Tissue-Specific B-Cell Dysfunction and Generalized Memory B-Cell Loss during Acute SIV Infection

Background Primary HIV-infected patients display severe and irreversible damage to different blood B-cell subsets which is not restored by highly efficient anti-retroviral therapy (HAART). Because longitudinal investigations of primary HIV-infection is limited by the availability of lymphoid organs, we studied the tissue-specific B-cell dysfunctions in acutely simian immunodeficiency virus (SIV) mac251-infected Cynomolgus macaques. Methods and Findings Experiments were performed on three groups of macaques infected for 14, 21 or 28 days and on three groups of animals treated with HAART for two-weeks either initiated at 4 h, 7 or 14 days post-infection (p.i.). We have simultaneously compared changes in B-cell phenotypes and functions and tissue organization of B-cell areas in various lymphoid organs. We showed that SIV induced a steady decline in SIgG-expressing memory (SIgD−CD27+) B-cells in spleen and lymph nodes during the first 4 weeks of infection, concomitant to selective homing/sequestration of B-cells to the small intestine and spleen. SIV non-specific Ig production was transiently increased before D14p.i., whereas SIV-specific Ig production was only detectable after D14p.i., coinciding with the presence of CD8+ T-cells and IgG-expressing plasma cells within germinal centres. Transient B-cell apoptosis on D14p.i. and commitment to terminal differentiation contributed to memory B-cell loss. HAART abrogated B-cell apoptosis, homing to the small intestine and SIV-specific Ig production but had minimal effect on early Ig production, increased B-cell proportions in spleen and loss of memory B-cells. Therefore, virus–B-cell interactions and SIV-induced inflammatory cytokines may differently contribute to early B-cell dysfunction and impaired SIV/HIV-specific antibody response. Conclusions These data establish tissue-specific impairments in B-cell trafficking and functions and a generalized and steady memory B-cell loss in secondary lymphoid organs. Characterization of underlying mechanisms would be helpful in designing new therapeutic strategies to dampen B-cell activation and increases HIV/SIV specific antibody response.

Introduction B-cell dysfunction represents a central feature of HIV infection and an important pathogenic mechanism [1][2][3][4][5]. In the absence of highly active antiretroviral therapy (HAART), HIV-1 infection is associated with a wide range of B-cell defects, including polyclonal hypergammaglobulinemia and the presence of immature/transitional CD10+ or exhausted CD27 negative B-cells in blood [5,6]. Decreased expression of CXCR5 on blood B-cells [7] but increased proportions of CD38-expressing B-cells have been described as a consequence of abnormal trafficking of germinal centre (GC)-like B-cells into blood [8]. More recently, Cagigi et al have shown that the decrease in CXCR5 expression is concomitant to abnormal CXCL13 production by peripheral and lymph node B-cells and increased B-cell responsiveness to CXCL13 in HIV-1 infected patients with low CD4+ T-cells counts [9]. As CXCR5/CXCL13 pair is essential for the entry of naive B-cells and marginal zone (MGZ) B-cells into follicles [10,11], altered expression of this chemokine receptor-ligand pair may contribute to abnormal B-cell trafficking during the course of HIV-1 infection. In secondary lymphoid tissue from HIV-1 infected individuals, follicular hyperplasia and alterations in the architecture of GC and splenic MGZ were observed [12][13][14]. Despite polyclonal activation, the humoral response is strongly impaired, resulting in a decreased response to natural or vaccine T-independent and T-dependent antigens [15] and a loss of peripheral memory B-cells [16][17][18]. Most of these defects are considered as the hallmarks of the chronic phase of infection and are frequently correlated with increased plasma viral load (pVL) and loss of CD4+ cells [5]. However, recent studies in humans [15,19] suggest that early activation has a major role in shaping Bcell (memory and plasma cell) repertoire and trafficking [19][20][21]. In particular, primary HIV-infected patients have severe and irreversible damage to several blood B-cell subsets, which is not counteracted by HAART despite an increase in CD4+ T-cell counts and a decreased viral load [3].
The study of primary HIV-1 infection in humans is limited by the availability of lymphoid organs and the difficulties associated with performing longitudinal investigations using tissues other than blood or tonsils. We thus used a model of experimental pathogenic infection in cynomolgus macaques-a well-known, suitable model that reproduces long-lasting HIV-1 disease [22]-to investigate Bcell dysfunctions during acute HIV infection. Experiments were performed on three groups of macaques infected by SIV mac251 strain for 14, 21 or 28 days (placebo-treated groups) and on three groups of animals treated with HAART for two-weeks either initiated at 4 h, 7 or 14 days post-infection (p.i.) and sacrificed at D14, 21 and D28 p.i. (HAART-treated groups). After having characterized naive, memory and MGZ B-cell subsets in various lymphoid organs from non-infected animals, we have compared the changes in these B-cell subsets in blood, lymph nodes (LN) and spleen of placebo-and HAART-treated macaques and examined whether correlations could be established with immunological or virological parameters. Changes in B-cell functions in vitro (proliferation, apoptosis or immunoglogulins (Ig) production) as well as in tissue organization of areas populated by B-cells within lymphoid tissues (spleen, small intestine and mesenteric LN) were concomitantly examined in placebo-or HAART-treated animals.
Our data show that SIV induced a transient increase in B-cell apoptosis and SIV non-specific Ig production by D14p.i., a steady loss of memory B-cells in spleen and peripheral LN but promoted preferential B-cell trafficking to the small intestine and spleen. HAART initiated 4 hp.i. strongly decreased B-cell apoptosis and B-cell seeding of gut mucosa, but not memory B-cell loss. The production of SIV non-specific IgG was similar in placebo-and HAART-treated animals whereas that of SIV-specific antibodies, only detectable after D14p.i., was abrogated by HAART initiated on D7p.i.. These data establish tissue-specific impairments in Bcell trafficking and functions but a generalized and steady memory B-cell loss in secondary lymphoid organs. Our results provide a clearer understanding of the effects of SIV/HIV-1 on the B-cell compartment during the acute/primary phase of infection.

