B-cell activating factor (BAFF) and BAFF-R levels correlate with B-cell subset activation and redistribution in controlled human malaria infection

Abstract

Characteristic features of Plasmodium falciparum malaria are polyclonal B-cell activation and an altered composition of the blood B-cell compartment, including expansion of CD21−CD27− atypical memory B-cells. B-cell activating factor (BAFF) is a key cytokine in B-cell homeostasis, but its potential contribution to the modulation of the blood B-cell pool during malaria remains elusive. In the controlled human malaria model (CHMI) in malaria-naïve Dutch volunteers, we therefore examined the dynamics of BAFF induction and B-cell subset activation and composition, to investigate whether these changes are linked to malaria-induced immune activation and in particular induction of BAFF. Alterations in B-cell composition after CHMI closely resembled those observed in endemic areas. We further found distinct kinetics of proliferation for individual B-cell subsets across all developmental stages. Proliferation peaked either immediately after blood-stage infection or at convalescence, and for most subsets was directly associated with the peak parasitemia. Concomitantly, plasma BAFF levels during CHMI were increased and correlated with membrane-expressed BAFF on monocytes and dendritic cells, as well as blood-stage parasitemia and parasite-induced IFNγ. Correlating with elevated plasma BAFF and IFNγ levels, IgD−CD38lowCD21−CD27− atypical B-cells showed the strongest proliferative response of all memory B-cell subsets. This provides unique evidence for a link between malaria-induced immune activation and temporary expansion of this B-cell subset. Finally, baseline BAFF-receptor levels prior to CHMI were predictive of subsequent changes in proportions of individual B-cell subsets. These findings suggest an important role of BAFF in facilitating B-cell subset proliferation and redistribution as a consequence of malaria-induced immune activation.

Introduction

Humoral immune responses play a major role in conferring naturally-acquired immunity to malaria (1). This immunity, however, appears to be slow to develop and ineffectively maintained (2, 3), also demonstrated by the low prevalence of Plasmodium falciparum (Pf) malaria antigen-specific memory B-cells (MBCs) even in high endemic areas (4–7). While the complex nature of the parasite (2, 5) and the high degree of antigenic variation (8) certainly contribute to this apparently impaired generation of humoral immune memory, there is also increasing evidence that the malaria parasite actively alters B-cell function (9). This includes not only polyclonal activation and modified responsiveness of B-cells in vitro and in vivo (10–14), but also profound changes to the composition of the peripheral blood B-cell compartment as recently described in naturally malaria-exposed populations (15–20). These changes observed in acutely infected or continuously exposed individuals include increased levels of transitional B-cells (15, 17), reduced levels of IgD+CD27+ marginal zone-like non-switched MBCs (17) and an enlarged proportion of atypical MBCs (atypMBCs), which have become a recent research focus (16–20). In malaria-endemic areas, expansion of atypMBCs appears to be linked to both cumulative duration and frequency of parasite exposure (18–20). Due to the cross-sectional nature of most of these studies, however, conclusive evidence for a causal link is missing. Also unknown are the mechanisms governing these alterations of the blood B-cell pool.

A key cytokine in mediating B-cell homeostasis by regulating differentiation and survival is the constitutively expressed B-cell activating factor (BAFF) belonging to the tumor necrosis factor family (21). BAFF is initially synthesized in membrane-anchored form by cytokine-activated myeloid cells such as monocytes and dendritic cells (DCs), and subsequently released after enzymatic cleavage (22). In vitro, surface BAFF production and release by Pf-activated human monocytes and concomitant naïve B-cell activation has been demonstrated (23), while in a murine model of acute malaria infection, reduced surface BAFF-expression by antigen-presenting cells (APCs) corresponds with impaired MBC survival (24). Finally, in children with acute malaria increased plasma BAFF levels have been reported, which correlate with disease severity (25).

A unique tool to gain insights into immunomodulatory effects of the Pf parasite in humans is the controlled human malaria infection (CHMI) model, allowing analysis of sequential samples of previously malaria-naïve volunteers during a primary Pf infection in comparison to their pre-infection status (26–28). We therefore took advantage of the CHMI model to study the dynamics of B-cell activation and modulation during the very early stages of malaria infection. We further comprehensively investigated the kinetics and source of Pf-induced BAFF during CHMI, and addressed the question whether modified BAFF secretion or B-cell BAFF-R expression may provide an explanation for B-cell subset activation or reshaping of the human B-cell compartment during malaria.

Materials and Methods

Study subjects and controlled human malaria infection

Eighteen healthy, malaria naïve Dutch adult volunteers (age 19–30, median 23) were subjected to controlled human malaria infection (CHMI) by intradermal injection of 2500, 10000 or 25000 (n=6 in each group) aseptic, purified, cryo-preserved Pf sporozoites (PfSPZ Challenge, strain Pf NF54) in an open-label phase I clinical trial at the Radboud university medical center from October 2010 to July 2011 (29). The three groups were subjected to CHMI at different time points, in one month intervals. Written informed consent was obtained from each volunteer. The trial was performed in accordance with Good Clinical Practice and an Investigational New Drug application filed with the U.S. Food and Drug Administration. The study was approved by the Central Committee for Research Involving Human Subjects of The Netherlands (CMO CCMO NL31858.091.10). The trial was registered at Clinicaltrials.gov, identifier: NCT 01086917.

As reported previously (29), 15 volunteers (n=5 in each group) developed patent parasitemia as determined by both thick-smear (TS; median pre-patent period with range: 12.6 days (11–14.3)) and retrospective quantitative (q)PCR (10.3 days (9–12)). When TS+ (or at day 21 for volunteers remaining TS−), volunteers were treated with atovaquone/proguanil. There was no significant difference between the three groups by either time to positive qPCR or TS, parasite densities on day of TS positivity (day of treatment; DT) or peak parasite density (measured at time of TS positivity ± 18h).

