Relative Impact of Complement Receptors CD21/35 (Cr2/1) on Scrapie Pathogenesis in Mice

Mammalian prion diseases are caused by prions, unique infectious agents composed primarily, if not solely, of a pathologic, misfolded form of a normal host protein, the cellular prion protein (PrPC). Prions replicate without a genetic blueprint, but rather contact PrPC and coerce it to misfold into more prions, which cause neurodegeneration akin to other protein-misfolding diseases like Alzheimer’s disease. A single gene produces two alternatively spliced mRNA transcripts that encode mouse complement receptors CD21/35, which promote efficient prion replication in the lymphoid system and eventual movement to the brain. Here we show that CD21/35 are high-affinity prion receptors, but mice expressing only CD21 die from prion disease sooner than CD35-expressing mice, which contain less prions early after infection and exhibit delayed terminal disease, likely due to their less organized splenic follicles. Thus, CD21 appears to be more important for defining splenic architecture that influences prion pathogenesis.

other. Human CD35 expression serves to clear immune complexes by providing a phagocytic signal for macrophages (15) and neutrophils (16). CD21 forms the B cell coreceptor (BCCR) with CD19 and CD81 on B cells that provides a costimulatory signal to reduce the activation threshold upon engagement of B cell receptor with its specific antigen (17)(18)(19). CD21/35 have been shown to facilitate prion infection administered peripherally (3)(4)(5), but the relative importance of each isoform remained unknown. Here we show that while both CD21 and CD35 bind PrP Sc with high affinity, mice lacking CD21 form PrP Sc -replicating follicular networks closer to proximal splenic nerves, replicate PrP Sc more slowly, and resist prion disease longer than mice lacking CD35. Follicular networks in CD21-and CD21/35-deficient mice appeared more fragmented and less organized and contained fewer Mfge8-positive follicular dendritic cells (FDCs) and/or tingible body macrophages (TBMs) than wild-type (WT) and CD35-deficient mice, which could explain impaired prion pathogenesis in these mice.

RESULTS
CD21 and CD35 both bind PrP Sc . Previous reports indicate a crucial role for CD21/35 in prion disease, but whether this role can be attributed to direct interactions of CD21/35 with PrP Sc remained unknown. To address this possibility, we performed a previously published protocol to remove soluble PrP C and highly enrich insoluble PrP Sc from infected hamster brains (20)(21)(22). We then tested whether PrP Sc interacts with purified full-length CD21 or truncated CD21 containing either SCRs 1 and 2 or SCRs 1 to 6. All of these forms of CD21 bound to PrP Sc with similar binding kinetics ( Fig. 2A). Monoclonal antibody (MAb) 171, which binds and prevents CD21 binding to its endogenous ligands at SCRs 1 and 2 (23-25), inhibited, but did not eliminate, the interaction between CD21 and PrP Sc (Fig. 2B), the dissociation constant (K d ) of which we FIG 2 CD21 and CD35 both bind PrP Sc . PrP Sc was enriched from elk brain infected with chronic wasting disease as previously described (21,22). (A) Equimolar full-length CD21 or protein containing the first two or first six SCRs on CD21 bound PrP Sc . (B) Addition of inhibitory monoclonal antibody 171 (iMAb 171) partially reduced the interaction. (C) Kinetic analysis revealed a K d of 16 nM between full-length CD21 and PrP Sc . Colored traces denote nanomolar ligand concentrations. estimate to be 16 nM (Fig. 2C). These data suggest CD21/35 partially bind PrP Sc at its first two SCRs, but may also bind PrP Sc at additional SCRs. Alternatively, the prion binding site on CD21 may overlap the MAb 171 binding site, but CD21 binds PrP Sc with higher affinity than MAb 171.