SIV infection induces a B-cell increase in spleen
We compared the proportions of lymphocyte subsets in blood, spleen and peripheral (inguinal and axillary) LN from various groups of animals infected for 14, 21 or 28 days. Baseline values for blood samples were obtained from each animal before infection, whereas those for spleen and LN were obtained from non-infected animals (controls).

Memory B-cell loss is induced early in acute SIV infection
Consistent with previous data obtained in humans [25][26][27], the combined analysis of CD20, CD27 and Surface IgD (SIgD) expression allows the delineation of three blood B-cell subsets in the macaque: SIgD + CD272 naive B-cells, SIgD2CD27 + memory B-cells and SIgD + CD27 + MGZ B-cells (Figure 1.B). In constrast to humans, we found that the proportion of peripheral memory Bcells was higher than that of naive B-cells (6864% versus 1162% of B-cells, respectively) in non-infected macaques (Figure 1.C). The proportions of blood MGZ B-cells were similar in humans (1262%) [25] and macaques (1765%). In LN, memory cells accounted for 5965%, naive cells for 1564% and MGZ B-cells for 2266% of the total B-cell population ( Figure 1.D), whereas they accounted for 39610%, 3065% and 1765% of the total Bcell population in spleen (Figure 1.E). In agreement with our data, Vugmeyster et al reported that smaller proportions of naive (CD20 + CD272) B-cells and higher proportions of CD20 + CD27 + B-cells are present in blood and LN of cynomolgus macaques than those in humans [28]. A population of GC B-cells of variable size has been characterized in human lymphoid organs by the coexpression of CD20 antigen with CD10, CD38 and CD77 antigens. Unfortunately, staining with any of the specific antibodies (Ab) was not suitable to define a similar GC population in macaques (data not shown).
SIV infection induced an early but transient loss of peripheral memory B-cells, with a two-and 1.4-fold decrease observed in the percentage of cells at D7p.i. and D28p.i., respectively. A 4.3and three-fold increase in peripheral naive B-cells was observed on D7 and D28p.i., respectively. Changes in MGZ B-cells were only detected on D21 and D28 p.i. (1.4-and 1.2-fold decrease, respectively) ( Figure 1.C). In LN, the proportion of memory Bcells had decreased by factors of 1.3 by D14 p.i., 2.6 by D21p.i. and 2.2 by D28 p.i., respectively, whereas the percentage of naive B-cells increased by a factor of 2.7 from D14p.i. (Figure 1.D). The percentage of MGZ B-cells transiently decreased with a nadir observed at D14p.i. (1.6-fold decrease) and values approaching baseline at D21 p.i.. Results were similar for axillary and inguinal LN (data not shown). In spleen, memory B-cells were decreased by a factor of 2.3 by D14, 5.6 by D21 and 3.9 by D28 p.i.. The percentages observed for MGZ B-cells between D14 and D28 p.i. were between 1.2 and 1.5 times lower than baseline. In contrast, the percentage of naive B-cells steadily increased from D14p.i. by a factor of 2.3. (Figure 1