PBMC isolation, cryopreservation and thawing

Blood samples for peripheral blood mononuclear cell (PBMC) isolation were collected at baseline (challenge C−1), during liver-stage infection (C+5), during developing blood stage infection (C+9), at TS positivity just before treatment (DT), 3 days after treatment (DT+3) and 35 and 140 days after challenge infection (C+35, C+140). PBMC were isolated by density gradient centrifugation from citrate anti-coagulated blood using vacutainer cell preparation tubes (CPT; BD Diagnostics). Following four washes in ice-cold phosphate-buffered saline (PBS), cells were counted and cryo-preserved at a concentration of 10×10⁶ cells/ml in ice-cold FCS (Gibco)/10% DMSO (Merck) using Mr. Frosty freezing containers (Nalgene). Samples were stored in vapour-phase nitrogen. Immediately prior to use, cells were thawed, washed twice in Dutch-modified RPMI 1640 (Gibco/Invitrogen) and counted.

Flow cytometry analysis

Phenotypic analysis of sequential PBMC samples collected at different time points prior to, during and after CHMI was conducted simultaneously for each individual donor in one experiment to avoid confounding influences of day-to-day inter-experimental variation. Ki67 expression on B-cell subsets was determined for all donors (n=18; n=15 TS+). FcRL4, BAFF-R, TACI and BCMA expression on B-cell subsets and surface BAFF expression on antigen-presenting cells was determined in those donors for which sufficient cells were available (n=14; n=11 TS+). Antibodies used for flow cytometry are listed in Table 2. For immunostaining, 500,000 to 1,000,000 cells per stain were transferred into a 96-well v-bottom plate, washed once with 200μl PBS and incubated with 50μl fixable dead cell stain dilution in PBS for 30min on ice. Cells were then washed twice with staining buffer (PBS containing 0.5% bovine serum albumin (Sigma)), and stained with 50μl antibody cocktail diluted in staining buffer for 30min at room temperature (RT), followed by another wash step with staining buffer. This was repeated for the secondary surface staining step. Cells were then re-suspended in 50μl fixation/permeabilization buffer (eBioscience) and incubated for 30min on ice, followed by a wash step with 150μl permeabilization buffer (eBioscience). For intracellular staining, cells were incubated for 30min at RT with 50μl antibody cocktail diluted in permeabilization buffer. Cells were washed with permeabilization buffer, re-suspended in 200μl PBS/1% paraformaldehyde and kept on ice until analyzed. 50,000 to 200,000 events per sample were acquired on a Cyan ADP 9-colour flow cytometer (Dako/Beckman Coulter). Flow cytometry data were analyzed using FlowJo v9.6 software.

Table 2.

Antibodies used for flow cytometry

B-cell panel
purpose	target antigen	fluorochrome	clone	supplier
viability	fixable viability	eFluor 450	n/a	eBioscience
dump channela	CD3	FITC or PE	HIT3	Biolegend
	CD56	FITC or PE	HCD56	Biolegend
	CD14	FITC or PE	HCD14	Biolegend
B-cell lineage	CD19	APC-eF780	HIB19	eBioscience
B-cell subset	CD10	ECD	ALB1	BeckmanCoulter
	CD38	PerCp	HIT2	Biolegend
	CD27	PC7	IA4CD27	BeckmanCoulter
	CD21	APC	B-ly4	BD Biosciences
	IgD	biotin	IA6-2	BD Biosciences
	Streptavidin	Pacific Orange	n/a	Invitrogen
variable marker	BAFF-R	FITC	11C1	Biolegend
	Ki67	FITC	B56	BD Biosciences
	TACI	PE	1A1	Biolegend
	BCMA	PE	polyclonal	R&D Systems
	CD86	FITC	2331 (FUN-1)	BD Biosciences
	CCR6	FITC	53103	R&D Systems
	CXCR5	PE	51505	R&D Systems
	CD24	FITC	ML5	Biolegend
	FcRL4	PE	413D12	Biolegend

DC/monocyte panel
purpose	target antigen	fluorochrome	clone	supplier
viability	lived/dead stain	aqua	n/a	Invitrogen
dump channela	CD3	PerCP/Cy5,5	HIT3	Biolegend
	CD19	PerCP/Cy5,5	HIB19	Biolegend
	CD56	PerCP/Cy5,5	HCD56	Biolegend
Monocyte lineage	CD16	PECy7	3G8	Biolegend
	CD14	Pacific Blue	HCD14	Biolegend
	HLA-DR	APC-Cy7	L243	Biolegend
DC subset	BDCA-1	PE	AD5-8E7	Miltenyi Biotech
	BDCA-2	biotin	AC144	Miltenyi Biotech
	Streptavidin	ECD	n/a	BeckmanCoulter
	BDCA-3	APC	AD5-14H12	Miltenyi Biotech
variable marker	BAFF	FITC	137314	R&D Systems

^aDump channel comprised of lineage markers to gate out non-relevant PBMC subsets

n/a = not applicable

ELISA

Plasma BAFF and IFNγ levels were determined in cryo-preserved EDTA anti-coagulated plasma samples using the Human BAFF/BLyS/TNFSF13B Quantikine ELISA Kit (R&D Systems) and the Human IFNγ ELISA Ready-SET-Go kit (eBioscience) according to the manufacturer’s recommendations.

Statistical analysis

Statistical analysis was performed using graph pad prism software v5. We used parametric tests since the majority of data analyzed was normally distributed as determined by D’Agostino & Pearson omnibus normality test, and since non-parametric tests have limited power to detect significant differences in small data set. Data for more than two time points were analyzed by repeated measures one-way ANOVA. Dunnett’s post-hoc test was used when comparing all time points to C−1 baseline data, while Bonferroni post-test was applied when comparing different B-cell subsets. Data for cell subsets at several time points were analyzed by repeated measures two-way ANOVA with Bonferroni post-hoc test. One sample t-tests were used to determine whether fold changes in B-cell proportions at a given time point compared to C−1 were significantly different from 1 (no change). Relationships between plasma BAFF and surface BAFF or BAFF-R levels were analyzed by Pearson correlation. If parameters were not normally distributed, non-parametric Spearman correlation was used for analysis (relationships between peak parasitemia, plasma BAFF and IFNγ levels, and B-cell subset proliferation).

