell biology and provide a starting point for future studies that aim to enhance Plasmodium parasite clearance while avoiding immunopathology.

Study participants
In Sabah, an area of low malaria transmission, patients were enrolled from a tertiary referral centre (Queen Elizabeth Hospital 1, Kota Kinabalu) 66,67 and two district hospitals (Kudat and Kota Marudu) as part of concurrent prospective clinical studies 68 . During the sample collection period, the region was co-endemic for P falciparum, P. vivax and the zoonotic parasite P. knowlesi 69,66 . Criteria for enrolment included a blood film that was positive by microscopy for any Plasmodium species, fever or history of fever in the last 48 h, no major concurrent illness or comorbidity, and no prior antimalarial therapy in the preceding 24 h. Patients were excluded if pregnant or lactating. Hospitalisation was mandatory in all patients with malaria in Sabah. All patients with malaria were treated according to the local guidelines. Uninfected participants were visitors or relatives of patients with malaria, with no fever or history of fever in the preceding 14 days and with a blood film negative for malaria parasites. Heparinised venous blood samples were collected at presentation, and follow-up samples were collected 28 days post-treatment. Peripheral blood mononuclear cells (PBMCs) were isolated and cryopreserved until analysis.
In a lowland region of Papua with perennial, unstable transmission of P. falciparum, P. vivax and P. malariae, patients were enrolled with malaria attending the Mitra Masyarakat Hospital in Timika or a regional health clinic. The following groups of participants were enrolled following the informed consent, and patients with acute uncomplicated falciparum malaria (UM) presenting with P. falciparum monoinfection by microscopy, fever or history of fever in the preceding 48 h and no alternative cause were identified (n = 20) [70][71][72] : patients with P. falciparum parasitaemia and ≥ 1 modified WHO criteria of severe malaria (SM), acute renal failure (creatinine > 265 µmol L À1 ), or hyperbilirubinaemia with renal impairment (creatinine > 130 µmol L À1 ) and/or parasitaemia of > 100 000 parasites lL À1 , or blackwater fever, or hyperparasitaemia (> 10% parasitised red cells), or cerebral malaria (Glasgow coma score < 11), or hypoglycaemia 72 . In the SM cohort (n = 11), 3 patients (27%) had cerebral malaria, 3 patients (27%) had hyperparasitaemia, 5 patients (45%) had renal failure, 4 patients (36%) had hyperbilirubinaemia, and 7 patients (64%) had more than one WHO criterion for severe disease. Asymptomatic malaria-exposed adults, resident in Timika district for at least two years, with no fever or symptoms of malaria within the preceding two weeks were enrolled as controls. Those without parasitaemia were grouped as healthy controls (n = 7) and those with asymptomatic parasitaemia as asymptomatic controls (n = 6). Peripheral blood mononuclear cells (PBMCs) were isolated from heparinised venous blood samples and cryopreserved until analysis.

Study approval
The human studies were approved by the ethics committees of the Malaysian Ministry of Health, the Indonesian National Institute of Health Research and Development, the NT Department of Health and Menzies School of Health Research and QIMR Berghofer Medical Research Institute. Written informed consent was obtained from all participants.

Plasma biomarkers
Total body parasite biomass was quantified by detecting plasma concentration of P. falciparum Histidine Rich Protein 2 (PfHRP2) using ELISA as described 73 . Plasma Ang-2 was measured as a biomarker of endothelial activation by ELISA (Quantikine, R&D Systems) as previously described 74 .

Gene expression analysis by qPCR
Cells as detailed in the text were sorted by FACS directly into RLT buffer and stored at À80°C. On the day of extraction, frozen cell lysates were thawed quickly on ice and mRNA was extracted using the RNeasy Micro Kit (Qiagen) according to the manufacturer's instructions. cDNA was synthesised using oligo-dT and Superscript III (Invitrogen) according to the manufacturer's instructions.
Gene expression was measured using individual TaqMan gene expression assays (Applied Biosystems) and Platinum Taq Polymerase (Life Technologies) in a 15 lL volume reaction using a Rotor-Gene 3000 PCR machine (Corbett Research) with the following conditions: 2 min at 50°C for calibration of fluorescent gain values, then denaturing for 2 min at 95°C, followed by 45 cycles of 5 s at 95°C and 30 s at 60°C. Gene expression was quantified relative to a standard curve generated from a titration of cloned cDNA, and relative gene expression was calculated by dividing the expression value of the test gene by that of a reference gene (RPL13A for human studies and b2m for mouse study) except for the data in Figure 4d, which was normalised as molecules/cell as the input cell number was accurately determined beforehand.

