CD28 deficiency leads to accumulation of germinal-center independent IgM+ experienced B cells and to production of protective IgM during experimental malaria

Protective immunity to blood-stage malaria is attributed to Plasmodium-specific IgG and effector-memory T helper 1 (Th1) cells. However, mice lacking the costimulatory receptor CD28 (CD28KO) maintain chronic parasitemia at low levels and do not succumb to infection, suggesting that other immune responses contribute to parasite control. We report here that CD28KO mice develop long-lasting non-sterile immunity and survive lethal parasite challenge. This protection correlated with a progressive increase of anti-parasite IgM serum levels during chronic infection. Serum IgM from chronically infected CD28KO mice recognize erythrocytes infected with mature parasites, and effectively control Plasmodium infection by promoting parasite lysis and uptake. These antibodies also recognize autoantigens and antigens from other pathogens. Chronically infected CD28KO mice have high numbers of IgM+ plasmocytes and experienced B cells, exhibiting a germinal-center independent Fas+GL7-CD38+CD73- phenotype. These cells are also present in chronically infected C57BL/6 mice although in lower numbers. Finally, IgM+ experienced B cells from cured C57BL/6 and CD28KO mice proliferate and produce anti-parasite IgM in response to infected erythrocytes. This study demonstrates that CD28 deficiency results in the generation of germinal-center independent IgM+ experienced B cells and the production of protective IgM during experimental malaria, providing evidence for an additional mechanism by which the immune system controls Plasmodium infection.


Introduction
Protection against clinical blood-stage malaria in humans and mice typically involves parasitespecific IgG antibody production [1] [2]. Data from mouse malaria models suggest that PLOS  production of these antibodies depends on CD4 + T cells and mostly occurs after control of acute infection [3] [4]. Among the malaria mouse models, Plasmodium chabaudi (Pc) infection has been used to investigate the development of adaptive immunity due to its similarities to the human disease that is caused by Plasmodium falciparum [5]. The early CD4 + T cell response to Pc infection provides large amounts of pro-inflammatory cytokines and helps B cells to secrete polyclonal IgG [6] [7]. However, Pc-infected mice also produce IgM in a T-cell independent manner [4] [7]; IgM production is also observed in humans exposed to malaria [8] [9], however a protective role for these antibodies is unclear [10]. Additionally, somatically hypermutated IgM+ memory B cells are found in both humans and mice infected with Plasmodium [11]. CD28 is a costimulatory molecule fundamental for the full development of CD4 + T cell responses [12] and CD4 + T cell-driven antibody class switch [13]. We previously showed that mice lacking CD28 do not eliminate chronic Pc parasitemia, due to the lack of memory CD4 + T cells and anti-parasite IgG [14]. However, despite the absence of full protective immunity, parasitemia in these mice persists at low levels during chronic infection, suggesting the contribution of other protective mechanisms. IgM participates in several immune effector mechanisms, such as complement system activation [15], antigen agglutination [16], dead and damaged cell scavenging [17] and lymphocyte activation through Fcμ receptors [18]. During encapsulated bacterial infections, IgM opsonizes bacilli, facilitates their removal by phagocytic cells and effectively combats the infection [19] [20].
A full characterization of IgM produced in response to Plasmodium infection, as well as its potential anti-pathogenic roles have not been studied yet. We hypothesized that CD28KO mice would offer a good model to investigate the protective role of IgM against malaria given their deficiency in developing acquired immunity. The present study shows that CD28KO mice accumulated serum anti-parasite IgM in response to chronic parasitemia. The IgM response was associated with high numbers of IgM-producing plasmocytes and IgM + experienced B cells in the spleen. Our results show that IgM produced in response to chronic parasitemia promotes parasite control in CD28KO mice, suggesting an additional antimalarial mechanism for protection against malaria.