CD21 influences early splenic PrP Sc accumulation and terminal prion disease onset. We previously showed that CD21/35 expedited early splenic prion accumulation, neuroinvasion, and terminal prion disease (3,5). To determine if one isoform preferentially facilitates this early pathogenesis, we inoculated mice with RML5 prions and assessed PrP Sc loads in spleen at 30 days postinfection (dpi). Mice deficient in CD21 accumulated less PrP Sc in their spleen than CD35-deficient mice ( Fig. 5; P ϭ 0.0459). We also monitored time to terminal prion disease in these mice to ascertain whether splice variant CD35 or CD21 promotes disease onset. We found that CD21-deficient mice resisted disease significantly longer than wild-type, CD35-deficient, or hemizygous mice ( Fig. 6; Table 1), although the levels of PrP Sc accumulation in their brains at terminal disease appeared similar (Fig. 6B). We detected no difference by fluorescence-activated cell sorting (FACS) in the number of PrP C -positive brain cells or splenocytes or the amount of PrP C those cells express to account for the differences in prion disease kinetics we observed (Fig. 6C).
CD21 alters CD19 expression and follicular networks. Previous reports indicate that lack of CD21/35 leads to increased expression of another member of the BCCR, CD19 (29,30). B cells from CD21/35-deficient mice express 33 to 49% more CD19 than those from CD21/35-sufficient mice. Additionally, changes in CD19 expression have been shown to impact prion disease. CD19-deficient mice exhibit accelerated prion neuroinvasion, most likely due to their forming FDC networks that replicate PrP Sc in spleens closer to adjacent nerve fibers (31). To determine whether CD21/35 could be indirectly affecting prion pathogenesis by altering CD19 expression and/or function, we first investigated CD19 expression on CD21-and CD35-deficient mice. Flow cytometry revealed a significant 38% increase of CD19 on B cells from CD21-deficient mice compared to wild-type or CD35-deficient mice (Fig. 7). Since CD19's absence results in FDC networks closer to proximal nerves, we investigated whether CD19 upregulation positions FDC networks farther from them. We found that, like CD19-deficient mice, CD21-and CD21/35-deficient mice also contain splenic follicles positioned closer to proximal nerves than wild-type or CD35-deficient mice ( Fig. 8A and B). Thus, despite reduced neurofollicular distances in their spleens, CD21-deficient mice still survived prion infection longer than wild-type or CD35-deficient mice, coinciding with the reduced splenic PrP Sc replication we observed early after infection. Efficient prion replication in lymphoid follicles requires intact, organized follicular networks where PrP C and CD21/35-expressing FDCs and B cells can efficiently capture and replicate PrP Sc . We observed more fragmented, less organized follicular networks composed of significantly less FDCs and TBMs in spleens from CD21-and CD21/35-deficient mice than CD35-deficient and wild-type mice ( Fig. 8A and C), which could explain their differential prion replication. We observed little or no differences in B cell architecture and PrP C expression, save slightly increased marginal zone B cells in CD21/35 Ϫ/Ϫ spleens (Fig. 9) as previously described (30).

DISCUSSION
Attempts to ascertain CD21/35 functions previously relied on Cr2 genetic manipulations that render mice deficient in both CD21 and CD35. However, Donius et al. (10,11) generated mice that express one splice variant or the other, allowing researchers to assess the function and relative importance of each splice variant to cells that express them. For example, CD35 on macrophages and neutrophils is generally thought to promote phagocytosis and immune complex clearance, whereas CD21 on follicular dendritic (FDC) and B cells is thought to enhance antigen presentation and adaptive immune responses, respectively. These mice offer the ability to challenge or confirm these proposed roles.
The role of complement receptors in prion disease is well established. Mice deficient Here we show that both variants derived from Cr2 transcripts can biochemically interact with PrP Sc (Fig. 2). This finding supports the previous in vivo data and suggests CD21/35 are cell-surface prion receptors. Interestingly, in the mouse model of chronic wasting disease, C3-deficient mice eventually succumbed to terminal prion disease, whereas CD21/35-deficient mice did not (5,6). Our current data support the idea that CD21/35 impacts prion disease by directly interacting with PrP Sc independent of its endogenous complement ligands. These data suggest that CD21/35 binds PrP Sc in a less-site-specific manner than its endogenous ligands, but do not exclude the possibility of CD21/35 binding to C3 and C4 cleavage products bound PrP Sc . CD21/35 likely exerts its effects along with C3/4 opsonization, although perhaps not crucially, nor do our current data exclude the remote possibility that other proteins mediate this interaction.