Differential changes in memory B-cells from various lymphoid organs during SIV infection
It is well established that memory SIgD2CD27 + B-cells preferentially express SIgG/A over IgM in humans. We therefore quantified the relative levels of CD27 + B-cells expressing SIgG and SIgM in macaques. The ratio between SIgM and SIgG in blood B-cells did not differ significantly between non-infected and SIV-infected animals at any time point tested during the four weeks of infection ( Table 2). By contrast, this ratio progressively increased in LN and spleen, with 2.4and 2.8-fold increases detected at D28 p.i. in these two tissues, respectively. These data suggest that switched memory B-cells were preferentially depleted in secondary lymphoid organs.
As in humans, macaque memory B-cells can express CD80 and CD86 co-stimulatory molecules: most peripheral memory B-cells  (SIgD2CD27 + ) co-expressed CD80 and CD86 in non-infected animals ( Table 2). After infection, the percentages of both peripheral CD80 + and CD86 + memory B-cells decreased, reaching a nadir at D7 p.i. (1.8-and 1.9-fold decreases, respectively). A significant correlation was observed between the percentages of SIgD2CD27 + memory B-cells and those of SIgD2CD86 + (Rho = 0.93, p,0.003) or SIgD2CD80 + (Rho = 0.92, p,0.003) B-cells. In LN, most memory B-cells were either CD80 + or CD86 + in non-infected animals (data not shown). By D14 p.i., the percentage of CD80+ cells had decreased by a factor of 1.5, CD86+ had decreased by a factor of 1.2 and total memory B-cells by a factor of 1.3. At D21 and D28 p.i., the percentage of CD80 + had increased, approaching the baseline value, whereas the percentages of both CD86 + and total memory B-cells continued to decrease by similar amounts. The percentage of total SIgD2CD27 + memory B-cells was correlated with that of SIgD2CD86 + (Rho = 0.65, p,0.004) but not SIgD2CD80 + Bcells. In spleen, the percentage of CD80 + was lower than that of  Table 2). A significant correlation was found between these percentages in blood (Rho = 0.98, p,0.002) and LN (Rho = 0.79, p,0.001) during acute SIV-infection.
Our data thus show a transient depletion of blood CD80 + CD86 + CD95 + memory B-cells without modification of the ratio between SIgM-and SIgG-expressing cells. In contrast, a preferential and sustained depletion of SIgG-, CD86-, CD95expressing memory B-cells was observed in LN and spleen.

SIV infection transiently affects B-cell survival, proliferation and Ig production
To assess whether the decrease in memory B-cells results from apoptosis, anergy or terminal differentiation, we investigated the effect of SIV infection on spleen B-cell function. The proliferation rate of SIV-infected spleen B-cells to B-cell receptor (BCR) stimulation was three (IL2+ SAC) and four (IL4+ anti-Ig H+L Ab) times lower than in non-infected animals at D14 p.i., but three and 1.4 times higher thereafter. Similarly, the response to CD40 stimulation was four times lower at D14p.i. and six and 8.4 times greater at D21 and D28 p.i., respectively (Figure 2.A). Levels of spontaneous and BCR-induced apoptosis were 20% higher in SIV-infected animals than in non-infected ones at D14p.i.. (Figure 2.B). Spontaneous and BCR-induced apoptosis at D21 and D28 p.i. were two and three times lower, respectively, than those observed in non-infected animals.
In addition to these effects on B-cell proliferation and apoptosis, a strong increase in plasma Ig concentration was observed in SIVinfected animals from D7 p.i.. However, the kinetics of IgG and IgM production differed (Figure 2.C): as compared to baseline values, IgM production was significantly increased by 29% up to D14 p.i. and by 54% to 69% thereafter, whereas IgG production was transiently increased by 50 to 60% by D14p.i. (p,0.001). Changes in IgG production were a direct result of an impaired B-cell response. Indeed, spontaneous and cytokine (IL2+IL10)mediated IgG production by purified spleen B-cells was increased in SIV-infected animals at D14p.i. (Figure 2.E), probably reflecting the in vivo commitment to plasma cell differentiation. At D21p.i., IgG levels were still higher than control values in supernatants from spleen cells cultured with medium or stimulated with cytokines. The addition of CD40 mAb further enhanced cytokine-mediated IgG production in culture supernatants from non-infected and SIV-infected macaques on D21p.i. but not on D14p.i.. This suggests that available SIV-responsive B-cells are committed to plasma differentiation by D14p.i., whereas new SIVresponsive B-cells can be recruited into plasma cell differentiation from D21p.i.. Consistent with this notion, SIV-specific Ab were detected in only two out of seven plasma samples from D14p.i. but were present in most plasma samples from D21p.i. The median Ab plasma concentration thus significantly increased between D21 and D28 p.i. (Figure 2.D).