CD28KO mice develop long-lasting non-sterile protective immunity against blood-stage Pc malaria
In accordance with our previous study [14], CD28KO (Cd28 -/-) mice infected with Pc-infected red blood cells (Pc-iRBCs) controlled the first parasitemia peak, but developed increased chronic parasitemia as defined by the presence of detectable parasitemia percentages in the circulating blood (i.e., above 0.1%) (Fig 1A). In C57BL/6 (Cd28 +/+ ) mice, Pc-iRBCs were no longer detected by microscopic examination after clearance of acute parasitemia. Because generation of classic memory T and B cell responses to Pc infection requires CD28 signaling [14], it is intriguing how CD28KO mice survive acute infection and maintain relatively low levels of chronic parasitemia. To investigate whether this protection depends on parasite persistence, C57BL/6 and CD28KO mice at 30 days post-infection (p.i.) were submitted to a curative chloroquine treatment and then challenged with a lethal parasite dose at 40 or 80 days p.i. (c40 and c80 mice, respectively) ( Fig 1B). In C57BL/6 c40 mice, the parasites were no longer detected by microscopic examination after 2 days of challenge (Fig 1C), while C57BL/6 c80 mice had limited parasitemia at <0.1% (Fig 1D). Interestingly, CD28KO c40 and c80 mice almost completely controlled the Pc re-infection, limiting parasitemia at~0.1% and~1%, respectively. In both cases, CD28KO and C57BL/6 negative controls failed to control challenge-induced parasitemia and succumbed ( Fig 1C and 1D and data not shown). Furthermore, all the re-infected CD28KO mice (as well as re-infected C57BL/6 mice) survived (data not shown). Our results suggest the existence of an alternative effector mechanism to ensure long-lasting immunity in CD28KO mice.

Anti-parasite and parasite-unrelated IgM serum levels correlate with delayed parasitemia in CD28KO mice
We next investigated the effector mechanisms responsible for Pc control in the absence of CD28. First, the anti-parasite serum IgM kinetics were determined in infected C57BL/6 and CD28KO mice. In C57BL/6 mice, anti-parasite IgM peaked at 15 days p.i., and subsequently decreased during chronic infection (Fig 2A). In contrast, infected CD28KO mice had a gradual increase in IgM, achieving maximal levels at 100 days p.i. As previously reported [14], anti-parasite IgG production was completely abrogated in infected CD28KO mice (Fig 2B and 2C). The total IgM kinetics in CD28KO mice (but not C57BL/6) were similar to those observed for the anti-parasite IgM (S1A Fig). Total IgG, however, was not apparently increased in either of the infected mouse groups (S1B and S1C Fig).
The augmented total IgM levels observed led us to investigate whether these antibodies also recognize Pc-unrelated antigens, such as autoantigens (e.g., thyroglobulin, histone, dsDNA and myelin) and antigens from other pathogens (e.g., Trypanosoma cruzi-T. cruzi and Bacille  Calmette Guerin-BCG). Enhanced IgM titers for all these antigens were observed in CD28KO mice at 100 days p.i. (Fig 2D and 2E). The kinetics of parasite-unrelated IgM followed those of anti-parasite IgM in infected C57BL/6 and CD28KO mice, except for anti-BCG antibodies, which reached higher levels on day 100 p.i. in both mouse groups.
To verify whether persisting infection was responsible for the high IgM serum levels found in the absence of CD28, mice at 30 days p.i. were treated with chloroquine to eliminate residual parasitemia. At day 40 p.i., reductions in the anti-parasite and total IgM levels were observed in cured CD28KO mice ( Fig 2F and S1D Fig). Indeed, anti-parasite IgM levels in these mice were similar to those found in C57BL/6 mice (data not shown). Notably, anti-parasite and total IgM levels remained increased in cured CD28KO mice at 80 days p.i. compared with the non-infected controls ( Fig 2G and S1E Fig). Secondary infections in cured C57BL/6 and CD28KO mice (at 40 and 80 days p.i.) led to further increases in these levels. C57BL/6 mice also produced high amounts of anti-parasite IgG after re-infection (data not shown).