Results

Sequential blood samples were collected prior to, during and after controlled human malaria infection (CHMI) by intradermal administration of cryo-preserved P. falciparum sporozoites (29). Fifteen volunteers became thick-smear (TS) positive between days 11 and 14.3 after infection and reached peak parasitemia of 3–759 parasites/μl blood (median 56, IQR 15–102). Three donors remained negative by both TS and qPCR until day 21 after challenge (C+21), when they were presumptively drug-treated. These donors were analyzed in parallel with TS+ donors to ensure that any changes observed after treatment were not solely related to drug-treatment. Hematological parameters showed significant changes over the course of infection (Table 1). Decreased total leukocyte, lymphocyte and platelet counts in TS positive donors were most evident at three days after treatment (DT+3; all p<0.001), with lymphocyte counts also significantly decreased on day of treatment (DT; p<0.05). The proportion of total B-cells within PBMCs mirrored this trend, showing a significant decrease at DT+3 compared to baseline (C−1; p<0.05).

Table 1.

Hematological parameters at baseline and during CHMI in thick-smear positive volunteers

	C−1	DT	DT+3	C+35	p-value compared to C−1d
Leukocyte counta	5.1 (4.3–6.2)	4.9 (4.5–5.6)	2.6 (2.0–3.8)	5.4 (4.4–5.8)	DT+3***
Lymphocyte counta	1.7 (1.52–2.31)	1.1 (0.84–1.67)	0.88 (0.81–1.18)	2.14 (1.93–2.62)	DT*, DT+3***
% B-cellsb	8.92 (6.03–12.23)	10.66 (6.54–15.17)	5.35 (3.43–7.15)	8.41 (6.13–12.2)	DT+3*
Platelet counta	240 (190–251)	224 (186–281)	149 (95–191)	243 (213–297)	DT+3***
Hbc	8.2 (7.7–8.9)	8 (7.7–8.7)	8.1 (7.6–8.5)	7.9 (7.3–8.5)

^aLeukocyte, lymphocyte and platelet counts x 10⁹/L, median ± IQR

^bB-cells as % of viable PBMC by flow cytometry, median ± IQR

^cHb in mm/L, median ± IQR

^d*p<0.05, ***p<0.001 as determined by one-way ANOVA with Dunnett’s post-hoc test

IQR = interquartile range, C−1 = day before challenge, DT = day of treatment

Increased plasma BAFF during CHMI is associated with parasitemia and IFNγ secretion

During CHMI, a statistically significant increase in plasma BAFF was detected from the day of TS positivity (day of treatment, DT) onwards (Figure 1A, B). This increase was absent in the three donors who did not develop blood-stage parasitemia (Figure 1B). BAFF levels peaked at DT+2 or DT+3, with a median 3.3 fold increase (IQR 2.3–5.9) and an absolute increase of 1053 pg/ml (median; IQR 616–2956 pg/ml) compared to baseline. Increased plasma BAFF concentrations were preceded by elevated plasma IFNγ (Figure 1B), an important factor in mediating BAFF release (22). The rise in IFNγ closely followed the increase in parasitemia, and peak IFNγ levels correlated with peak parasitemia (Spearman r=0.70, p=0.004, Figure 1C). The interval between peak IFNγ (on DT+1) and peak BAFF levels (on DT+3) in plasma was two days (Figure 1B), and peak BAFF levels positively correlated with peak IFNγ concentrations (r=0.78, p=0.001: Figure 1D). The correlation between peak BAFF and parasite load, however, was only weak and did not reach significance (r=0.47, p=0.08: Figure 1E).

Figure 1. Plasma BAFF levels during CHMI.

Plasma BAFF was quantified for (A,B,D,E) n=15 thick-smear positive (TS+) and (B) n=3 thick-smear negative (TS−) donors on baseline (C−1), and from one day prior to (DT−1) until three days after treatment (DT+3) and after resolved infection (C+35). (A) Dots depict individual TS+ donors, error bars the median and IQR. Asterisks show significant differences compared to C−1 by one-way ANOVA with Dunnett’s post-hoc test (* p<0.05; ** p<0.01; *** p<0.001). (B) Kinetics of Pf parasite load (open triangles), plasma IFNγ (grey filled circles) and BAFF (black filled squares) were analyzed in TS+ donors, depicted as mean and SEM. BAFF in TS− donors is shown in open squares (light grey). The individual Pf load per day was calculated per donor as the mean of all PCR samples taken on that day for this individual donor. Relationships between (C) peak Pf load and peak IFNγ, (D) peak IFNγ and peak plasma BAFF and (E) peak Pf load and peak plasma BAFF in TS+ donors were analyzed by non-parametric Spearman correlation.

Surface BAFF expression on antigen-presenting cell populations is increased during CHMI

