Antibody Production in Murine Polymicrobial Sepsis—Kinetics and Key Players

Although antigen-specific priming of antibody responses is impaired during sepsis, there is nevertheless a strong increase in IgM and IgG serum concentrations. Using colon ascendens stent peritonitis (CASP), a mouse model of polymicrobial abdominal sepsis, we observed substantial increases in IgM as well as IgG of all subclasses, starting at day 3 and peaking 2 weeks after sepsis induction. The dominant source of antibody-secreting cells was by far the spleen, with a minor contribution of the mesenteric lymph nodes. Remarkably, sepsis induction in splenectomized mice did not change the dynamics of the serum IgM/IgG reaction, indicating that the marginal zone B cells, which almost exclusively reside in the spleen, are dispensable in such a setting. Hence, in systemic bacterial infection, the function of the spleen as dominant niche of antibody-producing cells can be compensated by extra-splenic B cell populations as well as other lymphoid organs. Depletion of CD4+ T cells did not affect the IgM response, while it impaired IgG generation of all subclasses with the exception of IgG3. Taken together, our data demonstrate that the robust class-switched antibody response in sepsis encompasses both T cell-dependent and -independent components.


INTRODUCTION
Sepsis is still associated with astoundingly high morbidity and mortality despite improvements in intensive care (1)(2)(3)(4)(5). A systemic hyper-inflammatory phase (systemic inflammatory response syndrome, SIRS) is followed or accompanied by a compensatory anti-inflammatory response (compensatory anti-inflammatory response syndrome, CARS), with the risk of lethal (secondary) infections (6,7). During the initial hyper-inflammatory phase, 40-50% of the T and B cell populations as well as innate immune cells go into apoptosis (8). Antigen presentation and T cell proliferation are impaired in the subsequent hypo-inflammatory phase, with a concomitant increase in concentrations of stress-induced anti-inflammatory glucocorticoids. These aforementioned effects, together with a Th2 cytokine bias, impair an effective immune response against primary or secondary infections (9)(10)(11)(12)(13)(14)(15)(16). This explains the fact that mortality from sepsis mostly occurs during this later phase (17,18).
It is well-documented that the antigen-specific B cell response in sepsis is strongly reduced (19)(20)(21)(22). For example, Mohr et al. have shown an impaired primary B cell response against defined antigens (22). However, they have also observed an unspecific increase of serum IgM and IgG concentrations after cecal ligation and puncture, a commonly used mouse model of sepsis (22). However, details of the B cell response in sepsis that could explain that discrepancy are largely unknown (21,23).
During an antigen-driven T cell-dependent (TD) immunoreaction against protein antigens, follicular B cells, which belong to the group of B-2 cells, are activated via the B cell receptor. With the help of activated T cells, they start to differentiate and form germinal centers, where class switch and somatic hypermutation take place. By the end of this process, affinity-matured plasma cells have developed that continuously secrete antibodies (24).
On the other hand, microbial components, which are systemically disseminated during sepsis, can activate B cells in a T cell-independent (TI) manner. For instance, TI-2 antigens (e.g., polysaccharides) crosslink B cell receptors and initiate a strong and long-lasting antigen-specific primary response (25). TI-1 antigens (e.g., lipopolysaccharide, LPS and bacterial DNA, CpG) activate B cells independent from the B cell receptor via toll-like receptors (TLRs), thereby inducing proliferation and antibody secretion (26,27). In addition, TLR ligation itself can induce class switch recombination (28)(29)(30)(31).
Though all naive and memory B cells in the mouse constitutively express TLRs (32)(33)(34)(35), there are mainly two B cell subtypes, namely B-1 and marginal zone (MZ) B cells, which differentiate into antibody-secreting cells (ASC) soon after TLR-activation (34). Their antibody repertoire is restricted, polyreactive and lacking somatic hypermutation (36)(37)(38). These antibodies are produced to bridge the time gap until the adaptive response has sufficiently matured.
B-1 cells differ in their mode of activation, development, specificities and locations from follicular B cells. Their main reservoir are the pleural and peritoneal cavities, where they can be further subdivided based on their CD5 expression into B-1a (CD5+) and B-1b (CD5-) cells. In addition, they can be found in small proportions in all lymphoid organs and are prone to TI responses. They are selected during development based on a certain strength of self-binding. In strong contrast to follicular B cells, their BCR engagement does not lead to activation. They are able to switch to all IgG subclasses in vitro, whereas in vivo they produce natural antibodies mainly of the IgM, IgG3 and IgA isotype [reviewed extensively in (24,38,39)].
MZ B cells are located close to the marginal sinus in the murine spleen (40,41), where they have direct access to bloodborne antigens (42,43). Although they have the capacity to generate TD and TI responses (44)(45)(46), their main function is the TI response against blood pathogens. They differentiate very early into IgM-or IgG-secreting cells (43). Follicular B cells (or B-2 cells) react only moderately or weakly to TI-1 antigens (34,47,48), but are classically the main producers of T cell-dependent, class-switched and hypermutated antibodies, which are produced in response to an antigen-specific TD response. They are found in peripheral lymphoid organs but also in the peritoneal cavity (PC) (49,50).
In the present study, we set out to examine B cell reactions and antibody secretion in polymicrobial abdominal sepsis, with the aim of explaining disparities in research findings. For that purpose, we used two murine models for sepsis induction: (i) fecal-induced peritonitis (FIP): intraperitoneal (i.p.) injection of pooled cecal content of donor mice into recipient mice (51)(52)(53); (ii) colon ascendens stent peritonitis (CASP): continuous leakage of own gut content over a certain time, which mimics the clinical setting (54,55). Whether splenic follicular or MZ B cells have a key role in the humoral response in sepsis was examined by explanting the spleen parallel to sepsis induction. In addition, CD4+ T cells were depleted before sepsis induction to determine the portion of the T cell-dependent and -independent humoral response.

