C4d, a high-affinity LilrB2 ligand, is elevated in Alzheimer’s disease and mediates synapse pruning

Significance

A complex interplay of mechanisms and cell types regulates synapse stabilization and pruning throughout life. The complement cascade contributes by activating glia, leading to pruning. Neurons also express receptors implicated in pruning, including murine PirB and human LilrB2, a GWAS risk locus for AD. Unexpectedly, C4d, cleaved from complement C4, binds with high affinity to LilrB2/PirB. In the human cerebral cortex, these molecules colocalize at synapses. In vivo, C4d drives PirB-dependent synapse pruning on cortical pyramidal neurons. These observations link C4 processing fragments to coordinate synapse elimination both on glia and on neurons. Present attempts to treat cognitive loss in AD have targeted glial receptors; our observations imply that targeting neuronal pruning receptors such as PirB/LilrB2 may also be advantageous.

Abstract

Synapse pruning sculpts neural circuits throughout life. The human Leukocyte immunoglobulin-like receptor type B2 (LilrB2)/murine Paired immunoglobulin receptor B (PirB) receptors expressed in neurons and complement protein C4 have been separately implicated in pruning. Here, we report that C4d, a C4 cleavage product with unknown function, binds LilrB2/PirB with nanomolar affinity. C4d and LilrB2 colocalize at excitatory synapses in the human cerebral cortex as well as with beta amyloid in Alzheimer’s disease (AD). C4d, as well as C4, increase with age and more so in AD. To examine whether C4d-PirB interactions can drive pruning, dendritic spines—the postsynaptic structure of excitatory synapses—were monitored on L5 pyramidal neurons in the mouse cerebral cortex: A significant decrease in dendritic spine density occurred in WT with C4d exposure, but KO of PirB completely prevented this loss. Together, our findings reveal an unexpected physiological role for C4d in pruning and imply that different complement cascade components may collaborate to engage both neuronal and glial-specific effectors of synaptic pruning.

Synapse pruning—a process by which synapses are eliminated—is a key mechanism that remodels neurons and circuits throughout life and is perturbed in neurodegenerative and neurodevelopmental disorders such as Alzheimer’s disease (AD), autism and schizophrenia (1–9). Paired immunoglobulin-like receptor B (PirB) is a murine transmembrane receptor expressed in pyramidal neurons and located at synapses in the cerebral cortex and hippocampus that is required for synapse pruning and Hebbian synaptic weakening throughout life (10–12). The human homolog, LilrB2, has been identified as a risk locus for AD in several large GWAS metanalyses (13, 14). In mice, PirB function in neurons is sufficient for synapse pruning: Spine density and functional excitatory synapses are greater than normal following germline or conditional deletion of PirB exclusively from pyramidal neurons. Moreover, creating a mosaic cerebral cortex in which PirB is deleted only from a small subset of pyramidal neurons in an otherwise WT brain and body prevents developmentally regulated pruning only in those neurons lacking PirB (10, 15–17). In adult PirB knockout (PirBKO), or in WT mice after acute PirB blockade with a soluble decoy receptor, spine and synapse density on cortical pyramidal neurons are elevated, spines are more stable with longer lifetimes, and memory and motor performance in mice are enhanced (10, 11, 15, 17). PirB is also implicated in Alzheimer’s disease (AD): AD model mice lacking PirB do not experience cognitive impairment or loss of LTP despite high levels of beta amyloid (18).

Complement cascade components and their classical receptors are also thought to participate in synapse pruning. Early cascade components including C1q, C3, and C4 are present at synapses (19–21), and microglia and astrocytes express complement receptors CR1, CR2 and CR3. Complement components are elevated in aging and implicated in synaptic loss in Alzheimer’s disease (8, 9, 22–24). C1q and C4 are also thought to contribute to synapse elimination during early development (21, 25). GWAS studies have associated the C4 locus with Schizophrenia (19) and in mice overexpressing the human C4 isoform C4A there is excessive pruning (1, 26). However, C4 itself is not thought to mediate pruning directly. Rather, a cascade of components cleaved from C4 (C4b) are thought to activate C3, which then binds to microglial or astrocytic CR3 receptors, leading to synapse engulfment and/or pruning (3, 27, 28).

Here, we present evidence that these two seemingly separate mechanisms—PirB acting in neurons and complement cascade components acting via glia—converge at the synapse. We show that C4d, a terminal split product of complement C4 with unknown function, mediates synaptic pruning by binding with nanomolar affinity to neuronal PirB and its human ortholog LilrB2. Our findings are consistent with a physiological role for C4d-LilrB2/PirB interactions in synapse pruning, particularly when the complement cascade is upregulated in aging, inflammation or AD. Our observations also offer a mechanism that can unite different views of complement-driven synapse pruning, in which separate C4 split products may act in parallel, leading on the one hand to glial activation while also engaging a neuronal pruning cascade via a C4d-LilrB2/PirB interaction.

Results

LilrB2 and PirB Are High-Affinity Receptors for the C4d Split Product of C4.

In humans, there are five family members belonging to the Leukocyte immunoglobulin-like receptor type B (LilrB) family of receptors. In mice, there is only one homolog: PirB. LilrB2 in particular is considered a risk locus for AD (13, 14). Because LilrB2 (also referred to in NCBI as ILT4) protein is expressed in human frontal lobe, and PirB is required for synapse pruning (10–12, 15–17) and implicated in mouse models of AD (18), we examined the interaction of LilrB2 and its most closely related family members, LilrB1 and LilrB3 (29, 30), with C4 and its split products.