Chronically infected CD28KO mice have increased splenic populations of IgM + experienced B cells and IgM + plasmocytes
Next, the splenic B cell populations in chronically infected CD28KO mice were characterized. These mice had a dramatic increase in spleen weight and cellularity compared to C57BL/6 mice at 100 days p.i. or to age-matched controls, which was dependent on parasitemia persistence (S2 Fig). The numbers of splenic B (CD19 + ) cells and intracellular (i)IgM + CD138 + plasmocytes were also significantly higher in infected CD28KO mice (Fig 3A and 3B). Supporting the requirement of CD28 for GC formation [14], the Fas + GL7 + B cell population was much smaller in infected CD28KO mice than in their C57BL/6 counterparts (S2C and S2D Fig). The few GC-like B cells were IgM + in CD28KO mice (Fig 3C), which were also found in C57BL/6 mice ( Fig 3D). Remarkably, on day 100 p.i., CD28KO mice showed a sharp increase in a splenic B cell population that had characteristics of GC-independent IgM + memory B cells The Fas + GL7 -IgM + CD38 + CD73 -CD19 + cell numbers per spleen at day 40 p.i. in mice that were or were not treated with chloroquine (CQ). In a-j, significant differences ( Ã p<0.05, ÃÃ p<0.01, ÃÃÃ p<0.001) between the indicated groups are shown. Data from three independent experiments (n = 6, mean ± SEM) is shown. [21]. The IgM + Fas + GL7 -B cells (Fig 3E and S2C Fig) expressed IgM and the memory B cell marker, CD38 (Fig 3F and 3G) and were predominantly negative for the ecto-nucleotidase CD73 (Fig 3H and 3I). Residual parasitemia was crucial for the accumulation of this population, as the chloroquine-treated CD28KO mice had lower numbers of these cells (Fig 3J). We herein called this population "IgM + experienced B cells". We next assessed whether Pc parasites could directly induce expansion of IgM + experienced B cells in cured CD28KO mice as well as induce differentiation of these cells for antibody production. The cured CD28KO mice that were challenged with Pc parasites displayed increased numbers of Fas + GL7 -IgM + CD38 + CD73experienced B cells and IgM + plasmocytes (Fig 4A  and 4B). Fas + GL7 + B cells were predominant in cured and Pc-challenged C57BL/6 mice ( Fig  4C). Sorted IgM + experienced B cells from cured C57BL/6 and CD28KO mice were then stimulated in vitro with Pc-iRBCs or LPS, which are polyclonal B cell activators [22]. After 72 h of culture, a considerable proportion of IgM + experienced B cells from C57BL/6 and CD28KO mice proliferated and differentiated into iIgM + CD138 + cells in response to Pc-iRBCs, a phenomenon that was also observed following LPS stimulation (Fig 4D and 4E). iIgM -(IgG + ) plasmocytes, however, were only generated from WT IgM + experienced B cells, but not from CD28KO counterparts (Fig 4E). Anti-parasite and total IgM production levels were also increased in cultured cells that were stimulated with Pc-iRBCs (Fig 4F and 4G). Generally, IgM + experienced B cell responses in the CD28KO mice were higher than those in the C57BL/ 6 mice. Altogether, these results suggest IgM + experienced B cells can differentiate in iIgM + plasmocytes and produce IgM in response to parasite.

IgM from chronically infected CD28KO mice recognizes mature Pc-iRBCs and is effective in parasite control
We next evaluated whether IgM from CD28KO mice recognized Pc-iRBCs. Antibody-free Pc-iRBCs (isolated from infected RAGKO mice) were incubated with IgM from chronically infected C57BL/6 or CD28KO mice. IgM from 15 day-infected C57BL/6 and from 100 dayinfected CD28KO mice recognized higher proportions of mature Pc-iRBCs compared to IgM from naïve mice (Fig 5A). Furthermore, IgM from 100 day-infected CD28KO mice transferred protection to RAGKO mice (Fig 5B). A delay in parasitemia of several days was observed after this treatment, while IgM from non-infected CD28KO mice was not protective.
We also assessed the functionality of IgM from 100 day-infected CD28KO mice. These antibodies significantly enhanced complement-mediated Pc-iRBC lysis, while IgM from 100 dayinfected C57BL/6 mice or non-infected controls were ineffective (Fig 6A). IgM from 100 dayinfected CD28KO mice also increased the in vitro uptake of SYTO 16-labeled mature Pc-iRBCs by splenic CD11c + (dendritic cells) and F4/80 + cells (macrophages) (Fig 6B). Moreover, pre-incubation with these antibodies enhanced the in vivo uptake of mature Pc-iRBCs by splenic dendritic cells. This was demonstrated by inoculating cell tracker orange (CMTPX)labelled mature Pc-iRBCs, which was or was not pre-incubated with IgM from 100 dayinfected CD28KO mice, into CD11c.YFP mice. Intravital imaging revealed enhanced 3D colocalization of mature Pc-iRBCs with CD11c + YFP + cells inside the splenic red pulp in the presence of antibodies ( Fig 6C). Overall, these results suggest IgM produced in response to Pc infection induces multiple mechanisms that can recognize and act on Plasmodium parasites.