Major sources of plasma BAFF are myeloid cells, which initially express the cytokine in membrane-bound form (22). Indeed, HLADR+ lineage (CD3, CD19, CD56)-negative antigen-presenting cells (APCs) showed increased levels of surface BAFF expression at DT+3 (p<0.001; Figure 2A,B), and this increase in BAFF+ APCs correlated with the increase in plasma BAFF levels (Pearson r=0.70, p=0.016; Figure 2C). We next analyzed which APC populations contributed to CHMI-induced surface-BAFF expression. Based on differential expression of the LPS receptor CD14 and the low affinity Fcγ receptor III CD16, HLADR+lin- APCs were subdivided into classical monocytes (CD14+CD16−), intermediate monocytes (CD14+CD16+), inflammatory monocytes (CD14−CD16+). CD14−CD16− APCs were further gated on BDCA-1+ DCs, BDCA-2+ DCs and BDCA-3+ DCs (Figure 2D). The proportion of BAFF+ APCs within PBMCs was already markedly elevated on DT and then further increased until DT+3 (Figure 2E) – a pattern that was again not observed in TS negative individuals (Figure S1). CD14+CD16− classical monocytes constituted the largest proportion of BAFF+ APCs, followed by inflammatory monocytes (Figure 2E). Classical, inflammatory and intermediate monocytes as well as BDCA-1 DCs showed a significant increase in BAFF+ cells on DT+3 compared to C−1 (Figure 2F), while no such increase was seen for BDCA-2 and BDCA-3 DCs. The fold increase in BAFF+ cells was highest within classical monocytes (median 2.76, IQR 1.7–3.5), followed by inflammatory monocytes (2.63 (2.3–4.5)) and BDCA-1 DCs (2.18 (1.4–3.7)). When analyzing absolute percentages of BAFF+ cells, there was no difference between the three monocyte subsets at baseline (C−1; median with IQR: class mono (2.9% (2.1–5.1%); interm mono (3.0% (1.7–4.0%); inflamm mono (3.2% (2.8–4.2%)). On DT+3, however, inflammatory monocytes (11.1% (8.1–14.5%) showed significantly higher levels of BAFF expression than classical (7.1% (5.4–10.7%; p<0.05) and intermediate monocytes (5.2% (3.3–8.2%); p<0.001).

Figure 2. Surface BAFF expression on DC and monocyte populations during CHMI.

Surface BAFF expression on HLADR+ CD3/CD19/CD56-negative antigen-presenting cell (APC) subsets was determined by flow cytometry. Data are shown as representative flow cytometry plots for (A) one donor and (B) for all analyzed TS+ donors (n=11), with dots depicting individual donors, error bars the median and IQR. (C) The relationship between the increases (calculated by substracting C−1 from DT+3 values) in plasma BAFF levels and proportion of BAFF+ APCs was determined by Pearson correlation analysis. (D) APCs were further subdivided into (i) classical monocytes (class mono; CD14+CD16−), (ii) intermediate monocytes (interm mono; CD14+CD16+), (iii) inflammatory monocytes (inflamm mono; CD14−CD16+) and CD14−CD16− (iv) BDCA-1+ DCs, (v) BDCA-2+ DCs and (vi) BDCA-3+ DCs. (E) Median proportions of BAFF+ APC subsets were analyzed within viable PBMCs on C−1, DT and DT+3. (F) Percentages of BAFF+ cells within APC subsets are shown as individual data, medians and IQR. Asterisks show significant differences compared to C−1 by one-way ANOVA with Dunnett’s post-hoc test (* p<0.05; ** p<0.01; *** p<0.001).

CHMI induces low, transient FcRL4 expression and proliferation of B-cell subsets with distinct kinetics

Increased BAFF secretion following CHMI is likely to have an impact on B-cell activation. To analyze the peripheral blood B-cell compartment, we developed a 9-colour B-cell panel and gating strategy to delineate ten phenotypically distinct B-cell populations (Figure 3A) based firstly on IgD and CD38 expression (Figure 3B), followed by further subdivision using CD10, CD21 and CD27 (Figure 3C). As reported previously, CD21−CD27− MBCs in healthy, malaria naive donors (prior to CHMI) constitute only a small proportion of circulating B-cells (median 1.96%, IQR 1.76–2.66%) in contrast for instance to classical MBCs (cMBCs; median 12.43%, IQR 9.9–16.64%; Figure 3D). These CD21−CD27− MBCs closely resembled the phenotype (high expression of CCR6 and CD86, low expression of CXCR5 and CD24; Figure 3E) reported for so-called atypical CD21−CD27−MBCs expanded in malaria-exposed individuals living in highly-endemic areas (16, 19). In contrast to a subset of tonsil B-cells (Figure 3F), peripheral blood B-cells, including CD21−CD27− atypMBCs from healthy Dutch individuals did not express the inhibitory Fc receptor-like protein (FcRL)4 (Figure 3G, H). FcRL4 expression on B-cells was not induced during or immediately after CHMI, but two weeks post-treatment (C+35; p<0.001) in TS+ donors (Figure 3H), and not in those three which remained TS− (data not shown). This induction of FcRL4 expression was only temporary, occurred on a very small proportion of B-cells, did not correlate with peak parasitemia, IFNγ or BAFF levels, and was not confined to atypMBCs (p<0.05), but also observed in cMBC (p<0.05), non-switched (ns)MBC (p<0.001), activated (act)MBC (p<0.01), classical naïve B-cells (cN) (p<0.001) and CD21−CD27− double-negative naïve B-cells (dnN) (p<0.05) (Figure 3I).

Figure 3. Proportion and phenotype of B-cell subsets during CHMI.

Following exclusion of debris, doublets and dead cells, lineage (CD3/CD56/CD14)-negative, CD19+ B-cells were subdivided based on (A) IgD and CD38 and then on (B) CD10 expression. (C) CD38hi B-cells were divided into (i) CD10−IgD−CD38hiCD27+ plasma blasts (PB) and (ii) CD10+ IgD+CD38hiCD27− transitional B-cells (TBC). CD38lowCD10− B-cells were subdivided first based on IgD and then CD21 and CD27 expression into four pairs of switched and non-switched/naïve B-cell populations (C): (iii) CD21−CD27+ activated MBCs (actMBC) and (iv) activated naïve B-cells (actN); (v) CD21+CD27+ classical MBCs (cMBC) and (vi) non-switched MBCs (nsMBC); (vii) CD21+CD27− MBC (CD27− MBC) and (viii) classical naïve B-cells (cN); and (ix) CD21−CD27− atypical MBCs (atypMBC) and (x) double-negative naïve B-cells (dnN). Panel (D) shows proportions of individual B-cell subsets within total CD19+ B-cells at baseline (C−1). (E) PBMCs from healthy, malaria-naïve volunteers (n=10) were analyzed by flow cytometry to determine surface expression of CD86, CCR6, CXCR5 and CD24 on cN B-cells, cMBC and atypMBC. Data are depicted as whisker-box-plots, boxes indicating the IQR and whiskers the min/max values. Representative flow cytometry plots are shown for FcRL4 expression on (F) total tonsil B-cells as a positive staining control and (G) peripheral blood B-cell subsets in healthy, malaria-naïve donors (n=11) prior to (C−1) or after (C+35) CHMI. FcRL4+ cells were analyzed as proportion of (H) total CD19+ B-cells or (I) within individual B-cell subsets in n=11 TS+ volunteers. (H, I) Dots depict individual donors. (H) Error bars show the median and IQR. (I) Asterices show significant differences compared to C−1 by one-way ANOVA with Dunnett’s post-hoc test (* p<0.05; ** p<0.01; *** p<0.001).