Upon complement activation, C4 is processed into C4b, exposing an internal thioester group which enables a small fraction of C4b to covalently bind nearby proteins or carbohydrates, while the majority of C4b remains soluble (Fig. 1A). Bound or soluble C4b then complexes with C2a to activate C3. Subsequent cleavage of C4b generates the terminal split product C4d, which can be membrane-bound or soluble (28, 31, 32). To determine whether any C4 split products might be ligands for LilrB2, C4, C4a, C4b, and C4d proteins were spotted onto nitrocellulose membranes and probed with sLilrB2, a soluble Fc-tagged form of LilrB2 containing the entire extracellular domain (18). sLilrB2 bound strongly to C4d, while binding of C4 or C4b was barely above background; no binding was observed to C4a or the BSA control (Fig. 1B). Likewise, sPirB, a soluble myc-tagged form of PirB consisting of its extracellular domain (18), bound C4d, and weakly C4 and C4b (Fig. 1B). These dot blots suggest that C4d interacts with LilrB2, as well as with PirB.

Fig. 1.

LilrB2 and mouse PirB are high-affinity receptors for C4d. (A) Diagram of complement component C4 cleavage to form C4d. Activated C1q cleaves the C4 alpha chain to form C4a and C4b; then Factor 1 cleaves the C4b alpha chain on either side of the C4d peptide to release C4d, which can be soluble or covalently linked to cell membrane proteins. (B) Dot blots showing binding of soluble LilrB2 (sLilrB2; Left) or soluble PirB (sPirB; Right) to C4 complement components spotted on nitrocellulose membranes. sLilrB2 is detected with an LilrB2 antibody; sPirB is detected with an anti-Myc antibody. sLilrB2 and sPirB are positive controls for binding of anti-Myc or anti-LilrB2 antibodies; BSA, negative control. (C) Confocal images of C4d binding to LilrB family members in vitro. HEK293 cells transfected with Myc-tagged LilrB1, LilrB2, LilrB3, or empty vector (control) were treated with 570 nM C4d, then fixed and immunostained with antibodies against C4d and Myc. C4d: Cyan; LilrB1, 2, or 3: red; DAPI: blue. (Scale bar, 30 µm.) (D) Amount of C4d binding to LilrB family members, given as a comparison of pixel overlap between C4d and LilrB1, 2, or 3 in confocal images, expressed in arbitrary units (AU). *P < 0.05, t test. (E) Dose-dependent saturable binding of C4d to LilrB2 in HEK293 cells. Data are mean ± SEM (n = 3 experimental replicates) of mean fluorescence intensity (MFI). Kd = 37 nM. (F) Dose-dependent saturable binding of C4d to PirB expressed in HEK293 cells. Kd = 56 nM. (G and H) Surface Plasmon Resonance Sensorgrams of C4d binding to LilrB2 (G) or LilrB1 (H), at 160 RU surface ligand density. C4d analyte concentrations: 64 nM, 32 nM, 16 nM, 8 nM, 4 nM, 2 nM, 1 nM, 0.5 nM, and 0.25 nM. Raw sensorgram data are red. Fitted sensorgram curves are black. Kd at 160 RU = 3 nM. (I) Confocal images of C4d immunostaining (cyan) of WT or PirBKO C4d-treated cortical neurons coimmunostained with Tuj (red). Merged image with C4d, Tuj, and DAPI (blue). (Scale bar, 40 µm). (J) Quantification of bound C4d onto neurons from WT (circles) or PirBKO (squares) as shown in (I). Data are means ± SEM (n = 3 experimental replicates) of mean fluorescence intensity (MFI). Kd = 380 nM for neuronal PirB; black dashed line (triangles) shows the difference between wild type (WT) and PirBKO (WT-PirBKO). The red dashed line represents best fit line from which Kd of PirB-dependent binding was determined (Materials and Methods). (K) Diagram of PirB and LilrB2 extracellular domain mutants. (L) Binding of soluble Fc-tagged LilrB2 and PirB domain mutants to C4d. Constructs were transfected into HEK293T cells, and supernatants containing soluble domain mutants were incubated with C4d and pulled down with Protein-A agarose beads. Pellets were then Western blotted with anti-C4d to detect C4d associating with domain mutants (C4d) or with anti-human IgG antibodies to detect associating LilrB2 or PirB domain mutants (Fc-tagged domain mutants). (M) Four experiments were performed for each set of mutants (except D1-D3 for LilrB2) and values for binding to full-length LilrB2 and PirB constructs were normalized to 1.0 and graphed.

To learn more about specificity of the LilrB2–C4d interaction, LilrB1, 2, and 3 were expressed in HEK293 cells and C4d binding was assessed by measuring the colocalization of C4d with LilrB1-3 in vitro. C4d bound to LilrB2 and LilrB3; little binding to LilrB1 was detected (Fig. 1 C and D). In the human prefrontal cortex (PFC), Western blot analysis of brain lysates revealed a clear LilrB2-specific band, while LilrB3 was not detectable (SI Appendix, Fig. S1A). The LilrB2 and LilrB3 antibodies showed little or no cross-reactivity when tested in Western blots of HEK293 cells expressing either LilrB2 or LilrB3, supporting the specificity of the LilrB2 signal detected in the human cortex (SI Appendix, Fig. S1B). Therefore, the LilrB2–C4d interaction was characterized further in subsequent experiments.