Discussion
In this study, we evaluated the role of IgM produced in response to experimental malaria using chronically infected CD28KO mice as a model, which showed increased parasitemia due to a defect in developing classic memory T and B cells [14]. Interestingly, Pc-infected CD28KO mice acquired long-lasting non-sterile immunity and survived a lethal parasite challenge. The protective immunity in CD28KO mice is associated with a progressive increase in anti-parasite and total IgM and, notably, these antibodies recognize Pc-iRBCs and promote parasite clearance. Additionally, the analysis of chronically infected CD28KO mice allowed us to  characterize a population of Fas + GL7 -IgM + CD38 + CD73 -B cells, which was also present in lower numbers in chronically infected C57BL/6 mice. These IgM + B cells differentiated into IgM + plasmocytes and produced anti-parasite IgM in response to Pc-iRBCs, in a T cell-independent manner, suggesting these cells could be a source of IgM-producing cells in our model. As showed in Fig 2G, there is an increase in production of IgM in response to challenge in cured mice, indicating a "memory-like" response for B cells in both C57BL/6 and CD28KO mice. Although this could indicate the IgM + B cells found in this context to be memory cells, the presence of patent blood parasitemia was fundamental for their accumulation, which could indicate these cells as being "constantly activated/effector" B cells rather than bona-fide memory. Hence, this population can be better defined as "IgM + experienced B cells". We also found that the vast majority of the few cells with a GC phenotype in CD28KO mice are IgM + . Although this could indicate the generation of functional germinal centers in the absence of CD28 co-stimulation generate preferentially unswitched cells, the low numbers of these cells In vitro uptake by splenic F4/80 + macrophages and dendritic cells of mature iRBCs, which were or were not previously incubated with purified IgM from chronically infected CD28KO mice. SYTO 16-labeled Pc-iRBCs from RAGKO mice were incubated with purified IgM for 30 min and then cultured with phagocyte-enriched splenocytes for 20 min. The SYTO 16 + F4/80 + cell and SYTO 16 + CD11c + cell percentage data are shown, which was obtained by flow cytometry. (c) In vivo uptake by splenic red pulp dendritic cells of mature iRBCs, which were or were not previously incubated with purified IgM. The CMTPX-labelled iRBCs from the RAGKO mice was incubated with purified IgM and inoculated i.v. into CD11c-YFP mice. The scale bars represent 50 μm. The CMTPX + YFP + cell percentage data in the YFP + cell population is shown. In a-c, significant differences ( Ã p<0.05, ÃÃ p<0.01, ÃÃÃ p<0.001) between the experimental groups are shown. Data from three independent experiments (n = 6, mean ± SEM) is shown. make this assumption unlikely. Overall, this study helps to understand the role of IgM in chronic malaria.
The role of IgM in protective immunity against Plasmodium infection is still controversial, where both a positive role via production of neutralizing IgM [23] and reduced effector function associated with IgM were reported [24]. Our results indicate that IgM plays an important protective role in situations where conventional acquired immunity is not optimal, such as immunodeficiency and prolonged chronic parasitemia in face of impaired classical immunological memory. The function of these cells in C57BL/6 mice is unclear. The high levels of IgG specific antibodies, together with higher levels of CD4+ T cell memory, might mask some of the protective functions that could be attributed to IgM (and IgM + experienced B cells). Moreover, the presence of patent chronic parasitemia favors the expansion of these cells (Fig 3), which does not happen in C57BL/6 mice. Nevertheless, IgM isolated from d100 C57BL/6 mice can lead to phagocytosis and complement induced lysis, albeit not in the same extent to the observed in CD28KO mice (Fig 6). This makes it possible that IgM from B6 mice is indeed functional. This finding supports the speculation that IgM + memory B cells are remnants of primitive immune system that persisted throughout evolution [25] and thus act as a backup protective mechanism upon infection. Like innate B cells [26], these cells could have been evolutionarily selected to produce germ-line gene encoded polyreactive antibodies that bind to multiple conserved microbial molecules and damage-associated self-components. Reinforcing this idea, we showed that IgM produced by chronically infected CD28KO mice (in reaction to the presence of patent blood Pc infection) recognized the parasite as well as autoantigens and antigens from other pathogens. Uninfected CD28KO mice were not producing these antibodies-as we stated above, this is believed to be mainly a result of polyclonal activation in response to patent blood-stage infection. Alternatively, the prevalence and accumulation of IgM did not prevent parasite persistence in CD28KO mice, which could indicate that the parasite evolved mechanisms that can subvert the immune response into a less effective type in certain circumstances. The notion of malaria-induced autoantibodies is not new; in humans, autoantibodies are produced during malaria infection [27]. A logical prediction would be that the recognition of auto-antigens by malaria-induced IgM could induce autoimmunity at some extent in chronically infected mice. We did not observe any effect on our experimental mice. Perhaps at later time points (which we did not explore) we could have observed such consequences.
The protective role of IgM was suggested by our data showing that IgM from chronically infected CD28KO mice was more efficient in limiting parasite replication in vivo than IgM from naïve CD28KO mice. IgM from chronically infected CD28KO mice efficiently recognized and opsonized Pc-iRBCs, which led to phagocytosis by splenic phagocytes, as well as inducing complement-dependent Pc-iRBCs lysis. These antibodies may target Pc-iRBCs by binding to parasite antigens that are expressed on the erythrocyte surface or damage-associated self-molecules, such as Band 3 clusters [28] [29]. In support of our data, anti-parasite IgM has been previously shown to protect against malaria in mouse models [30] [31]. In humans, binding of non-immune or 'natural' IgM to the P. falciparum-iRBC surface is supposedly due to the interaction of a subset of the parasite variant surface antigen, P. falciparum erythrocyte membrane protein 1 (PfEMP1), with the Fc regions of IgM [32]. This interaction is often interpreted as deleterious to the host as it leads to the formation of rosettes that induce immunopathology [33], and it contributes to immune system evasion by masking of protective IgG epitopes [34]. However, the positive correlation of IgM response to parasite antigens with resistance to P. falciparum infection in African ethnic groups indicates a protective role of these antibodies [35].
Altogether, our study reveals a protective role for IgM in experimental malaria. It also shows that increased IgM production occurs when infection develops in the absence of classic memory T and B cells. Moreover, IgM production is dependent upon parasite persistence. These results may help to explain why IgM-producing B cells are expanded in malaria patients living in endemic areas [35]. The rapid decline of anti-parasite IgG serum levels after interruption of parasite exposure suggests classical humoral memory is functionally impaired in malaria patients [35,36]. As a counterpart, IgM + memory B cells might arise as a primitive, innate-like defence mechanism against infection.