Early HAART strongly decreases plasma viral load but not depletion of blood B-cells
To improve our understanding of the kinectics of B-cell dysfunctions during the acute phase of SIV infection, we compared data obtained in placebo-and HAART-treated macaques. Animals were subjected to HAART for two-weeks initiated at 4 h (beginning of the infection), 7 or 14 (peak of pVL) days p.i. and sacrificed at the end of HAART (i.e.: on D14 (H14), 21 (H21) and 28 (H28) p.i.) (Figure 3.A). At each time point, comparisons were done between placebo and HAART-treated groups infected for the same time period. A 10 6 fold decrease in pVL was detected at D14 p.i. in H14 animals as compared to P14 animals whereas treatment initiated at D7 and D14 p.i. only led to 10 3 fold (H21) and 10-fold (H28) decreases in pVL (Figure 3.B). Despite its efficiency in decreasing pVL in H14 animals, CD4 Tcell (Figure 3.C) and CD20 cell counts (Figure 3.D) were less than 10% higher in these animals than relevant placebo-treated animals. CD4+ T-cell and B-cell counts were correlated in HAART-treated animals (Rho = 0.925, p,0.0005).
The percentage of total blood B-cells was 20% and 10% higher in H14 and H21/28 animals respectively, than in placebo-treated animals (P14, 21 or 28). These differences were not statistically significant ( Figure 3.E, left panel). However, percentages of memory and MGZ blood B-cells were partially restored at D21 and D28p.i.. In H28 animals, percentages of naive, MGZ and memory B-cells almost reached baseline values ( Figure 3.E right panel). Only the difference between percentages of MGZ B-cells between P28 and H28 was statistically significant (p,0.05). The SIV-mediated increase in total spleen B-cells (Figure 3.F, left panel) was less marked in animals receiving HAART than in placebo-treated animals but loss of memory and MGZ B-cells was still prominent at D21 and D28 p.i. (Figure 3.F, right panel). The differences between percentages of total B-cells in placeboand HAART-treated groups were not statistically significant but those between percentages of naive and memory B-cells on D21/ 28p.i. and of MGZ B-cells on D14p.i. were statistically significant (p,0.05).
Although HAART initiated at 4 h p.i. abrogated both spontaneous and BCR-induced apoptosis (Figure 2.B), the impairment in BCR-and CD40-mediated proliferation was only partially counteracted (Figure 2.A). Early HAART had no effect on the plasma IgG/M increase detected before D14p.i. (Figure 2.C), but led to a 50% decrease in IL2+IL10-induced IgG production in vitro (data not shown). HAART potently reduced SIV-specific Ab and total IgM production when initiated on D7 p.i. but was less efficient when initiated on D14p.i. (Figure 2.C).  Table S2). The number of follicles and the size of GC were decreased in MLN on D14p.i. whereas the sizes of both follicles and GC were increased from D21 p.i.. (Figure 4.A). Although the numbers of spleen follicles were similar in non-infected and SIV-infected animals during the 28 days of infection, their size rapidly increased, peaking at D21p.i. (1.6-and 1.3-fold increase observed at D21 and D28p.i., respectively). The spleen GC size was 1.5 times greater at D14p.i.. and 3.6 times greater at D28p.i. than at baseline. (Figure 4.B). Changes in size of the spleen MGZ were analyzed after staining with CD20 ( Figure 4 Whereas B-cell follicles were generally rare in the duodenum of non-infected animals, increasing numbers of B-cell follicles accumulated in this area in SIV-infected animals from D14p.i. (Figure 4.D). The size of these follicles also progressively increased and was frequently associated with well-developed GC from D28p.i.. Thus, trafficking to intestinal mucosa could transiently decrease the B-cell numbers in MLN before D14p.i. but this loss could be balanced by homeostatic mechanisms thereafter.