To determine the Kd of C4d binding, varying C4d concentrations were applied to LilrB2 expressing HEK293 cells. Binding was dose-dependent and saturable, with a Kd of 37 nM (Fig. 1E). Saturable binding was also seen for murine PirB (Kd of 56 nM) (Fig. 1F). To determine whether C4d can bind to endogenous PirB in neurons, cultures of either WT or PirBKO cortical neurons were treated with C4d (Fig. 1 I and J): C4d binding on neurons was dose-dependent and saturable with a Kd of 380 nM calculated from the nonlinear isotherm (Fig. 1J). Furthermore, C4d binding to PirBKO neurons was reduced, indicating that binding in these neuronal cultures is partially dependent on PirB. The persistence of residual C4d binding in PirBKO is consistent with additional putative C4d receptors known to be expressed in immune cells such as neuropilin (33, 34). Because neuropilin is also known to be expressed in cortical neurons (35), it is possible that this explains the residual binding present in PirBKO cultures.

To validate further the interaction between C4d and LilrB2 in isolation, surface plasmon resonance (SPR) analysis was performed with Fc-tagged LilrB2 captured on Protein A sensor chip surfaces using C4d analytes of increasing concentrations. Concentration-dependent binding profiles with clear resonance signals above background were observed (Fig. 1G), consistent with a specific ligand–analyte interaction. Indeed, the Kd of interaction is about 3 nM (SI Appendix, Fig. S2A), indicating an extremely high-affinity interaction at multiple capture densities (SI Appendix, Fig. S2 D, F, and H). As in the HEK293 cell experiments (Fig. 1 C and D) binding of LilrB1 to C4d was not detected (Fig. 1H and SI Appendix, Fig. S2 E, G, and I). Together, the cellular binding data and SPR results show that C4d binds to LilrB2 with high specificity and affinity.

C4d Binding Requires the D1D2 Extracellular Ig Domains of LilrB2/PirB.

To identify which specific Ig domains mediate the C4d interaction with LilrB2 or PirB, Fc-tagged extracellular domain mutant constructs were used (Fig. 1 K−M) (18). LilrB2 has four immunoglobulin-like extracellular domains, while mouse PirB has six (Fig. 1K). Secreted soluble Fc-tagged domain mutants were incubated with C4d and then pulled down with Protein A beads. For LilrB2, C4d bound to a D1D3 and to a D1D4 construct (Fig. 1 K and L); no binding was seen for constructs containing only D3 or D4. The LilrB2 D1D2 domain construct was unstable and expressed poorly (Fig. 1L) and could not be used to assess binding. For PirB, all constructs containing domains D1 and D2 bound C4d efficiently, including just the D1D2 domain construct (Fig. 1 K and L). C4d did not bind to PirB constructs containing domains D3 through D6. These results suggest that the D1 plus D2 extracellular immunoglobulin domains of PirB or LilrB2 are the minimum required to interact with C4d.

LilrB2 Is Present in Pyramidal Neurons and Excitatory Synapses of the Human Prefrontal Cortex and Colocalizes with C4d.

To detect whether LilrB2 and C4d are present at synapses, the human temporal cortex was examined in ultrathin serial sections using Array Tomography (AT) (36). This high-resolution immunohistochemical technique permits individual synaptic profiles to be imaged without background contamination from fluorescence signals originating in other cells or compartments above or below the 70 nanometer thin section (36). Synaptic puncta (Fig. 2 A and B) and pyramidal neurons (SI Appendix, Fig. S3) were immunostained with the LilrB2 antibody. In neuropil, LilrB2 immunostaining overlapped extensively with C4d puncta, as well as with C4 puncta and also markers for excitatory synapses (Fig. 2A). Synaptograms containing four 70 nm sequential thin sections through four different synapses further illustrated colocalization (Fig. 2B): LilrB2, C4, and C4d immunostaining colocalized at the same excitatory synapse, either as a direct overlap of all markers within one 70 nm section (e.g. Synapse 1: section #3), or with the sequential appearance of markers in the four serial sections defining a particular synapse (Fig. 2B: synapses 2, 3, 4). Note that Pearson correlation analysis of these markers reveals significant and nonrandom colocalization between LilrB2 plus C4d, C4d plus PSD95, and LilrB2 plus PSD95 (Fig. 2C), consistent with the idea that the postsynaptic compartment of excitatory synapses is enriched for C4d-LilrB2 ligand–receptor interactions.

Fig. 2.