Colon Ascendens Stent Peritonitis (CASP)
Colon ascendens stent peritonitis (CASP) surgery was performed as described before (54,56). Briefly, mice were anesthetized with i.p. Ketamin (Ketanest, Parke-Davis GmbH, Berlin) and Xylazin (Rompun, Bayer Health Care, Leverkusen), 100/10 µg per g body weight, respectively. The abdomen was opened through a small incision and a 18G stent (Ohmeda AB, Helsingborg, Sweden) was implanted into their colon ascendens. After surgery, all animals were carefully monitored every 6 h (h) until the end of the experiment. Control animals received sham operations, without stent implantation. Animals were euthanized 10 or 14 days following CASP surgery.

Fecal-Induced Peritonitis (FIP)
Sepsis was induced by introducing feces into the peritoneum using the method described by Wang et al. (57). In brief, littermates were anesthetized and euthanized. Fecal content (FC) was collected by cutting the Ampulla ceci and squeezing out the content. FC was homogenized in PBS to a final concentration of 100 mg/mL. The recipients received 7.25 × 10 5 CFU i.p., whereas control animals were treated with PBS instead. At certain time points after sepsis induction (days 1, 3, 7, 14, 28, as well as at 12 weeks), animals were euthanized and the splenocytes isolated.

Splenectomy
Following anesthesia, a midline laparotomy was performed. The cranial-dorsal and caudal-ventral spleen blood vessels were ligated with Mariderm 7/0, after applying yasergil-clips, and cut. The spleen was subsequently explanted.

Antibody Assay
Mice were anesthetized, and blood was collected via the retrobulbar venous plexus using a microhematocrit capillary. Serum was collected after centrifugation of the coagulated blood at 16,000 × g for 10 min. Total serum IgM and IgG concentrations in murine serum were measured with the Milliplex R Mouse Immunoglobulin Isotyping Immunoassay

Enzyme Linked Immuno Spot Assay (ELISpot)
On assigned days, mice were euthanized under deep anesthesia and then spleen, mesenteric lymph nodes (MLN), femur and omentum were harvested for the preparation of single-cell suspensions. For spleen and MLN, 70 µm cell strainers (Sigma-Aldrich) were used. Bone marrow cells were prepared by flushing the femur with 10 mL cold PBS containing 5% fetal bovine serum (5% FBS/PBS). Cells were washed with cold 5% FBS/PBS (250 × g, 6 min, 4 • C), and erythrocytes were lysed with sterile filtered ammonium chloride-buffer followed by another washing step.
A single-cell suspension of omentum was prepared by collagenase and DNAseI digestion (Roche Diagnostics GmbH, Mannheim, Germany). Briefly, the omentum was washed with PBS containing 5 mM EDTA for 1 min to get rid of the attached cells, followed by washing in HBSS containing 10% FBS and 0.01 M HEPES. The omentum was then cut into small pieces with a sterile scissor and incubated in 500 µL digestion buffer (PBS, 10% FBS, 0.01 M HEPES, 1.5 mg/mL collagenase D, 2 mg/mL DNAseI) for 30 min at 37 • C with constant shaking (500 rpm). The resulting tissue was then mashed through a 70 µm cell strainer and washed twice in HBSS containing 10% FBS and 0.01 M HEPES (500 × g, 5 min, 4 • C). The last step was then repeated with a 30 µm cell strainer.
All cells were resuspended in cold culture media (RPMI1640 supplemented with 50 µM 2-mercaptoethanol, 100 U/mL penicillin/streptomycin, 2 mM glutamine, 1 mM sodium pyruvate, 0.2% D-glucose, and 1% non-essential amino acids). The numbers of DAPI-negative and CD45-positive cells were determined as described in the flow cytometry section. The numbers of IgM-and IgG-secreting cells were determined using a mouse IgM and IgG ELISpotPLUS kit (Mabtech AB, Nacka Strand, Sweden). ELISpot was performed according to the manufacturer's instructions for in vivo activated cells (no additional activation required). Cells, titrated to 5,000-50,000 per well, were seeded in triplicates and incubated at 37 • C for 16 h.
Spots were imaged using an ELISPOT plate reader (ImmunoSpot S5 Versa, Cellular Technology Limited) and