Loss in follicular helper T-cells and infiltration of granzyme B + and CD8 + cells in GC during acute SIV infection
According to the loss in CD4 + T-cells during primary SIV/ HIV-1 infection [29,30] and the infiltration of cytotoxic CD8+ Tcells within the GC of various lymphoid organs in chronically infected individuals [12][13][14], we stained sections from spleen (data not shown) and MLN on D28 p.i. (Figure 5) for CD3, CD4, CD8 and CD45RO. We observed less intense staining with CD3 and CD4 mAb in the extrafollicular zone (EFZ) of placebo-treated animals than of non-infected animals, suggesting that there were fewer, positive cells. CD4 and CD3 staining on sections from HAART-treated animals appeared similar to that observed in non-infected animals. Due to their high number and low intensity staining, the number of CD4 T-cells present in the GC of noninfected and SIV-infected animals could not be quantified. However, we found that the number of CD45R0-positive cells in GC was 1. Although an increase in the number of CD8 + cells in the EFZ could be detected from D14p.i., infiltrating CD8 + cells were only detectable in GC from D21p.i. (Figure 5.D). The numbers of CD8 + cells were four (D21p.i.) and twenty (D28p.i.) times higher in the spleen of infected animals and five (D21p.i.) and 11 (D28p.i.) times higher in MLN ( Table 3). We then examined their potential cytolytic activity by granzyme B (GrB) staining. This revealed increased numbers of positive cells in MLN and spleen from D14p.i., not only in EFZ but also in GC. In spleen and MLN, GC contained two to three times more GrB+ cells in placebo-treated animals than in control animals at D14 p.i. and about four times more at D28 p.i. (Table 3). However, the number of CD8+ cells was four times higher than that of GrB+ cells at D28p.i.. The mean number of CD8+ and GrB+ cells strongly decreased following HAART. The number of CD68 cells present in the EFZ zone and GC increased by a factor of two in spleen (24611.

Changes in GC organization and plasma cells during acute SIV infection
As described above, the sizes of total (CD20 staining) and GC (staining with Bcl6) B cell-populated areas were already increased on D14p.i. and even more at D28p.i. in spleen ( Figure 6) and MLN (supplemental Figure S3). In contrast, the mantle zone was thinner in SIV-infected animals than in controls on D28p.i.. Several other striking differences between control and SIVinfected animals were observed from D14p.i. on: (i) a strong increase in Ki67+ proliferating cells, with Ki67 staining delineating the dark zone (DZ) of GC in non-infected individuals; (ii) a strong increase of Ki67+ cells in EFZ; (iii) reorganization of the follicular dendritic cell (FDC) network, revealed by staining with CD23 or ICAM1 (data not shown). CD23 staining was distributed throughout the GC of control animals, but was frequently patchy, with preferential staining at one side of GC, probably in the light zone (LZ) in placebo-treated animals. These observations are indicative of an impaired expansion of the FDC network and LZ during early SIV infection. HAART significantly reduced these alterations (Figure 7 and supplemental Figure S2).
In the absence of appropriate CD138 and CD38 mAb for IHC analysis in cynomolgus macaques, we used IRF4, IgG and -IgA Ab to detect plasma cells. (Figure 6). An increase in the number of IRF4-positive cells was detected in GC and EFZ (including red pulp) in spleen on D14. and D28p.i. (Figure 6). From D14p.i.on, increased numbers of IgG, but not IgA, plasma cells were observed   (Figure 6) or MLN (data not shown). HAART significantly reduced these alterations (Figure 7 and supplemental Figure S2).