LilrB2 is detected in pyramidal neurons and excitatory synapses of the human cerebral cortex and colocalizes with C4d. (A) High magnification image of a 70 nm thick Array Tomography section of the human temporal cortex (age 60 y) immunostained for pre- and postsynaptic markers VGluT1 (blue) and PSD95 (magenta), combined with LilrB2, C4, C4d (green). Synapses are identified by adjacency of VGluT1 and PSD95 puncta. Synapse enclosed in the white box is also shown in higher power synaptogram (Synapse 1) in B. (B) Synaptograms of four different synapses, imaged through four consecutive serial sections (70 nm each). (Scale bar, 1 mm.) (C) Pearson correlation coefficient as a function of lateral offset of pairs of fluorescent channels showing nonrandom spatial overlap between synaptic markers Vlut1, PSD95, LilrB2 and C4d. The VGluT1–PSD95 pair is a positive control; GAD2-PSD95 is a negative control. (D) Bar-graph with dot-plot comparing the fraction of total excitatory synapses coimmunostained for C4, C4d or LilrB2. Fifty random synapses from six regions (a total of 300 synapses) were analyzed. (E) Western Blot analysis of synaptosomes from postmortem frozen human prefrontal cortex from three different patients showing LilrB2 and C4d proteins, along with PSD95, and actin-associated Tuj1 (See also SI Appendix, Fig. S7). (F) Anti-LilrB2 or control Goat IgG immunoprecipitations from prefrontal cortex synaptosome preparations of age-matched Control and Alzheimer’s patient human brains, blotted with LilrB2 antibody. Also, a lane containing recombinant LilrB2 expressed in HEK293 cells was included as a marker for LilrB2 molecular weight. (G) Western blot of initial lysates or synaptosomes from (F), immunostained with PSD95 to show enrichment of synaptic markers such as PSD95 in synaptosomes.

To assess overall abundance of LilrB2, C4, and C4d at human cortical synapses, the localization of each of these three markers to excitatory synapses (defined as pairs of Vglut1+PSD95 puncta) was scored in AT micrographs. Nearly 60% of excitatory synapses colocalized C4 or LilrB2 immunostaining, and about 30% of synapses had C4d signal (Fig. 2D). Signals for C4 and C4d were frequently found together at the same synapses (Fig. 2 A−C), consistent with the idea that C4d may remain membrane-bound following cleavage from synaptically located C4 (33) (Fig. 1A). Finally, C4d- and LilrB2-specific bands were detected in Western blot analysis of synaptosomes prepared from the human prefrontal cortex (Fig. 2 E and F). Enrichment of synaptic proteins in synaptosome preparations was confirmed by Western blotting for the postsynaptic marker PSD95 (Fig. 2G). Both LilrB2 and PSD95 levels were lower in the Alzheimer’s synaptosomes. The decrease in PSD95 and LilrB2 expression per synapse are consistent with previous reports of synapse weakening and loss in advanced Alzheimer’s brains (Fig. 2 F and G). Notably, these findings confirm the presence of LilrB2, as well as C4 and C4d, in cortical synapses.

Using the same tissue sections from the human temporal cortex, immunolabeling for the inhibitory presynaptic marker GAD2 and postsynaptic marker gephyrin was also performed, along with LilrB2 and C4d (SI Appendix, Fig. S4). Pearson correlation analysis indicates nonrandom overlap between LilrB2 plus gephyrin signals and for C4d plus gephyrin (SI Appendix, Fig. S4C), but at ~3x lower levels than at excitatory synapses (compare with Fig. 2C). In sum, these observations point to a robust presence and colocalization of C4d and LilrB2 at excitatory synapses of the human neocortex, and to a lesser degree, at inhibitory synapses.

In mice, previous work demonstrated that PirB mRNA and protein are detected in neurons and at synapses using isotopic in situ hybridization or immunohistochemistry in tissue sections, or in Western blot analyses of synaptosome preparations (12). Here, we confirmed these findings using the BaseScope in situ hybridization method that permits high-sensitivity fluorescence detection of PirB mRNA combined with immunostaining using the neuronal marker NeuN (SI Appendix, Fig. S5). Using BaseScope, PirB mRNA was found in cortical (SI Appendix, Fig. S5 A−H) and hippocampal pyramidal neurons (SI Appendix, Fig. S5 I−L), as well as in cerebellar granule cells as expected (SI Appendix, Fig. S5 M and N) (12, 37). These BaseScope results also confirmed findings from the single-cell Nuc-Seq database for the mouse cerebral cortex published by the Broad Institute, in which most of the PirB signal is detected in excitatory neurons (38, 39). PirB mRNA was also detected in synaptosomes from mouse forebrain (SI Appendix, Fig. S5O), consistent with previous published observations of PirB protein in Western blots of synaptosome preparations (12, 18). Thus, similar to human LilrB2, murine PirB is in an ideal location to contribute to synapse pruning.

C4d Colocalizes with LilrB2 and Beta Amyloid (Abeta) in Human Alzheimer’s (AD) Cortex.

Biochemical and structural studies have shown that soluble oligomers of Abeta (oAbeta) also bind LilrB2 and PirB (18, 40, 41), and oAbeta has been observed at synapses in array tomography studies of human AD brains (42, 43). Therefore, postmortem samples of superior frontal gyrus from an AD patient were also examined (Fig. 3). Ultrathin sections for Array Tomography were immunostained for the same markers as above, as well as with an antibody recognizing soluble oligomeric Abeta (44). Low magnification views (Fig. 3A) revealed that C4d and LilrB2 signal were detected in and near amyloid plaques (Fig. 3 A Left). In addition, in the plaques there was little synaptophysin or LilrB2 signal, but C4d was very evident.

Fig. 3.