B-cell activation was assessed by Ki67 expression (30), which is only found in currently dividing cells. Elevated levels of proliferation following CHMI were observed in nearly all B-cell subsets (Figure 4A) with the exception of actMBCs (Figure 4Aiii) and plasma blasts (PBs), the latter being consistently more than 95% Ki67+ (Figure 4Ai). This proliferative response showed distinct kinetics for individual B-cell subsets: Proliferation of transitional B-cells (TBCs), cMBCs, CD27− MBCs and atypMBCs (all p<0.001; Figure 4Aii,v, vii and ix) as well as dnN B-cells (p<0.05, Figure 4Ax) peaked at DT+3. Amongst MBC subsets, atypMBCs showed the strongest proliferative response at DT+3 (Figure 4B). Finally, significant proliferative responses were found three weeks after resolved malaria infection (C+35) for cN B-cells (p<0.01; Figure 4Aviii), activated naïve (actN) and nsMBCs (both p<0.001; Figure 4Aiv and vi) and were still ongoing for dnN (p<0.001, Figure Ax) and cMBCs (p<0.01; Figure 4Av). The three donors remaining negative for parasitemia showed no proliferation (Figure S2). Despite the strong proliferative responses, also in the MBC compartment, we found no evidence for hypergammaglobulinaemia, with plasma IgG levels remaining stable during CHMI (median with IQR: C−1 8.98 mg/ml (8.2–12.3); DT 9.11 mg/ml (6.4–13.1); DT+3 7.47mg/ml (6.5–9.3); C+35 (8.98 mg/ml (6.9–10.4).

Figure 4. Proliferative response of B-cell subsets during CHMI.

Ki67 expression by individual B-cell subsets was determined by flow cytometry of PBMCs collected prior to CHMI (C−1) and during liver (C+5) and developing blood-stage (C+9), immediately prior to (DT) and three days after treatment (DT+3) and after parasite clearance (C+35, C+140). (A) Flow cytometry plots showing Ki67 gating in TBCs, cMBCs and atypMBCs. (B) Data are expressed as percentage of Ki67+ cells within each individual B-cell subset. Dots depict individual TS+ donors (n=15), error bars the median and IQR. Asterisks show significant differences compared to C−1 by one-way ANOVA with Dunnett’s post-hoc test (* p<0.05; ** p<0.01; *** p<0.001). (C) The fold change in the percentage of Ki67+ cells within by individual MBC subsets at DT+3 was compared to C−1 was calculated. Dots depict individual TS+ donors (n=15), error bars the median and IQR. Asterisks show significant differences between memory B-cell subsets by one-way ANOVA with Bonferroni post-hoc test (*** p<0.001). Relationships between atypMBC proliferation at DT+3 with (D) peak parasitemia, (E) DT+3 plasma BAFF levels and (F) peak IFNγ in TS+ donors (n=15) were analyzed by non-parametric Spearman correlation.

The proportion of Ki67+ cells correlated with peak parasitemia within atypMBC (Spearman r=0.56, p=0.03; Figure 4C) and cMBCs (r=0.71, p=0.003) on DT+3, as well as within cMBC (r=0.63, p=0.01), CD21−CD27− dnN (r=0.55, p=0.03) and cN B-cells (r=0.62, p=0.014) on C+35. Moreover, on DT+3 the proportion of proliferating atypMBCs (r=0.68, p=0.005; Figure 4D) and their non-switched dnN counterparts (r=0.64, p=0.01), but not other B-cell subsets, correlated with plasma BAFF. There was no correlation between DT+3 plasma BAFF levels and B-cell subset proliferation on C+35. Since plasma BAFF levels strongly correlated with plasma IFNg, we also assessed the relationship between peak IFNg levels and B-cell subset proliferation. As for BAFF, peak IFNg concentrations correlated with the proportion of proliferating atypMBCs (r=0.86, p<0.0001; Figure 4E), their non-switched dnN counterparts (r=0.64, p=0.01) and, in contrast to BAFF, also with cMBCs proliferation (r=0.57, p=0.03).

Altered B-cell subset proportions associate with BAFF-R expression, but not proliferation

Selective proliferation of individual B-cell subsets at different time points during CHMI might affect the composition of the peripheral blood B-cell compartment. Indeed, the percentage of PB, TBC, atypMBC and dnN within CD19+ B-cells was significantly elevated at DT+3 (Figure 5A). There was, however, no correlation between the increased proportion and proliferative response for any of these subsets at this time point. Proportions of other subsets were significantly decreased during exposure to blood-stage parasitemia (cN and cMBC at DT+3; nsMBC at DT and DT+3; Figure 5A and data not shown), despite active proliferation (cMBCs at DT+3). Finally, increases in the proportion of activated naïve and memory B-cell subsets were particularly evident after parasite clearance (C+35; Figure 5B), but again not associated with proliferation at this time point. These patterns of altered B-cell subset proportions were again not found for the three donors remaining negative for parasitemia (Figure S3).

Figure 5. BAFF-R expression and its association with altered B-cell subset proportions during CHMI.