Discussion
Impaired production of neutralizing HIV-specific antibodies during natural exposure to virus or during vaccination protocols is one major hallmark of HIV infection [15,31]. Even long-term HAART is only partially efficient in preserving the humoral response in HIV + patients, suggesting that alternative immunemodulators are required. Given that the nature of the events taking place during primary infection is critical for initiation of antiviral responses and predictive of long-term disease progression, a better understanding of the early pathogenic effects of HIV on Bcells is needed. Since early time points of acute HIV infections are very difficult to study in humans, we specifically examined changes in B-cell subsets from several lymphoid organs during the acute phase of the disease in SIV mac251 -infected cynomolgus macaques. Our data reveal a transient peripheral B-cell lymphopenia (count and percentage) with a nadir at D14p.i., coinciding with the plasma viremia peak and the nadir in NK and CD4 + or CD8 + Tcells. Similar panleukopenia is observed during the acute phase of most viral infections in primates and mice.
In contrast to blood, the percentage of total B-cells increased in spleen from D14p.i. despite few changes in the percentage of Tcells in spleen before D21p.i. (data not shown). During the same time, the percentage of total B-cells did not vary in peripheral LN (axillary and inguinal LN) whereas a consistent influx of B-cells into the small intestine occurs from D14p.i. with the number and size of follicles progressively increased between D14 and D28p.i.. Altogether these data suggest that blood B-cells traffick to lymphoid organs but preferentially accumulate in the spleen and the small intestine. A drop in blood B-cells associated with a selective tissue homing has previously been reported in mice acutely infected by various pathogens [32][33][34]. Local sequestration of B-cells in these mice correlated with type I IFN-induced upregulation of CD69. In these murine infection models, membrane CD69 interferes with surface expression of sphingosine 1phosphate receptor 1 and blocks B-cell egress from particular lymphoid organs [34,35]. Consistent with the B-cell hyperresponsiveness to CXCL13 observed in HIV-infected patients [9], a similar mechanism may underlie B-cell accumulation in spleen follicles and MGZ during acute SIV infection. Indeed, early peaks of type I IFN and TNFa are detectable in plasma during HIV/SIV infection, before or at the same time as the viremia peak [24,[36][37][38][39] Regardless of changes in total B-cells, a selective loss in CD27 + memory B-cells occurred in blood, LN and spleen. This loss was also demonstrated by decreases in CD80-, 86-and 95-expressing B-cells in these three organs. The decrease in circulating memory B-cells was transient, correlating with that of total B-cells, and occurred without any change in the ratio of SIgM + /SIgG + memory B-cells. This suggests that this decrease mainly resulted from B-cell trafficking to lymphoid organs, where a B-cell response is initiated. The relative percentage and number of memory B-cells increased in blood from D21p.i., suggesting that memory B-cells are newly formed or are sent back into circulation from tissues. This coincides with reduced levels of inflammatory cytokines in plasma, observed in SIV-infected macaques after 2 weeks p.i. [36,40]. Given previous findings in primary HIV-infected patients [15], the blood B-cell repertoire in SIV-infected macaques must undergo significant changes at four weeks p.i. and probably also later on. This hypothesis remains to be verified in SIV-infected macaques previously vaccinated with recall antigens.
Memory B-cells in spleen or LN steady declined until D28p.i., correlating with a preferential depletion in SIgG-expressing memory B-cells. This cell loss peaked at D14p.i. in spleen but was more progressive in LN. This loss may have been initiated by SIV-induced B-cell apoptosis by D14p.i. and early commitment to terminal differentiation. Most of the IgG molecules produced in vitro were non-specific for SIV, with SIV Ab remaining undetectable in culture supernatants (data not shown). Given that HAART initiated 4 h p.i. fully prevented apoptosis but partially reduced in vitro IgG production, independent control mechanisms may be involved. Previous data obtained in SIV-or HIV-infected individuals suggest that apoptosis is induced by various virusinduced signals [41,42], whereas Ig production is rather promoted by increased cytokine production [43,44].
Plasma IgG and IgM levels were increased by D14p.i., whereas SIV-specific antibodies were detectable only from D21p.i. on. Therefore, we suggest that an early T-independent response takes place before D14p.i.. In humans, unlike in mice, MGZ contains both CD27 + SIgM + SIgD + MGZ and CD27 + SIgD2SIgG + memory B-cells poised to differentiate rapidly into plasma cells in response to T-independent signals, including BAFF, APRIL and IL21 [45]. Elevated BAFF plasma levels have only been reported in chronically HIV-infected patients with a low CD4 count [46,47]. Inflammatory cytokines and CD40 ligand are wellestablished inducers of BAFF production by monocytes, macrophages and dendritic cells [48] and, moreover, He et al have recently shown that gp120-induced production of BAFF by monocytes enhances the polyclonal IgG and IgA production by a subset of gp120-activated MGZ-like B-cells [19]. BAFF production may therefore be triggered during acute SIV-infection and contributes to the early, potentially T-independent increase in plasma IgG/M. Although the histological size of MGZ did not significantly vary, the frequency of (CD27 + SIgD + ) MGZ B-cells decreased transiently in LN between D14 and D21p.i. and steadily in spleen and blood from D14p.i. which could be due to their commitment to T-independent terminal differentiation. Similarly, a loss in IgM-only memory B-cells, associated