C4d colocalizes with LilrB2 and Beta Amyloid (Abeta) in the human AD cortex and is elevated in patients and APP/PS1 mice. (A) Array Tomography micrographs of one 70 nm section of the human prefrontal cortex from an AD patient (age 70 y), showing a plaque and surrounding neuropil immunostained for oligomeric Abeta (magenta, Left panel), C4d (green, Left and Middle panels) and LilrB2 (green, Right panel). Plaques are largely devoid of the presynaptic marker synaptophysin (blue). C4d and LilrB2 puncta are visible in the plaque and also in neuropil at some distance from the plaque. (Scale, 10 μm.) (B) Synaptograms of four different synapses, four serial sections each. Synapses 1 to 3 are located near the plaque; synapse 4 is further away (white boxes mark locations in panel A, Left). Abeta is not detected at synapse 4. Overlap between magenta/green or blue/yellow signals appears white. (C and D) Anti-C4 (C) and C4d (D) Western blot of brain lysates from AD (ages 81, 85) and non-AD samples (both ages 85) (Control). C4 and C4d-specific bands are much more intense in AD patients relative to non-AD (see also SI Appendix, Fig. S7). To confirm identity of C4 and C4d bands, 1.0 ng of human C4 protein complex or C4d were also included. (E) Quantification of data from anti-C4 and C4b Western blots of brain lysates from five AD and four age-matched non-AD samples from prefrontal cortex (SI Appendix, Fig. S7 A and B). C4 and C4b levels are significantly increased in AD patients. (LiCor Revert total protein stain was used for loading controls; see Materials and Methods and SI Appendix, Fig. S7). *P < 0.05, U-test. (F) qPCR of C4 mRNA expression in 13-mo-old APP/PS1 mouse cerebral cortex compared with age matched WT controls. n = 3 to 5 mice per age and genotype; *P < 0.05, ***P 0.001, t test. (G) Elevated levels of C4d protein in APP/PS1 mouse forebrain. C4 and split products were immunoprecipitated with mouse anti-C4 from lysates of 16-mo-old WT, APP/PS1 or C4KO forebrain, and detected with anti-C4 antibodies. C4d runs at ~42 kD; in C4KO lane; note absence of specific bands for C4, and cleavage products including C4b and C4d. (H) Averages of C4d levels in five C4 immunoprecipitations from WT, APP/PS1, or C4 KO brains in mice aged 15 to 18 mo. *P < 0.05, t test.

To visualize individual synapses, higher magnification synaptograms are shown. These illustrate that LilrB2, C4d, and Abeta signals colocalized at synapses near plaques (Fig. 3B: synapses 1, 2, 3). At some synapses further from plaques, Abeta was not detected (e.g. synapse 4 in Fig. 3A). Colocalization of all 3 markers was also seen in the non-AD age-matched temporal cortex (SI Appendix, Fig. S6). These observations place C4d and LilrB2 along with Abeta at a subset of excitatory synapses, particularly those located near amyloid plaques in the AD brain.

C4d and C4 Are Elevated in Human and Mouse Cerebral Cortex with Age and in AD.

Levels of C4 and C4d protein were assessed quantitatively in the human cerebral cortex using Western blot analysis of brain lysates from AD patients and controls (Fig. 3 C−E and SI Appendix, Fig. S7). Identification of the C4 and C4d bands was confirmed by coelectrophoresis of lysates with purified human C4 and C4d (Fig. 3 C and D). C4 levels were elevated, and C4d protein levels were fourfold higher, in the human AD prefrontal cerebral cortex relative to age-matched controls (Fig. 3E and SI Appendix, Fig. S7). Similarly, in the cerebral cortex of APP/PS1 AD model mice, increases in C4 mRNA (Fig. 3F), as well as C4 and C4d protein (Fig. 3 G and H) relative to WT were also observed, consistent with previous studies (22, 45, 46). Age-related increases in C4 and LilrB2 mRNA also are reported to occur in the human cerebral cortex from 24 y to plateau at about 40 y of age (SI Appendix, Fig. S8 A and B: replotted from Allen Brain Atlas; see also reference (45) for C4. In the mouse cerebral cortex, C4 mRNA also increased significantly with age, along with a modest but significant increase in PirB (SI Appendix, Fig. S8 C and D).

C4d Infusion In Vivo Drives PirB-Dependent Spine Loss on Cortical Pyramidal Neurons.

Genetic overexpression or deletion of the C4 gene itself cannot distinguish between roles for any of C4’s many cleavage products (Fig. 1A) and could also alter the complement cascade downstream of C4, including activation of C3. Therefore, to examine possible functional roles for a specific interaction between C4d and PirB in synapse pruning, a gain-of-function experiment was conducted. C4d protein levels were selectively increased by adding recombinant C4d in vivo in the mouse cerebral cortex. Since PirB is required for synapse pruning in vivo (10, 11), application of exogenous C4d should drive excessive elimination of dendritic spines and functional excitatory connections on pyramidal neurons. Moreover, if this effect is specific, then there should be little or no effect on spine density in PirBKO. When PirB is blocked or deleted from pyramidal neurons in development or adulthood, spine density is greater than normal on pyramidal cells in the CA1 hippocampus and in the visual and motor cortex (10–12, 15, 17). To determine whether C4d can increase spine loss in vivo, C4d was infused via osmotic minipumps into primary visual cortex of adult (P70) WT or PirBKO mice crossed to Thy-1 YFP-H transgenic mice (47), in which layer 5 (L5) pyramidal neuron dendrites and spines are labeled. Following 4 d of infusions of C4d or control (BSA) (Fig. 4 A and B), spine density was assessed (Fig. 4 C−E). The presence of intact C4d following infusion was confirmed in cortical lysates by Western blot (SI Appendix, Fig. S9). Spine density in WT infused with C4d was significantly lower compared with controls. In contrast, in the PirBKO cortex, spine density was indistinguishable from controls (Fig. 4E and SI Appendix, Table S2). Analysis of spine density changes in the WT cortex as a function of distance also revealed a clear dose–response in which the C4d-induced spine density decrease occurred within about 800 microns from the infusion site (Fig. 4F), closely matching the extent of C4d diffusion as assessed by immunostaining for C4d (Fig. 4B). No distance-related change in spine density was observed with any other treatment condition or most importantly, in the PirBKO cortex (Fig. 4F). Together this in vivo “gain of function” experiment reveals that exogenous addition of C4d can generate excessive structural pruning at cortical synapses and reveals a specific requirement for PirB in mediating spine loss in the presence of elevated C4d.