Flow cytometry was conducted on PBMCs collected prior to, during and after CHMI. B-cell subsets were analyzed as percentage of total viable CD19+ B-cells, and their fold change in proportion compared to C−1 was calculated for (A) DT+3 and (B) C+35. Data are depicted as whisker-box-plots, indicating median, IQR and min/max values of n=15 TS+ donors. Asterisks show significant differences compared to 1 (no change, dashed line) as tested by one sample t-test. BAFF-R levels on (C, E) individual B-cell subsets and (D) total B-cells were determined by flow cytometry analysis of PBMC samples from n=11 TS+ donors and are expressed as median fluorescence intensity (MFI). Panel (C) shows the fold change in the proportion of each individual B-cell subset on DT+3 (compared to C−1) plotted against baseline (C−1) BAFF-R levels, analyzed by non-parametric Spearman correlation. Small grey dots depict all individual B-cell subsets from each individual donor, while large black dots represent the median (of n=11 TS+ donors) fold change in subset proportion and BAFF-R expression for each of the B-cell subsets: i (PB); ii (TBC); iii (actMBC); iv (actN); v (cMBC); vi (nsMBC); vii (CD27− MBC); viii (cN); ix (atypMBC); x (dnN). (D) Data for total B-cells are depicted for individual donors (dots), error bars show the median and IQR. Asterisks show significant differences compared to C−1 by one-way ANOVA with Dunnett’s post-hoc test. (E) Data for individual B-cell subsets are shown for C−1 (white) and DT+3 (grey), depicted as whisker-box-plots, indicating median, IQR and min/max values of n=11 TS+ donors. The dashed line in (D) and (E) indicates median BAFF-R expression levels on CD19-negative lymphocytes. Asterisks show significant differences compared to C−1 by two-way ANOVA with Bonferroni post-hoc test. (F) The relationship between the DT+3 plasma BAFF levels and median B-cell BAFF-R expression levels on DT+3 was determined by Pearson correlation analysis. (* p<0.05; ** p<0.01; *** p<0.001).

Another potential cause of altered proportions of individual circulating B-cell subsets is re-distribution between blood and lymphatic tissues, and this chemotactic B-cell migration can be augmented by BAFF signaling through its receptor BAFF-R (31). Intriguingly, we found an inverse correlation between baseline BAFF-R expression levels of the ten different B-cell subsets and their change in proportion at DT+3 (Spearman r=−0.33, p=0.0004): B-cell C−1, while B-cell subsets (nsMBCs and cMBCs) that were decreased the most had the highest baseline BAFF-R levels (Figure 5C).

During acute infection (DT and DT+3), B-cell BAFF-R expression was strongly reduced (Figure 5D), particularly on B-cell subsets with high baseline BAFF-R levels (Figure 5E). DT+3 B-cell BAFF-R expression levels inversely correlated with DT+3 plasma BAFF concentrations (Pearson r=−0.63, p=0.04; Figure 5F), suggesting a negative impact of high plasma BAFF levels on BAFF-R expression. Expression levels of the two other BAFF receptors, TACI and BCMA, however, remained unaltered by CHMI on all B-cell subsets, except cMBCs for which we observed a slight increase at DT+3 (both p<0.01 compared to C−1; data not shown).

Discussion

In this study, we investigated the kinetics and source of Pf-induced BAFF, and its association with B-cell subset activation and modulation of the composition of the human B-cell compartment during the very early stages of malaria infection. Within three days after peak parasitemia and anti-malaria treatment, we found a significant increase in both surface BAFF-expression on various APC subsets and plasma BAFF concentrations, as well as strong proliferative responses and altered proportions of numerous B-cell subsets. Both BAFF induction and B-cell subset proliferation directly correlated with peak parasitemia. For CD21−CD27− B-cell subsets (atypMBC and dnN), proliferation correlated with plasma BAFF and IFNγ levels, and for B-cell subsets with particularly high or baseline BAFF-R levels, these were inversely associated with their change in proportion following CHMI.

Increased plasma BAFF during CHMI is in line with previous findings in naturally exposed individuals, showing elevated BAFF in plasma from acutely malaria-infected children (25) and in placental tissue from malaria-infected pregnant women (32). In vitro, both malarial hemozoin and soluble parasite extract are capable of inducing BAFF release from monocytes (23). We now show for the first time that in vivo, malaria infection also results in increased expression of the membrane-bound form of BAFF on various monocyte populations as well as on myeloid BDCA-1 DCs up to three weeks after resolved infection. Myeloid cells expressing BAFF can act on B-cells in two manners: directly by cross-linking the memory B-cell BAFF receptor TACI via surface-expressed BAFF, and indirectly by enzymatic release of soluble BAFF with high affinity for BAFF-R (33). The herein observed increase in surface BAFF-positive BDCA-1 DCs stands in contrast to findings in a murine malaria model showing a loss of surface BAFF-expressing myeloid DCs from the spleen (and as a result reduced survival of MBCs) during acute infection (24). Since we only examined circulating, but not lymphoid tissue-resident APCs, we cannot exclude that a similar depletion of BAFF expressing DCs might also be observed in these organs following CHMI – possibly due to transition to the circulation. Our data are reminiscent of findings in human HIV patients (34), who also show elevated levels of monocyte- and DC-expressed membrane BAFF, plasma BAFF and alterations/activation in the B-cell compartment similar to what is observed in malaria (35–40). In as how far in either setting increases in APC surface-expressed BAFF via TACI, and soluble BAFF via BAFF-R, indeed contribute to activating B-cells, however, remains to be established. The monocyte subset showing the highest proportion of BAFF-expression three days after drug treatment were CD14low/−CD16+ monocytes. This subset is one of two “non-classical” CD16+ monocyte subsets that are known to be increased upon immune activation, as also found for P. falciparum malaria (41, 42). CD14low/− monocytes have pro-inflammatory properties (including TNFa secretion), while CD14+CD16+ intermediate monocytes secrete the anti-inflammatory cytokine IL-10 (43, 44). Significantly higher BAFF expression on the CD14low/− subset as found in our volunteers during CHMI is consistent with this functional division.