We also confirmed CD21/35 as prion receptors and that soluble CD21 (sCD21) can act in a dominant-negative fashion to inhibit prion replication and infection of N2a cells in cell culture. Cells infected with prions pretreated with CD21 appeared to contain less total PrP signal (without protease K [PK] digestion) than PBS-treated controls. Prion replication likely increased the PrP Sc signal that contributed to the higher total PrP signal in PBS-treated samples, but we cannot definitively discount the possibility that CD21 caused a reduction of PrP C . In either case, these data suggest sCD21 could serve as a therapeutic, and we are currently exploring the use of sCD21 or soluble SCRs as small, prion-inhibitory molecules in vivo.
CD21/35-expressing B cells and FDCs are known to impact prion disease. B cells likely promote prion trafficking more so than replication, because deletion of prnp specifically in B cells did not inhibit prion disease (32,33). Thus, B cells promote prion disease independent of PrP C expression, and the data presented here support that hypothesis. We have previously shown that optimal prion trafficking to, capture by, and replication in lymphoid follicles require CD21/35 expression on FDCs and B cells (3,5). We propose CD21 acts as the prion receptor on B cells that mediates these facets of prion pathogenesis.
FDCs previously have been shown to be perhaps the most important immune cell type for peripheral prion pathogenesis. They express the most PrP C of any immune cell, and ablating them or manipulating follicular networks impacts peripherally initiated  prion disease more than any other cell type (34,35). However, B cells help orchestrate follicular development by providing lymphotoxin ␣ and ␤ as maturation and maintenance signals to FDCs (35,36). Loss of the BCCR member CD19 alters follicular development, moving FDC networks closer to proximal nerves and expediting neuroinvasion and terminal prion disease (31). We observed the same phenomenon in CD21and CD21/35-deficient and, to a lesser extent, CD35-deficient spleens. However, despite reduced neurofollicular distances in CD21-deficient mice, they still survived prion disease significantly longer than wild-type and CD35-deficient mice. CD21-deficient mice express full-length CD35 and a slightly truncated form missing SCRs 5 and 6 (11), which may not bind prions as efficiently. However, CD35-deficient mice also lack these and another four SCRs, and our SPR and coimmunoprecipitation data showing prions binding multiple SCRs argue against this explanation. CD21-and CD35-deficient mice develop normal splenic architecture, express normal populations and proportions of B cells, and express normal levels of PrP C . CD21/35 mice do express slightly expanded marginal zone B cells (30), but one would expect that expansion to result in enhanced prion replication, if any effect at all. We did observe fragmented, less organized follicles that contained less Mfge8-positive FDCs and TBMs in spleens from CD21-and CD21/35-deficient mice that we conclude impaired early prion capture, replication, and eventual neuroinvasion. CD21 forms the BCCR with CD19 and CD81 and appears to be the more important isoform for proper follicular development. Along with altered FDC networks observed in CD19-deficient mice, these data strongly promote the BCCR as an important signaling complex in developing normal follicular architecture. We conclude that proper follicular development and organization require CD21, leading to more efficient lymphoid prion replication and expedited neuroinvasion in CD21-expressing mice than in mice expressing the CD35 isoform. Known CD35 functions include phagocytosis and clearance of opsonized pathogens and immune complexes on neutrophils and macrophages (15,16). CD35 can also biochemically interact with PrP Sc , but likely promotes its phagocytosis by and destruction and/or sequestration within macrophages, the only cell type known to help resolve prion infection (37). CD21/35 likely act as PrP Sc receptors on FDCs, which act as efficient PrP Sc bioreactors in the lymphoreticular system. CD21, however, engages C3d/g opsonized pathogens and provides a costimulatory signal when a mature, naive B cell encounters its specific antigen. Therefore, in conjunction with previous reports, our ) and antityrosine hydroxylase (anti-TH) Ab to identify nerve fibers (red). Proximal nerves appeared closer to follicular networks in spleens from CD21-, CD35-, and CD21/35-deficient mice than wild-type C57BL/6 mice. Follicular networks appeared more fragmented and less organized, with fewer Mfge8-positive FDCs and TBMs forming networks in CD21-and CD21/35-deficient spleens than in CD35-deficient or wild-type spleens. (B) Mean neurofollicular distances were shortened in the absence of CD21 (n ϭ 95), CD35 (n ϭ 92), or CD21/35 (n ϭ 94), with no statistically significant difference between distances observed in CD21-and CD21/35-deficient spleens. Distances measured in C57BL/6 spleens (n ϭ 57) were significantly different from those of all other groups. ****, P Ͻ 0.0001; **, P Ͻ 0.01. (C) CD21 (n ϭ 54)-and CD21/35 (n ϭ 52)-deficient splenic follicles contained significantly fewer Mfge8-positive FDCs and TBMs than CD35-deficient (n ϭ 50) and C57BL/6 (n ϭ 54) follicles. **, P Ͻ 0.01, and ***, P Ͻ 0.0001, compared to CD21-and CD21/35-deficient follicles. data clarify the role of B cells in prion disease, including promotion of optimal follicular development mediated by the BCCR and interaction of CD21/35 with PrP Sc .