Fig. 4.

Minipump infusion of C4d into the visual cortex results in spine loss requiring PirB. Minipump infusion of C4d into the visual cortex results in spine loss requiring PirB. (A) Experimental protocol: 9 to 10-wk-old mice received continuous minipump infusion of C4d into the visual cortex for 4 d. (B) Fluorescent image of the visual cortex (sagittal section) infused with C4d and immunostained for the T7 tag (cyan) on C4d protein. (Scale bar, 400 μm.) (C) Example image of layer 5 Thy1-YFP-positive basolateral dendrites (white box) following infusions. (Scale bar, 20 μm.) (D) Confocal images of basolateral dendrite segments from indicated conditions; (Scale bar, 5 μm.) (E) Quantification of spine density on basolateral dendrites of L5 Thy1-YFP pyramidal neurons following infusion of Control (BSA) or C4d (10ug total) into WT or PirBKO cortex. Individual data points represent mean density per cell per animal (n = number of animals given by numerical values in histograms). *P < 0.05 by one-way ANOVA and Tukey post hoc test; see SI Appendix, Table S2 for exact P-values. As expected from previous studies, spine density in PirBKO is higher than WT on L5 neurons. (F) Data from (E) plotted as a function of spine density vs distance from infusion site in WT mice receiving either BSA or C4d. *P < 0.05, **P < 0.005, Welch’s t test. Additional statistics in SI Appendix, Table S2.

Discussion

The results of this study identify an unexpected function for C4d, a terminal split product of the complement component C4, in pruning and elimination of dendritic spines—the postsynaptic site of excitatory synapses—on pyramidal neurons in the cerebral cortex. The mouse PirB receptor is essential for this function, consistent with our findings that C4d is a high-affinity ligand of PirB and that PirB knockout prevents C4d driven spine pruning. C4d binds the PirB receptor, and human homolog LilrB2 with nanomolar affinity. Binding is saturable, and Surface Plasmon Resonance experiments provide further validation, demonstrating a direct interaction. Array tomography immunohistochemistry places C4d and LilrB2 proteins directly at excitatory synapses in the human cerebral cortex, making it possible that this receptor acts in neurons at synapses to transduce C4d signals generated by complement cascade activation. Thus, high-affinity interactions with LilrB2/PirB receptors may provide a possible molecular explanation for the well-known synapse loss that occurs when C4 and C4d increase with age and AD.

Unlike other split products in the complement cascade (C3d, C3b, C4b) that work as opsonins to mediate interactions via known complement receptors on a target structure including glia, C4d has been considered an orphan ligand without an in vivo function. Here, we report that C4d binding to LilrB2/PirB is saturable not only in HEK293 cells, but also in primary neurons. C4d-LilrB2 binding has been noted previously in an in vitro study of immune cells (33), but crucially here we have proven with surface plasmon resonance the presence of a genuine very high-affinity biophysical interaction. Using domain analysis, we have further narrowed down the region of interaction to involve at minimum the D1D2 N-terminal Ig-like domains. Most notably, we have presented evidence that in vivo, this interaction has structural consequences for cortical pyramidal neurons. These observations are consistent with the conclusion that C4d can generate significant pruning and elimination of excitatory synapses. The fact that there is no effect of C4d application in PirBKO strongly supports the idea that PirB is required for C4d-driven pruning. Together, our results add an unexpected neuronal and synaptic milieu for C4d and describe a function for C4d in the context of its interaction with LilrB2/PirB.

By combining this knowledge of synaptic location in human brain with results from C4d and PirB manipulations performed in the mouse brain, we propose that the human C4d-LilrB2 interaction, similar to murine C4d-PirB, is a key component of the molecular machinery regulating synapse elimination and pruning. Our results suggest a working model (Fig. 5) in which cleavage products of complement C4 at excitatory synapses play a dual role. Many previous studies have shown that C4 is cleaved into C4d and C4b and then the C4b cleavage product (Fig. 1A) further activates downstream components of the complement cascade, leading to engulfment of weakened synapses by glia (3, 9, 22, 24, 27, 45, 48–50). Here, we have added a missing link by showing that C4d binds to PirB at synapses to promote synapse weakening and elimination. These observations suggest a scenario in which all cellular constituents at the synapse—neurons, astrocytes and microglia—can be actively engaged in a coordinated complement-driven pruning process via their own cell-type specific receptors.

Fig. 5.