Parasite-induced BAFF secretion from monocytes might be further augmented by IFNγ derived from innate or adaptive immune cells activation (22). Although previous experiments were performed with highly enriched monocytes/B-cell co-cultures, it could not be definitively concluded that this process was entirely T-cell independent (23). Indeed, we also found only a weak and not significant relationship between peak parasitemia and plasma BAFF levels, indicating that in vivo this direct effect may be of lesser importance. Instead, peak plasma BAFF levels correlated with peak plasma IFNγ levels, an important factor in mediating BAFF release from myeloid cells (22), which preceded them by two days, and is in line with previous findings (25). This strongly suggests that Pf-driven immune activation is an important contributing factor in parasite-mediated BAFF secretion. Peak IFNγ levels were are only reached one and three days after initiation of treatment, likely due to the fact that release of Pf material upon mass parasite killing further enhances immune activation. The two day delay between peak plasma IFNγ and BAFF concentrations can be explained with the time it takes for APCs to get activated by IFNγ and initiate sufficient surface BAFF protein expression (which also peaked two days later) and cleavage, and is consistent with in vitro studies showing that BAFF release from IFNγ-treated monocytes is only minimal after 24 hours, peaks after 48 hours and surface expression continues to increase until 72h (23).

Concomitantly with rising plasma BAFF levels, we observed reduced expression of BAFF-R on B-cells, corroborating findings in acutely malaria-infected children (25), HIV infection, Sjogren’s Syndrome and systemic lupus erythematosus (SLE) (34, 35, 38). At least in autoimmune diseases, this down-regulation is not transcriptionally regulated (38). Instead, physiological regulation by receptor internalization or shedding, but also partial masking by BAFF binding may be the explanation. Two other members of the BAFF receptor family, TACI and BCMA, which have low or no affinity for soluble BAFF (33), did show no down-regulation during acute infection.

Despite very low parasite densities during CHMI, we observed a number of profound albeit temporary changes in B-cell composition either during or immediately after the blood-stage of CHMI. These included increased proportions of TBCs, atypMBCs and PBs as well as the reduction in marginal zone-like nsMBCs (CD21+CD27+IgD+), and are consistent with previous observations in persistently exposed or acutely infected individuals in malaria-endemic areas (15, 17–20, 45, 46). A number of factors can influence the proportions of individual subsets within the peripheral blood B-cell compartment, including redistribution between blood and tissues, cell death or proliferation. Our data demonstrate that different B-cell subsets respond with different proliferation kinetics to Pf exposure. Proliferative responses during acute infection are likely related to either direct interaction with blood-stage parasite products or as a bystander-effect of Pf-induced immune activation and release of soluble mediators, both of which would be driven by the level of parasitemia encountered during infection. Indeed, proliferative responses of several B-cell subsets correlated with the degree of parasite exposure. Although antigen-specific expansion cannot be excluded, this is unlikely true for the majority of cells, seeing the large proportion of responding cells across subsets after a primary exposure. Moreover, while atypMBCs have been shown to contain a similar proportion of malaria-antigen specific cells as cMBCs (16), any circulating proliferating B-cells seen as early as three days after drug-treatment and thus only 5 days after blood-stage parasites became detectable by PCR are unlikely to stem from a memory-generating germinal centre response, which would still be ongoing at this time point. Instead, any changes in cell proportion or proliferation seen already at the peak of infection are more likely related to generalized immune activation during acute infection rather than antigen-specific activation and the generation of memory responses. Whether this is mediated by cytokines such as IFNgamma or BAFF (suggested by the correlation between plasma levels of those cytokines and atypMBC and dnN proliferation) (23, 25) or by direct parasite-B-cell interaction (10, 13, 14) remains to be established. One possible consequence of generalized B-cell activation could be hypergammaglobulinaemia (1), however, we did not detect any increase in plasma IgG during CHMI, nor does this necessarily occur in the field (25).

Next to B-cell subset proliferation during acute infection, we also observed proliferative responses as late as 35 days after infection (and thus three weeks after parasite clearance) in cN, activated and double-negative naïve B-cells as well as non-switched and classical MBCs. The majority of these subsets also showed reduced proportions in the B-cell compartment 3 days after drug treatment. While these reduced proportions are likely due to redistribution of B-cells from the circulation for instance to secondary lymphoid organs such as the spleen, we cannot exclude that there might also be a loss in B-cells due for example to apoptosis (47, 48). An explanation for these late proliferative responses might therefore be either a) the release of elsewhere recruited activated cells back into the circulation or b) a physiological counter reaction to replenish the apoptosis-diminished B-cell pool.

For atypMBCs, as well as their non-switched dnN counterparts, their particularly strong proliferative response correlated not only with peak parasitemia, but also plasma BAFF levels – a relationship not observed for other B-cell subsets. This may appear counterintuitive, since CD21− B-cells express lower levels of BAFF-R than other B-cell subsets. A possible explanation is that BAFF may be a co-stimulating factor rather than driving force of proliferation (49–51). If CD21−CD27− naïve and MBCs are more receptive than other B-cell subsets to BAFF-co-stimulated proliferation-driving stimuli, low BAFF-R expression would not necessarily be a limiting factor. In this scenario, factors other than BAFF should also correlate with CD21−CD27− B-cell proliferation. This is true for peak IFNγ levels, and may further extend to other B-cell stimuli not assessed in the present study. To establish whether there is indeed a causal link between the expansion of CD21−CD27− B-cells and Pf-induced cytokines, and to unravel potentially synergistic effects of BAFF, IFNγ and other cytokines or stimuli, future mechanistic in vitro studies are needed. Of note, in other diseases with elevated plasma BAFF, including HIV and SLE, atypMBC or phenotypically similar populations are also expanded (34, 38–40).