Flow Cytometry
B cells were characterized using specific antibodies listed in Table 1, together with the necessary isotype controls ( Table 2). B cell subpopulations were phenotypically defined according to the criteria listed in Table 3. Spleen cell suspensions were obtained as described before (62). Cell numbers were determined using BD TruCOUNT TM beads. One million cells were incubated with 2 µL Fc-Block for 15 min at 4 • C. Then, 50 µL of the appropriate antibody-cocktail was added and incubated for further 30 min at 4 • C. After washing (300 × g, 6 min) with FACS-buffer (BD FACSFlow Sheath Fluid, 2% FBS, 0.02% sodium azide), the pellet was resuspended in FACS-buffer and analyzed on a BD LSRII flow cytometer.

Depletion of CD4+ T Cells
For the depletion of CD4+ T cells, 150 µg rat anti-mouse CD4 mAb (GK1.5, in-house) was injected i.p. 1 and 3 days before CASP surgery. This efficiently depleted CD4+ cells without affecting CD8+ T lymphocytes (Supplementary Figure 2) as shown by FACS analysis using antibodies listed in Frontiers in Immunology | www.frontiersin.org  Table 3. Control mice received PBS instead. 14 days after depletion, the CD4 population in depleted, nonseptic control mice had recovered to 40% compared with the non-depleted controls (own unpublished data).

Statistical Analysis
Statistical analyses were performed using GraphPad Prism 6 for Windows (GraphPad Software, CA, USA). Data were assessed for significant differences using One-Way ANOVA with Bonferroni correction (Bonferroni post-hoc test) for selected pairs. P < 0.05 were considered significant.

The Spleen Is the Main Source of IgM-and IgG-Secreting Cells After Sepsis
Next, the source of the strong antibody reaction to sepsis was determined. Abdominal sepsis starts in the PC and is characterized by the systemic dissemination of pathogens and their products. Thus, both local and systemic immune responses are expected to take place. Locally, the parathymic lymph nodes are draining the PC (49,63). They increase in size after sepsis induction but still have a much lower cell count compared to the spleen, ruling out a major contribution to the serum Ig response. Furthermore, the omentum and its lymph follicle-like structures, the so-called milky spots, have been ascribed a role in lymphocyte migration to and from the PC (49,64,65), while the mesenteric lymph nodes (MLN), an accumulation of relatively large lymph nodes in the PC, drain the gut and are probably not directly involved in the immune cell migration to or out of the PC. On the other hand, a systemic immune reaction will take place in the spleen due to the hematogenous spread of microbial compounds. Finally, the bone marrow might be involved as a source of immature as well as memory B cells, and a niche of long-lived plasma cells (66)(67)(68). ELISpot analyses clearly revealed the highest amount of ASCs in the spleen. At the peak of the response, namely 10 days after sepsis induction, around 10 6 ASCs were counted in the spleens of septic animals. In addition, the MLNs seem to make a contribution to the antibody response, but the means differ by more than 20-fold (Figures 2A,B). In accordance with this, splenic follicular B cells, marginal zone B cells as well as germinal center B cells were rapidly activated in sepsis induced by FIP (within 24 h). The latter remained activated over a period of 12 weeks (Supplementary Figure 3).