withan impaired response to T-independent antigens has been reported in primary and chronically HIV-infected patients [15,17] whereas a splenic MGZ hypoplasia was reported only in chronically HIV-infected Percentages of blood B-cells before infection (E, pre, black square, n = 13), in non-infected animals (black square in F, n = 2 and G, n = 3), in placebo-(black circle, n = 3) and HAART-treated (open circles, n = 3) SIV-infected animals. Mean percentage6SEM is indicated for each group. Non significant p values between placebo and HAART-treated groups are indicated. Right panels. The relative percentages of naive (SIgD+ CD272, open area), memory (SIgD2 CD27+, hatched area) and MGZ (SIgD+ CD27+, black area) B-cells were determined in controls (Ctl, 13 blood samples before exposure to SIV) for blood (E) or three non-infected animals for spleen (F).Three placebo-(P14, P21 and P28) or HAART-treated (H14, H21 and H28) SIV-infected animals were similarly investigated. Mean percentage6SEM is indicated for each group. ND: Not Done doi:10.1371/journal.pone.0005966.g003  patients [12]. An increase in IgG-expressing plasma cells was observed in the EFZ of spleen and LN from D14p.i.. The fact that HAART had a moderate effect on increased plasma IgG/M by D14p.i. suggests that this increase was independent of viral replication per se but is correlated with SIV-induced inflammatory signals, including BAFF as also suggested by the work of He et al [19] IgG-expressing plasma cells were only detectable in GC from D21p.i.. The SIV-specific response was therefore likely to have been initiated between D14 and D21p.i.. Initiation of the specific response is likely to be delayed until viral antigens are locally available in sufficient amounts. Consistent with a recent study by Cinamom, showing that MGZ B-cells transport immune complexes into the GC [11,49], shuttling of MGZ B-cells to follicles could enhance the SIV-specific Ab response. According to recent data in HIV-infected patients, hyper-responsiveness to CXCL13 might favour the homing of MGZ B-cells into GC [9]. The marked and concomitant decrease in SIV-specific Ab and plasma IgM observed after D14p.i. in HAART-treated animals is consistent with SIV-specific Ab being essentially IgM until D28p.i.. Although IgM in nature, these early Ab would be protective and probably play an essential role in generating immune complexes to be transported into GC [49]. Viral particles, through interactions between trimeric gp120 and B-cells, could trigger a T-independent, SIV non-specific Ab response before D14p.i., predominantly targeting gp120-associated carbohydrates. As viral replication progresses, a T-dependent response might occur. Similar conclusions have been drawn by Trkola et al. in chronically infected patients subjected to short-term interruption of treatment [50]. Alternatively, soluble Nef present in GC of HIV-infected patients might inhibit CD40-dependent Ig-switching of GC B-cells [51]. The immunosuppressive effect of Nef on the Bcell response to SIV was previously suggested by the work of Chakrabarti et al, who showed that Nef-deleted SIV strain induced a more rapid development of GC and circulating SIVspecific Ab than pathogenic SIV strain [52]. Therefore, soluble Nef is likely to contribute to the delayed SIV-specfic Ab response observed during the acute phase of infection.
GC hyperplasia was detected in spleen and MLN of SIVinfected animals from D14 and D21p.i., respectively. The less intense staining observed in spleen and MLN suggests that fewer CD4+ T-cells may be present in EFZ of infected animals than in non-infected ones. However, the distribution of CD4+ cells in GC of infected animals was similar to those observed in non-infected animals before D28p.i.. This contrasts with the impaired polarization of GC observed in chronically infected patients [13]. A decrease in CD45R0 cells, likely CD4 follicular helper Tcells, was detected within GC on D28p.i.. Consistent with their activated/memory phenotype, SIV infection might lead to their apoptosis in situ or prevent their homing to GC. In contrast, infiltrated GrB+ cells and CD8+ cells were present in GC from D14p.i. and D21p.i. respectively. Given previous findings in chronically infected macaques [53] or patients [13,54], it seems likely that these cells are cytolytic T-cells involved in virus control. Similarly, GC of SIV-infected animals contained more Ki67+ positive cells, demonstrating an expanded DZ, whereas follicular dendritic cells (FDC) had a patchy distribution localized to one edge of GC. The numbers of total plasma cells (Vs38c, IRF4) and IgG-expressing plasma cells increased in the red pulp, EFZ and GC of SIV-infected animals, whereas IgA-expressing cells were present in similar proportions in the red pulp of non-infected and SIV-infected animals. In spleen, most of these changes were fully or partially (plasma cells) cured by HAART, no matter when HAART was initiated. HAART did not prevent the decrease in GC size in MLN, observed at D14p.i., but cured the increase in GC and other changes observed at D28p.i.. This dual effect is mainly independent of virus replication and is probably sustained by inflammatory signals.
In conclusion, we have combined IHC studies of lymphoid organs with phenotypic and functional analyses of B-cells to further elucidate the mechanisms underlying B-cell dysfunction during acute SIV infection. Consistent with unresponsiveness to early HAART, we suggest that T-independent signals, possibly triggered by direct contact with virions and relayed by inflammatory cytokines, are important in early B-cell dysfunction-a major feature of the SIV/HIV-specific B-cell response strongly favoring non-specific Ig production. A better understanding of the inflammatory signals acting on B-cell subsets should help to design of new therapeutic strategies, potentially delaying the initiation of HAART which strongly decreases the HIV/SIV humoral response.