T-cell stimulation with Plasmodium antigens
PBMCs (4 9 10 5 ) from adult patients with uncomplicated P. falciparum malaria (Sabah) were co-cultured for 24 h in 96well U-bottom plates with RPMI supplemented with 10% human AB serum and with an extract equivalent to 4 9 10 5 P. falciparum-parasitised RBCs (pRBCs), or a control extract made from equivalent numbers of uninfected RBCs (nRBCs). The pRBC and nRBC extracts were made as described previously 75 . Following the incubation period, the cells were harvested and sorted by flow cytometry into populations of CD4 hi CD38 hi or CD4 norm T cells. Following the FACS sort, the cells were resuspended in RLT lysis buffer and PCR was performed as described above.

mRNA expression using NanoString
Expression of 130 immune-related genes known to be involved in T-cell and B-cell responses was assessed using a custom 130-plex nCounter codeset of genes involved in immune recognition, survival, migration, adhesion, cytokine/chemokine secretion, activation, differentiation and exhaustion, as reported previously 76 . CD4 norm and CD4 hi CD38 hi T cells were sorted by FACS from PBMCs from samples as detailed in the text. The sorted cells were stimulated for 3 h with PMA (5 ng mL À1 ) and ionomycin (500 ng mL À1 ) in RPMI supplemented with 10% human AB serum and penicillin-streptomycin-glutamine (PSG). Following the stimulation, the cells were washed and resuspended in RLT lysis buffer (2 9 10 3 cells lL À1 ) and stored at À70°C until use. nCounter codeset hybridisation was performed as per the manufacturer's instructions using 5 lL of cell lysate (10 4 cells). Samples were processed using the NanoString GEN2 Prep Station and data acquired using the nCounter Digital Analyzer (NanoString Technologies). Data were normalised to the mean RPLP0 expression levels and processed using nSolver (NanoString Technologies) using the recommended settings. The limit of detection was set as the average reading for the negative controls plus 2 SD; all values below this were raised to this value, and all data were then log 2 -normalised for statistical analysis. Genes were considered as reproducibly detected if they were above the limit of detection in at least 50% of the samples. Principal component analysis, ANOVA, paired ttests, LEfSe and Random Forest analysis were performed using the Genomics Data Miner 21 . Heatmaps were generated using Heatmapper 77 , and data were ordered according to average fold change between CD4 norm and CD4 hi CD38 hi groups.

PBMC culture with IL-27
PBMCs from healthy Brisbane volunteers were co-cultured with an extract made from P. falciparum-infected RBC (as detailed above) at the equivalent of 1xpRBC:1xPBMC in media alone (RPMI, with 5% human AB serum, PSG and 20 units per mL rIL-2), or further supplemented with IL-27 (100 ng mL À1 ) in 2.5 mL at 2.5 9 10 6 PBMCs/well in 12-well tissue culture plates for 8 days; then, the cells were split into 2 wells of 6-well plates with 5 mL of fresh media and supplements and cultured for further 6 days (recombinant IL-2 was obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases and National Institutes of Health: human rIL-2 from Dr Maurice Gately, Hoffmann-LaRoche (Nutley, NJ); recombinant IL-27 from BioLegend. After 13 days of culture, the cells were restimulated for 24 h with the addition of pRBC extract (equivalent of 10 6 pRBC mL -1 ); then, the cells were harvested, and 10 9 10 3 CD4 + T cells were sorted by FACS and immediately resuspended into 200 lL of RLT lysis buffer and stored at À80°C until processed for qPCR.

Mice
Specific pathogen-free BALB/c and C57BL/6 mice were purchased from the Animal Resources Centre, Perth, WA, and were used at 6 or 9 weeks of age. Interferon-gamma reporter mice were purchased from the Jackson Laboratories (C.129S4(B6)-Ifng tm3.1Lky /J; stock No. 017580) and bred at QIMR Berghofer Medical Research Institute. All studies were approved by the QIMR Berghofer Medical Research Institute Animal Research Ethics Committee.

Mouse Plasmodium infections
To generate stocks of parasitised RBC (pRBC), BALB/c mice were infected by tail-vein injection of 1000 cryopreserved infectious P. yoelii sporozoites (17XNL non-lethal strain) kindly provided by Dr SL Hoffman (Sanaria Inc., Rockville, MD, USA) and blood-stage parasitaemia was monitored by FCAB assay 23 . On day 14, the mice were euthanised and the parasitised blood was collected and stored in freezing buffer in liquid nitogen. Blood-stage infections reported herein were initiated by tail-vein injection of 10 5 pRBC, and the infection was monitored daily by FCAB assay 23 . For kinetic analysis of CD4 and IFN-c expression in GREAT IFN-c reporter mice, GREAT mice were infected with P. yoelii as described above, and on the days indicated in the text, three mice were selected randomly for euthanisation, and organs and blood were collected for analysis. Single-cell suspensions were prepared and analysed by flow cytometry. Cytokine measurement in mouse plasma was assessed by cytokine bead array (BD Biosciences) using 10 lL of plasma taken from the tail tip of infected mice on the days shown in the text. Figures 1, 2, 4 and 5 were performed using GraphPad Prism 6 (GraphPad Software Inc.). Data in Figure 1c were analysed using the Kruskal-Wallis test followed by post-tests comparing each group with HC. Data in Figures 2, 4 and 5 were log 2 -normalised and analysed using a paired t-test. In Figure 3, all data were then log 2normalised for statistical analysis and principal component analysis, ANOVA, paired t-tests, LEfSe and Random Forest analysis were performed using the Genomics Data Miner 21 . Heatmaps were generated using Heatmapper 77 .