Mice, parasites, infection and chloroquine treatment
Six-to eight-week-old C57BL/6, CD28KO, RAGKO and CD11c.YFP (all in a C57BL/6 background) female mice (Jackson Laboratory, USA) were bred under specific pathogen-free conditions at the Isogenic Mouse Facility (ICB-USP). Pc parasites (AS strain) were maintained as described [37]. Mice were inoculated intraperitoneally (i.p.) with 1 x 10 6 Pc-iRBCs for primary infections and 1 x 10 8 Pc-iRBCs for secondary infections and intravenously (i.v.) with 1 x 10 8 Pc-iRBCs to assess in vivo phagocytosis. For complete parasite elimination, mice were treated i.p. with 10 mg/kg body weight/day of chloroquine (Sigma-Aldrich, USA) over 8 consecutive days. Parasitemia was determined by microscopic examination of Giemsa-stained blood smears. Mice were sacrificed using a CO 2 chamber.

Ethics statement
All procedures were in accordance with the national regulations of ethical guidelines for mouse experimentation and welfare of the Conselho Nacional de Saúde and Colégio Brasileiro em Experimentação Animal (COBEA)-Brazil, and the protocols were approved by the Health Animal Committee (Comissão de Ética no Uso de Animais-CEUA-ICB-USP), with permit numbers 0174/2011 and 0036/2007.