MATERIALS AND METHODS
Mice. All mice were bred and maintained at Lab Animal Resources, accredited by the Association for Assessment and Accreditation of Lab Animal Care International and approved on 14 January 2016 by the Institutional Animal Care and Use Committee at Colorado State University (protocol ID 09-1580A). CD21 (CR2)-or CD35 (CR1)-specific knockout mice on the C57BL/6 background were generated and characterized as previously described (10,11). We crossed the individual knockout mice to achieve hemizygosity.
Clinical sign scoring. Mice were monitored daily and sacrificed at the onset of terminal disease or specified time points. We employed a scoring system to assess the severity of disease, including: tail rigidity (0 to 2), akinesia (0 to 4), ataxia (0 to 4), tremors (0 to 4), and weight loss (0 to 2). Mice that scored above 10 were euthanized via CO 2 inhalation, replacing 20% of air per minute to effect.
Tissue collection and analysis. After euthanasia, the following samples were collected and frozen or fixed in 4% formaldehyde in PBS: serum, spleen (half fixed, half frozen), kidneys (one fixed, one frozen), tail clip, and brain (half fixed, half frozen). We assessed the presence of PrP Sc in 10% (wt/vol) homogenate after protease K (Roche) digestion (10 &micro;g/ml for spleen and 50 &micro;g/ml for brain) and Western blotting using anti-PrP monoclonal antibody (Ab) BAR 224 (Cayman Chemical) conjugated to horseradish peroxidase (HRP). Blots were developed using chemiluminescent substrates hydrogen peroxide and luminol for 5 min at room temperature and visualized using a GE digital imager and ImageQuant software. Tissues negative for PrP Sc on Western blots were subjected to serial protein misfolding cyclic amplification (PMCA [39]). Briefly, we used 10% normal brain homogenate (NBH) in PMCA buffer (PBS, 1% Triton X-100, 4 mM EDTA, 150 mM NaCl) from PrP C -overexpressing transgenic mice of strain Tga20 (40) as the substrate for amplification of previously undetectable PrP Sc . Twenty-five microliters of NBH and 25 &micro;l 10% sample homogenate were sonicated for 40 s at~150 W, followed by a 30-min incubation, which was repeated for 24 h (one round). Serial rounds were performed similarly, transferring 25 &micro;l of the previous round's sample to 25 &micro;l of fresh NBH. Each biological sample was run in at least technical duplicates, and round to positivity was determined by protease K (PK) digestion and Western blotting. Relative PMCA units were assigned as previously described (41). l/min. Excess activated groups were deactivated with 1 M ethanolamine HCl (pH 8.5) for 7 min at 10 &micro;l/min. Reference flow cells, built-in negative controls for this system, underwent rounds of activation and deactivation without protein ligand.