Proposed model for how C4 and C4d action at synapses can lead to weakening and elimination. (A) Strong synapse: Cofilin—an actin severing molecule—is in its inactive, phosphorylated form and synapse is stable. When the complement cascade is activated, for example by upstream components such as C1q (32), C4 is processed to generate split products C4b and C4d. (B) Weakened synapse: When C4d binds to the D1D2 domains of LilrB2/PirB, cofilin signaling is engaged (51, 52) leading to spine shrinkage and synapse disassembly (16–18). Meanwhile, C4b complexed with C2a binds C3 to generate iC3b (C3*). iC3b then binds to C3 receptors on glia (1, 8, 9). Thus, we suggest that C4 may have a dual function, engaging both neurons and glia via separate cleavage products downstream of C4: C4d (neurons) and iC3b (glia), resulting in synapse disassembly.

It is possible that a high-affinity C4d-LilrB2 interaction contributes to synaptic pruning in schizophrenia. In the prefrontal cortex of patients with schizophrenia, dendritic spine density is decreased on pyramidal neurons (53–55). In mice there is only one C4 gene, but in humans there are two alleles: C4A and C4B. Schizophrenia risk is increased in C4A carriers (19, 56), and previous studies have noted that synapse loss is increased in the cerebral cortex of mice overexpressing human C4A (26). Just how elevated C4A levels might lead to synapse pruning is unknown. Based on observations here, we suggest that elevated C4A protein in turn results in an increase in the C4d cleavage product that would be expected to exacerbate synapse pruning in schizophrenia via the interaction with LilrB2.

In aging there is also a dramatic rise in endogenous levels of C4 and C4d protein, as well as C4 mRNA, and levels are especially elevated in the cortex from human AD patients and the APP/PS1 mouse AD model (Fig. 3). Consequently, the interaction between C4d ligand and the LilrB2/PirB receptors could be a significant contributing factor to synapse loss occurring with aging and with early pathology in Alzheimer’s disease (57–59). We show that C4d protein is located at excitatory synapses, frequently colocalizing with LilrB2 (Fig. 2 and SI Appendix, Figs. S4 and S6). In the context of aging and AD, C4d, LilrB2, and oligomeric Abeta are all found at excitatory synapses in the human cortex (Fig. 3 and SI Appendix, Fig. S6). Previous studies have reported the presence of oligomeric Abeta at human cortical synapses (43), and C4d has been found in amyloid plaques in mouse AD models (22, 60). As levels of C4d and Abeta increase, their interactions with LilrB2/PirB are likely to exaggerate downstream consequences including synapse loss known to occur with age and in AD. These findings are consistent with previous observations that in AD mice models, PirB downstream signaling via cofilin to the actin cytoskeleton is hyperactivated, leading ultimately to synapse elimination (18, 51, 52) (Fig. 5). Thus, this system for synapse pruning joins other complement components including C1q, C3, CR3 that are also implicated in animal models of amyloid-mediated synapse loss (9, 22, 24, 45, 48–50).

Among risk factors for Alzheimer’s disease are both human genetic mutations and inflammation, systemic as well as in CNS (61–64). Familial mutations in APP and PS1 lead to early elevated levels of soluble beta amyloid and plaque formation, putting patients at high risk for AD (65, 66). Inflammation increases with aging and is associated with increased complement cascade components including C4 and its split products, both in blood and brain (22, 24, 50, 60). Inflammation is also known to increase expression of Major Histocompatibility Class I molecules, both peripherally and in CNS (67, 68). In mice, classical MHC class I molecules contribute to synapse pruning during development (69–71) and are present at synapses (20, 72). It is noteworthy that soluble oligomeric Abeta, C4d and classical MHC class I molecules all bind the D1D2 domains of PirB (mouse) and LilrB2 (primate) (18, 40, 41, 73). Thus LilrB2/PirB is a receptor with affinity for multiple ligands, consistent with previous observations for many other Ig-like immune receptors (30, 74). What is remarkable here is that PirB/LilrB2 and all three ligands are present in CNS and detected at synapses, suggesting that PirB/LilrB2 conveys signals about levels of inflammation. Genetically, MHC class I haplotypes have been associated with AD risk in a GWAS study (75) and LilrB2 has emerged recently in highly powered human AD GWAS studies (14), joining APP and other AD genetic risk factors. The LilrB2/PirB receptor may amplify synapse pruning in Alzheimer’s disease by acting as a common convergence point uniting both inflammation and genetics as dual risk factors. If so, then targeting the receptor may be far more effective in treating synapse pruning disorders than targeting any one ligand.

Materials and Methods

Ethics.

All experiments were carried out in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and approved by the Stanford University Institutional Animal Care and Use Committee. Methods are also in accordance with the Policies of the Society for Neuroscience on the Use of Animals and Humans in Neuroscience Research. All mice were maintained in a pathogen-free environment. Acquisition and analysis of dendritic spine data from the mouse cerebral cortex was performed blind to treatment and genotype. No statistical a priori sample size estimate was conducted. The sampling quota was determined when statistical power was reached. All chemical and biological reagents used, their sources, and experimental concentrations where applicable, are tabulated in the SI Appendix, Table S3.

Mouse Lines.