In vitro stimulation of IgM + memory B cells
Fas + GL7 -IgM + CD38 + CD73 -CD19 + cells were sorted with a FACSAria device (BD Biosciences), and then they were stained with 5 μM carboxyfluorescein succinimidyl ester (CFSE, Life Technologies) in PBS with 0.1% bovine serum albumin (BSA, Sigma-Aldrich) for 20 min at 37˚C. The cells were then stimulated with Pc-iRBCs (1 B cell: 3 Pc-iRBCs) or LPS (10 μg/ml, from Escherichia coli 0111:B4, Sigma-Aldrich). Proliferation was evaluated after 72 h of culture at 37˚C in a 5% CO 2 atmosphere, whereas supernatants were collected at 7 days to evaluate the total and anti-parasite IgM levels.

IgM purification
Mouse serum IgM was purified by affinity chromatography in a G-Sepharose protein column (GE Healthcare, USA). Sepharose-bound IgM was eluted with 0.01 M sodium phosphate buffer (pH 7.0) and monitored by absorbance readings at 280 nm. The eluted samples were dialyzed in PBS, concentrated and stored at -20˚C. The purity of IgM after isolation was >98% in all samples. Equilibration was performed with 20 mM sodium phosphate, 800 mM ammonium sulfate, pH 7.5.

IgM binding to mature Pc-iRBCs
Antibody-free Pc-iRBCs were obtained from the blood of infected RAGKO mice during a period of the circadian parasite cycle in which mature stages predominate (>95% late trophozoites/schizonts). For our convenience, the mice were placed in an animal room with an inverted light/dark cycle for at least 15 days before infection. Blood cells (5 x 10 7 ) were incubated with purified IgM (at a final concentration of 20 ηg/ml) for 45 min at 37˚C. The cells were stained with PE-labelled anti-IgM b mAb and 5 μM SYTO 16 (Invitrogen) and were then analysed by flow cytometry.

Adoptive transfer of purified IgM
Purified IgM samples (500 μl at 20 ηg/ml) were inoculated i.v. into RAGKO mice after 24 h of infection with 1 x 10 6 Pc-iRBCs, and parasitemia was monitored as described above.

IgM-mediated lysis of mature Pc-iRBCs
The ability of IgM to induce complement-mediated lysis of Pc-iRBCs was evaluated with an adaptation of the haemolysis assay [40]. In brief, 5 x 10 7 antibody-free mature Pc-iRBCs were pre-incubated with purified IgM as described above and then maintained for 45 min at 37˚C with 50 μl of RAGKO mouse serum, as an antibody-free source of complement components. The background and total cell lysis levels were evaluated by incubation of Pc-iRBCs with PBS (background) and H 2 O (total cell lysis). The absorbance of the supernatant was measured at 414 nm with a Spectra Max 190 spectrophotometer and expressed as lysis percentage.

In vitro phagocytosis of mature Pc-iRBCs
Mature Pc-iRBCs were pre-incubated with purified IgM and stained with SYTO 16, as described above. These cells (3 x 10 6 ) were then cultured with phagocyte-enriched splenocytes (1 x 10 6 ) for 20 min at 37˚C in a 5% CO 2 atmosphere and analysed by flow cytometry.

In vivo phagocytosis of mature Pc-iRBCs
Phagocytosis of Pc-iRBCs by splenic dendritic cells was previously reported [41]. CD11c.YFP mice were inoculated i.v. with CMTPX-stained mature Pc-iRBCs (1 x 10 8 ), which were or were not pre-incubated with purified IgM as described above. After 30 min, the mice were deeply i. p. anesthetized with 55 ng/g/body weight of ketamine (Imalgene 1000, Merial, USA) and 0.85 ng/g/body weight of xylazine (Rompun 2%, Bayer, Germany). The spleens were externalized, and live imaging was conducted with a LSM 780-NLO confocal microscope (Zeiss, Germany). The data were processed with the ZEN 2012 lite software (Zeiss). The 28 μm Z-sections with 4 μm Z-increments were acquired for 30 min. An Imaris X64 7.0.0 (Andor Technology) was used to edit images and to determine the CMTPX + CD11c + cell percentages.

Statistical analysis
Statistical analyses were performed with the GraphPad Prism 5 software (GraphPad, USA), and the differences between the groups were considered significant when p < 0.05 (5%). The simultaneous effects of two factors were analysed with the two-way ANOVA and the Bonferroni post-hoc test. The one-way ANOVA and the Tukey post-hoc test was used to assess the effects of only one factor. All samples were evaluated by the Kolmogorov-Smirnov test to assess Gaussian distribution of samples.