Spleen Cells, Including Marginal Zone B Cells, Are Not Necessary for the Production of Antibodies After Sepsis
Although we detected B cell activation and germinal center formation in the spleen, together with the majority of ASCs, it turned out that this organ was superfluous with regard to the observed strong increase of immunoglobulins after sepsis. To determine the input of splenic B cells to the overall humoral response, we splenectomized mice in parallel to CASP induction. Fourteen days later, there were no major changes in IgM or IgG serum concentrations as compared to the animals that received only CASP (Figure 3). Moreover, the lack of spleen had no effect on the induced IgG subclasses in the septic immune response (Supplementary Figure 4). The ostensible IgG-increase following splenectomy and CASP compared to CASP-only is due to three animals whose IgG2b concentrations increased strongly (Figure 3 and Supplementary Figure 4).

The Antibody-Response After Sepsis Is Partially T Cell-Dependent
The generation of germinal centers as well as the strong increase in serum concentrations of all IgG subclasses makes a case for an antigen-driven TD Ig response in sepsis. Depleting CD4+ cells with an antibody (Gk1.5) prior to sepsis induction (Supplementary Figure 2) had no influence on IgM secretion (Figure 4A), but led to reduced serum IgG concentrations 14 days after CASP (Figure 4B). This supports the notion of a TD component in the B cell response. Interestingly, the decrease in serum IgG concentrations was absolute for IgG1 (returning to background levels), intermediate for IgG2b and IgG2c, but only in tendency for IgG3 (Figures 4C-F). Therefore, class switch in sepsis is evidently not exclusively dependent on T cells, but additionally driven by T cell-independent processes/antigens.