Cell culture and in vitro prion infection. N2a mouse neuroblastoma cells were grown in RPMI 1640 medium containing 10% fetal bovine serum and 1% penicillin-streptomycin. RML5-infected brain homogenate was UV sterilized prior to infection of cells. RML5 was preincubated with PBS or CD21 (SCR1 and -2; 5 &micro;g/ml final concentration) for 10 to 20 min prior to infecting cells. N2a cells were seeded at 100,000 cells per well in a 12-well plate and infected with 0.3% RML5. Cells were grown at 37°C and 5% CO 2 for 4 days. Wells were rinsed 2ϫ in PBS, and cells were detached from the plate using 5 mM EDTA in PBS for 10 min. Cell pellets were resuspended in 100 &micro;l PMCA buffer containing 1% Triton X-100 and lysed on ice for 30 min. Lysates were assessed for PrP Sc using traditional PK digestion and Western blotting techniques.
FACS. Fresh brain and spleens were harvested and processed to single-cell suspensions in 3 ml PBS. Briefly, tissues were passed through 40-&micro;m-pore mesh filters using sterile plungers and cold PBS washes. Cells were pelleted at 250 ϫ g for 5 min at 4°C, and the supernatant was discarded. Red blood cells (RBCs) were lysed in ACK (ammonium-chloride-potassium) buffer for 5 min, and the remaining cells were pelleted and washed with FACS buffer (PBS, 1% fetal bovine serum [FBS], 150 mM EDTA). Primary splenocytes or brain suspensions were blocked in 7% mouse serum and 1:50 Fc block (BD Pharmingen) for 20 to 60 min on ice. Ab solutions (1:100 final) were added to cells and incubated in the dark for 1 h on ice. Cells were washed by adding 1 ml FACS buffer to existing Ab solution. Cells were pelleted and resuspended in 1 ml FACS buffer for a total of 3 washes. A 792-&micro;l cell suspension was added to 8 &micro;l of 100 &micro;g/ml propidium iodide (Sigma) immediately prior to data acquisition in a Cyan flow cytometer. Unstained samples were analyzed first to set detector voltage and gating parameters to place the mean fluorescent intensity of at least 99% of unstained cells in a well-defined peak in the first decade of a log scale. Mean fluorescent intensity (MFI) signals beyond this decade were called positive. MFI as well as frequency of parent gate values were imported into Excel and/or GraphPad Prism for analysis.
Immunofluorescent histology and morphometry of spleen sections. Spleens were removed from mice of each genotype and flash frozen in OCT medium in liquid nitrogen. Five-micrometer sections were cut onto glass slides, fixed in ice-cold acetone for 10 min, air dried overnight, and incubated in 1:50 dilution of Fc block (BD Pharmingen) and 10% rat serum in Ultra Cruz blocking reagent (Santa Cruz) for 1 h at room temperature. Samples were then incubated in anti-mouse tyrosine hydroxylase (TH) Ab, rinsed three times for 5 min with PBS, followed by incubation with CruzFluor 555 (CF555)-conjugated mouse IgG binding protein (Santa Cruz) to stain splenic nerves. Slides were then rinsed and blocked again and then incubated with anti-mouse Mfge8 Ab followed by CF488-cojugated mouse IgG binding protein (Santa Cruz) to stain follicular dendritic cells (FDCs) and tingible body macrophages (TBMs). B cells were stained with Alexa 488-conjugated anti-B220 Ab (Pharmingen). PrP C was detected using Alexa 650-conjugated BAR 224. Slides were rinsed again and coverslips mounted with ProLong Gold antifade mounting medium (Life Technologies, Inc.). Splenic follicles were visualized with an Olympus BX-60 fluorescence microscope, and images were captured using a DP-71 charge-coupled diode camera (Olympus). Neurofollicular distances were measured and cells counted in at least six nonconsecutive sections from two spleens from two mice of each genotype using a morphometric overlay module in GraphicConverter (Lemke Software).
Statistical analyses. All statistical analyses were performed using GraphPad Prism software. We used log rank tests to compare survival curves, Student's t test to compare PMCA scores, and one-way analyses of variance (ANOVA) for all other comparisons among genotypes. Data comparisons with P values of Ͻ0.05 were considered significantly different. Technical duplicates were averaged, and the mean value of each biological replicate was considered an n value of 1.