Animals, virus and treatment and plasma viral load
Thirty adult male cynomolgus macaques (Macaca fascicularis) were inoculated intravenously (i.v) with 50 AID50 of the pathogenic SIVmac251 strain, as previously described [23]. Animals were housed and cared for in accordance with European Guidelines for animal care. All protocols used in this study were reviewed and approved by an regional animal care and use committee. Three groups of five animals were killed on days (D) 14, 21 or 28 post-infection (p.i.) respectively; these were placebotreated groups P14, P21 and P28. The three other groups of five animals initiated a two-week HAART from 4 h, 7 or 14 days p.i.. HAART was a combination of 4.5 mg/kg 3-azido-2,3-dideoxythymidine (AZT), 2.5 mg/kg 2,3-dideoxy-3-thiacytidine (3TC), and 20 mg/kg indinavir given twice daily, via the oral route, as previously described [23]. These animals were sacrified at the end of HAARTand referred to as HAART-treated groups H14, H21 and H28, accordingly. All procedures with animals were performed after general anesthesia with ketamine chlorhydrate (Rhone-Merieux, Lyon, France). Plasma viral load was determined   by real-time PCR, as previously described [55]. Non-infected animals were used as controls.

Mononuclear cell counts and phenotypes
PBMC were obtained from peripheral blood on BD Vacutai-nerH CPT TM . Absolute numbers of leukocytes (cells/ml) were obtained by analysis of 15 ml of blood, collected on EDTA, in Micro Diff II (Beckman Coulter, Fullerton, USA). B-cells (CD20), monocytes (CD14), total T (CD3), CD4, CD8 cells and NK (CD32CD8+) cells were quantified by flow cytometry with 50 ml of titrated combinations of fluorochrome-conjugated mAbs added to 100 ml of whole blood, as previously described [55]. After incubation for 15 min at 20uC in the dark, cells were subjected to red blood-cell lysis and were then post-fixed with 300 ml of Cell Fix 16 (BD Biosciences) and kept at 4uC in the dark. Surface marker expression was analyzed with a LSR I flow cytometer and by FlowJo software (Tree Star Inc., Ashland, OR). Similar combinations of mAb were used on isolated lymph node mononuclear cells and splenocytes to determine the percentage of each lymphoid subset.

Immunophenotyping of B-cell subsets in blood, LN and spleen
Spleen, axillary and inguinal LN were collected from SIVinfected animals and four non-infected ones as controls. Splenocytes and LN mononuclear cells were obtained by mechanical disruption, passed through nylon mesh cell strainers with 40-mm pores and further purified by Ficoll gradient centrifugation (BD Biosciences, Franklin lakes, NJ). Freshly isolated cells were subjected to phenotypic analysis or kept at 280uC until use. Blood, spleen and LN B-cell subsets were identified by three-colour flow cytometry detecting the combined expression of CD20 with CD27, CD80, CD86, CD95, surface IgD (SIgD), SIgM or SIgG. Briefly, cells were stained with 50 ml of titrated combinations of fluorochrome-conjugated mAbs listed in supplemental Table S1. Data were analyzed as described above.

Immunohistochemistry (IHC) and quantitative image analysis
Paraffin-embedded tissues. IHC were performed on deparaffinized spleen, MLN or small intestine sections, with primary mAb (supplemental Table S2). Ab binding was visualized with the StreptABComplex/HRP duet kit and DAB (3,3 Diaminobenzidine) (Dako). Slides were counterstained with Mayer's hematoxylin and mounted in permanent mounting media (Dako).
Frozen tissues. MLN and spleen biopsies were frozen and 5mm cryostat sections were fixed in acetone, blocked in 1% BSA and stained with CD8 mAb (supplemental Table S2). Ab binding was visualized with StreptABComplex/HRP duet kit and DAB. Slides were counterstained with Mayer's hematoxylin and mounted in permanent mounting media (Dako).
Images from spleen and MLN sections were obtained with a light microscope Zeiss (Laboandco, Mandres-les-Roses, France), captured by a Microfire microscope camera system (Optronics, Goleta, CA) and analyzed with Mercator 4.42 software (Explor-aNova, La Rochelle, France).

Statistical analysis
Data are expressed as mean6SEM, unless otherwise indicated. Non-parametric analysis (Wilcoxon signed-rank test and Mann-Whitney test) and correlations (Spearman's rank coefficient) were performed using StatView software (SAS Institute, Cary, NC). p value#0.05 was considered as significant.  Figure S1 Correlation between plasma viral load and blood Tor B-cell counts during SIV infection. Correlation between plasma viral load (pVL, Log10copies/ml) and blood CD4 (A), CD8 (B) or CD20 (C) cell count (cells/ml) in SIV-infected animals is shown. Correlation between blood B-cell count and percentage (D) or CD4 cell count (E) in SIV-infected animals is shown. Statistical significance was assessed by Spearman's rank correlation test; Rho and p values are indicated.