DISCUSSION
While battling invading pathogens, the systemic immune response causes collateral damage to the host, impairing life-securing homeostasis. Compensatory anti-inflammatory mechanisms and the necessary apoptotic loss of immune effector cells lead to immunosuppression, culminating in immunoparalysis (7,8,10,12,69,70). It was hence assumed that the antibody response would also be impaired (70,71). We have shown that serum IgM as well as IgG concentrations start to increase three to seven days after sepsis induction. The IgG increase was distributed among all IgG-subclasses, with the strongest relative increase observed for IgG3. IgG1, IgG2b, and IgG3 reached similar absolute serum concentrations of 1-2 mg/mL. The robust increase in serum Ig appears to be a general phenomenon in sepsis. Mohr et al. observed a similar increase 10 days after CLP (22), and our group obtained analogous results in a mouse model of FIP (data not shown). Brunner et al. have detected increased serum IgG concentrations in septic patients as early as 48 h after diagnosis (72).
An important role has been attributed to the spleen in the early defense against bacterial dissemination in the blood. Phagocytosis, endotoxin detoxification and antibody production are the main effector mechanisms. Especially MZ B cells carry TI responses (43,(73)(74)(75) and were deemed essential for successful pathogen eradication (42,76). In accordance with this, the spleen was the main source of ASCs in our sepsis model, with a minor contribution of the MLNs. Moreover, we have shown an early activation of follicular and MZ B cells in the spleen, corroborating the results of other research groups (77). In accordance with what was observed by Kelly-Scumpia et al. (77), germinal centers were FIGURE 3 | Serum IgM and IgG concentration 14 days after sepsis and splenectomy. Female C57BL/6 mice were CASP-operated and their spleen was explanted in parallel. Untreated, splenectomized-only and CASP-only animals served as controls. 14 days later animals were anesthetized and blood was collected. IgM and IgG serum concentrations were measured by Luminex ® -technology. Shown is the mean of the collective data from two independent experiments with a similar tendency. We used the One Way ANOVA and Bonferroni post-hoc test for selected pairs for statistical evaluation. * p < 0.05; N = 8-17 per group. FIGURE 4 | Serum Ig concentration 14 days after CASP and CD4+ cell depletion. Female C57BL/6 mice received 150 µg of a depleting anti-CD4 antibody (αCD4; Gk1.5) i.p. three and one day before sepsis induction via CASP. Control animals remained untreated or received the depleting antibody only. Fourteen days after CASP, animals were anesthetized and blood was collected. IgM (A) and IgG (B) serum concentrations, as well as IgG subclass concentrations (C-F), were measured by Luminex®-technology. Shown is the mean of the collective data from two independent experiments with a similar tendency. We used the One Way ANOVA and Bonferroni post-hoc test for selected pairs for statistical evaluation *p < 0.05. N = 6-10 per group.
formed after 3 days. Four days later, high numbers of IgM and IgG-secreting cells were detected, especially in the spleen.
However, splenectomy did not impair the humoral immune response, as measured by the increase in serum antibody concentration. It seemed that splenic follicular as well as MZ B cells, despite being rapidly activated, were redundant. This was surprising in light of previous reports of a strongly reduced humoral response against bacterial antigens in splenectomized mice (78)(79)(80)(81). In those studies, mice were infected 7-70 days following splenectomy. In yet another study, splenectomy led to a 75% reduction in B-1a cells after 6 days (60), which might explain the reduced humoral response in this setting. B-1a cells are known to participate in anti-bacterial and anti-viral responses (38,(82)(83)(84). We splenectomized the mice in parallel to CASP induction; hence extra-splenic B cell populations could still react to the multitude of antigens. Furthermore, in studies with Borrelia hermsii, MZ B cells did not play an important role either, because B-1 cells were the main producers of protective serum IgM (85,86).
The fact that splenectomy did not decrease the serum IgM and IgG concentrations, despite the fact that we have clearly disclosed the spleen as the main source of ASCs, strongly argues for the hypothesis that most ASCs found in the spleen developed from immigrating B cells. These were probably peritoneal B cells, which -upon activation -left the peritoneum and migrated to spleen and peripheral lymph nodes (49,86,87). Some of these are obviously able to class switch to IgG. Our data also suggest that soon after splenectomy other lymphoid organs, be it mesenterial lymph nodes, the milky spots in the omentum and/or the parathymic lymph nodes, can compensate for the spleen.
Several groups have shown that antigen-specific priming is impaired after sepsis (21,22,88). So the question remains, what drives the strong antibody-increase in murine serum after sepsis? Obviously, microbial structures, such as LPS or CpG, flood the host system during sepsis, and are able to polyclonally activate B cells via their appropriate receptors (TLR4 or TLR9) (32,89). Especially MZ B cells and B-1 cells differentiate into plasma cells (34,35). Interestingly, LPS per se can induce class switch to IgG2b and IgG3, but also activated NKT cells, activated DCs and thrombocytes, all of which are abundant in sepsis, can at least partially compensate a lack of T cell help and promote TI antibody class switch (90)(91)(92)(93).
On the other hand, the disseminating bacteria as well as dying host cells also confront the adaptive immune system with a wealth of antigens in the setting of sepsis. The pronounced germinal center reaction, in conjunction with the increase in all IgG subclasses support the idea that, besides polyclonal B cell activation, there may also be a significant antigen-driven component in the Ig response to sepsis.
Indeed, depletion of CD4+ T cells before sepsis reduced IgGproduction, leaving the IgM-response intact. Especially IgG1 did not increase over the basal level in the absence of T cells. Although IgG2b, IgG2c, and IgG3 production was also less than in T cell competent septic animals, there was still a measurable increase. That shows that the IgG response to sepsis comprises both TD and TI components. Nevertheless, the observed significant TD IgG response seems to contradict reports of impaired antigen priming following sepsis induction (19,21,22,88,94,95). The contrasting findings can, however, be reconciled by the following observations made by our research group (88): at sepsis onset, the T cell response to a primary antigen stimulus was not only fully intact but even enhanced, presumably through the adjuvant effects of the abundant pathogen-associated molecular patterns (PAMPs) and danger-associated molecular patterns (DAMPs) (88,96). However, later during the disease, the T cell response to antigen priming was reduced, with a nadir 7 days after sepsis onset, where T cells in severely affected animals did not react at all (19,88).
Our data reveal a strong humoral immune response in animals who survived sepsis. It is composed of T cell-dependent as well as T cell-independent components, takes place mainly in the spleen and probably involves the activation of all B cell populations. The task would now be to determine the antigen specificity of this Ig response in sepsis. This is addressed in the companion paper by Nicolai and co-workers (Nicolai et al., under revision).
In summary, in the present study, the origin of the strong antibody increase in sepsis was investigated, which identified the spleen as the main source of ASCs. Explanting the spleen parallel to sepsis induction revealed that both splenic follicular and MZ B cells are redundant in the humoral response to sepsis. Moreover, depletion of CD4+ T cells prior to sepsis induction highlighted the fact that both T cell-dependent and T cell-independent components govern the IgG response to sepsis.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

ETHICS STATEMENT
The animal study was reviewed and approved by Animal ethics committee of the local animal protection authority (LALLF, State Office for Agriculture, Food Safety and Fisheries Mecklenburg-Western Pomerania).

ACKNOWLEDGMENTS
We are grateful to Maria Ozsvar Kozma for excellent technical support. We acknowledge support for the Article Processing Charge from the DFG (German Research Foundation, 393148499) and the Open Access Publication Fund of the University of Greifswald.