This study provides the first direct link between malaria-infection and increased proportions of atypMBCs. The temporary nature of this increase is in line with decreasing proportions of these cells in individuals from malaria-endemic areas after prolonged non-exposure (52). The fact that atypMBCs at baseline only showed little proliferation, but increased this following parasite exposure suggests that these cells may not be exhausted per se. This is also in line with a sizable proportion of atypMBCs showing proliferation in vivo in naturally exposed individuals (16). The notion of atypMBC exhaustion stems from a failed in vitro attempt to differentiate them into antibody-producing cells (19). Future studies will need to show whether activation of atypMBCs in vivo can result in the generation of antibody-producing plasma blasts after all, as suggested previously (16), or whether atypMBCs may have alternative functional properties upon activation.

While CHMI-activated CD21−CD27− atypMBCs closely resemble the phenotype of atypMBCs in viraemic HIV-patients and individuals from highly malaria-endemic areas in regards to CD21, CD27, CD86, CCR6, CXCR5 and CD24 expression (19, 40), they lack expression of FcRL4. FcRL4 is an IgA-binding inhibitory receptor (53) that impairs B-cell receptor signaling, but augments TLR responses (54, 55). IgD−CD27− cells (of which IgD−CD21−CD27− MBCs are a subpopulation) in healthy individuals or in those with SLE also lack FcRL4 expression (39). FcRL4 expression was induced by CHMI two weeks after parasite clearance, but this induction was temporary and occurred not only on atypMBCs, but numerous B-cell subsets. It is thus possible that atypMBC during CHMI may functionally differ from those found in malaria-endemic areas. However, even in frequently malaria-exposed individuals FcRL4 expression varies (19, 52). Future studies will be necessary to determine whether sustained, high FcRL4 expression is a necessary or specific feature of atypMBCs (the function of which is still unknown), may only be induced at high levels upon chronic immune activation, and what the triggers for this induction are. Very recently, the HIV envelope protein gp120 has been shown to trigger FcRL4 expression in primary human B-cells by direct interaction with B-cell expressed α4β7 and subsequent induction of TGFβ (56). If P. falciparum similarly expresses FcRL4-inducing molecules remains to be established. Of note, FcRL4 expression by gp120 in vitro was increased within 24 hours, while we only observed FcRL4 expression 3 weeks after peak parasitemia and drug treatment-mediated parasite clearance. We cannot exclude, however, that FcRL4 expression might have already peaked much earlier and remained stable for a prolonged period of time, since there were no PBMC samples collected between DT+3 and C+35. To our knowledge, this is the first study investigating changes in B-cell FcRL4 expression in vivo following an acute infection or immune activation, and follow-up studies are needed to further investigate the kinetics of FcRL4 expression after infection. The temporary induction of inhibitory FcRL4 expression might be yet another facet of negative immune regulation upon activation. In T-cells, this well-known process is mediated by inhibitory receptors such as PD-1 or CTLA-4 (57), pathways that have also been described and contribute to reduced T-cell responsiveness in malaria infection (20, 58). Importantly, negative immune regulation of T-cells not only leads to what is described as “exhaustion”, but is an important factor in preventing immune pathology (59) and in mediating contraction of immune responses when a pathogen is cleared (60). In analogy, temporary induction of FcRL4 on activated B-cells across all B-cell subsets following malaria infection may be a physiological response and simply serve to bring these cells back to the steady state.

Despite partially overlapping kinetics, we found no correlation between B-cell subset proliferation and the change in their individual proportions, suggesting selective B-cell subsets redistribution as a more important parameter in the altered composition of the peripheral blood B-cell compartment during malaria. BAFF has previously been shown to enhance B-cell chemotaxis to the CCR7, CXCR4 and CXCR5 ligands CCL21, CXCL12 and CXCL13 (31), which direct B-cell migration to lymphatic tissues. Amongst other chemokines, CXCL13 is induced during acute Pf infection (61). In a high BAFF environment, B-cell subsets with higher BAFF-R expression might thus be more readily induced to leave the circulation than those expressing little BAFF-R. In line with this hypothesis, we found an inverse association between baseline BAFF-R levels on individual subsets and their proportion within the B-cell compartment on DT+3, i.e. the time of highest plasma BAFF levels. This phenomenon might be further exacerbated by differential expression of the corresponding chemokine receptors by different B-cell subsets. AtypMBCs and PB, for instance, express low levels of CCR7, CXCR4 and CXCR5 (data not shown and (19, 40)).

In summary, by analyzing longitudinal samples collected during CHMI, we were able to extract potentially causal relationships between parasite exposure and B-cell activation and modulation during malaria. We show that plasma BAFF levels are increased in the context of Pf-induced immune activation and may be at least partially derived from monocyte subsets and BDCA-1+ DCs, which increase membrane BAFF-expression during CHMI. B-cell subsets were activated and proliferated with distinct kinetics, and these responses depended on peak parasitemia levels during CHMI. Finally, our data suggest that parasite-induced BAFF elevation may contribute to orchestrating the changes in the B-cell compartment by two distinct mechanisms, namely facilitating B-cell subset proliferation and redistribution. This phenomenon is likely not malaria-intrinsic but may be a common pathway of B-cell modulation, since malaria shares several features including polyclonal B-cell activation and specific alterations in the phenotype and composition of the peripheral B-cell pool with other diseases that are also characterized by excessive plasma BAFF levels, including HIV infection and SLE.

Abbreviations used in this paper

actMBC

activated memory B-cell

actN

activated naïve B-cell

atypMBC

atypical memory B-cell

BAFF

B-cell activating factor

BAFF-R

BAFF receptor

BCMA

B-cell maturation antigen

BDCA

blood dendritic cell antigen

C

challenge

CHMI

controlled human malaria infection

cMBC

classical memory B-cell

cN

classical naïve B-cell

DC

dendritic cell

DT

day of treatment

FcRL4

Fc receptor-like protein 4

MBC

memory B-cell

nsMBC

non-switched memory B-cell

PB

plasma blast

Pf

Plasmodium falciparum

PfSPZ

Plasmodium falciparum sporozoites

TACI

transmembrane activator and calcium modulator and cyclophilin ligand interactor

TBC

transitional memory B-cell

TS

thick-smear