Germline PirB−/− mutant mice, as well as corresponding WT cohorts, were generated as described previously (12). Briefly, PirBfl/fl line was crossed with a deleter strain targeting Cre-recombinase expression to mouse embryos via adenovirus EIIa promoter (B6.FVB-TGN(EIIa-cre)C5379Lmgd, Jackson). Heterozygote sibling matings were then used to generate both control WT (PirB+/+) and a homozygous PirB−/− line on the same mixed C57Bl/ 6 × SV/ 129 J genetic background. PirBfl/fl mice were generated as described previously (12). For antibody validation, C4−/− mice on B6 background (B6.129S4-C4b^tm1Crr/J) were obtained from Feng Lin (Cleveland Clinic, Cleveland OH). All studies were performed in male mice at P29-P30 or P60 (PirB−/− vs. WT). APP/PS1 mice were generated as described previously (18).

Binding Studies.

HEK293 cell transfections and immunohistochemistry.

HEK293 cells (ATCC) were seeded into 8-well chamber coverslips (Nunc Lab-Tek 2) at a density of 2.5 × 10⁵ cells per well. Cells were then transferred to Opti-MEM (ThermoFisher) and transfected with HA-PirB, Myc-LilrB constructs, or empty vector using Lipofectamine 2000 (ThermoFisher). After 48 h, cells were treated with C4d for 2 h and then washed 3× with 4 °C DPBS. Cells were fixed in 4% paraformaldehyde/PBS for 13 min. They were then washed 3× with PBS and blocked for 1 h in blocking buffer (3% BSA/0.1%NP-40/PBS). Anti-HA, Myc, or C4d antibodies were then diluted in blocking buffer and coverslips were incubated for 16 h at 4 °C. After 3× washes with PBS, secondary antibodies conjugated to fluorescent dyes were added and incubated for 2 h. After 3× washes with PBS, coverslips were mounted with ProLong® Gold antifade reagent (Invitrogen Corp., Carlsbad, CA).

Primary neuron cultures and immunocytochemistry.

Primary cortical neurons were cultured as described (76) with modifications. The neocortex was isolated from E16 (embryonic day 16) mouse pups (after removing hippocampus and midbrain), dissociated with 0.25% trypsin for 20 min and plated on PDL (Sigma) and laminin (Thermo Fisher) coated 12 mm glass coverslips (Bellco) in a 24-well plate at 150,000 cells/well in Neurobasal media (NB) (Invitrogen) supplemented with 2% B27 (Invitrogen), 1% P/S, and 1% glutamine. Every 3 to 4 d, cells were fed with fresh NB-B27. For C4d Kd experiments, wild type or PirBKO DIV21 neurons were treated with purified C4d protein (Abcam) (concentration as indicated) at 37 °C for 2 h and then fixed in PFA/Sucrose and incubated overnight at 4 °C with anti-C4d (Abcam, AB198640) and anti-Tuj (Abcam, 78078) antibodies. Next day, cells were labeled with fluorescently conjugated secondary antibodies for 2 h at RT and mounted onto glass slides with ProLong® Gold antifade reagent. For PirB overexpression studies, cortical neurons were obtained from P0 mouse pups and cotransfected at DIV4 with both GFP and either an empty vector control or PirB-HA using the calcium phosphate method (77). Following C4d treatment, cultures were fixed and stained at DIV18 as described above with anti-HA (BioLegend, MMS-101R), anti-GFP (Abcam, ab5450), and anti-C4d (Abcam) antibodies.

Kd analysis.

For HEK293 cell and neuron assays, cells were fixed and stained as described above after being treated with varying concentrations of C4d (Abcam AB198640). Fluorescence images were acquired with a Leica SP8 confocal microscope using a 40x objective. Approximately 5 to 10 images per concentration of C4d were acquired to assess C4d binding. To measure C4d binding onto HEK293 cells expressing either PirB/LILRB2, regions of interest were generated by creating masks from the thresholded intensity of either GFP or PirB/LILRB2 expression by using the threshold and convert to mask function in Fiji. The mean signal intensity (MFI) of bound C4d within ROIs were then measured and recorded. Background subtraction was performed by subtracting the signal intensity of bound C4d in the untreated cells from treated cells. To calculate Kd of specific binding, signal intensity of C4d bound to cells transfected with empty vector was subtracted from PirB/LILRB expressing cells that were treated with equivalent concentrations of C4d. Kd was generated in GraphPad Prism using the nonlinear isotherm model.

For experiments done in C4d-treated neuron cultures, image analysis was performed similarly as in HEK cells. Tuj staining was used to generate a mask and isolate neuronal signal which was overlaid onto C4d staining and the mean fluorescence intensity of overlapping C4d signal was measured and quantified using ImageJ (Fiji). To isolate PirB specific binding, signal intensity from PirB KO neurons was subtracted from WT neurons at each concentration and data was fit to a nonlinear isotherm using GraphPad PRISM.

For experiments comparing C4d binding to LilrB1, B2, or B3 expressed in HEK293 cells, cells were cultured, transfected, fixed and stained as described above after being treated for 2 h with 570 mM C4d. As described above, fluorescence images were acquired with a Leica SP8 confocal microscope using a 40x objective. Approximately 5 to 10 images were acquired for each LilrB family member to assess C4d binding. To measure C4d binding, regions of interest were generated by creating masks from the thresholded intensity of either GFP or PirB/LILRB2 expression by using the threshold and convert to mask function in Fiji. The mean signal intensity (MFI) of bound C4d within ROIs were then measured and recorded. Background subtraction was performed by subtracting the signal intensity of bound C4d in the untreated cells from treated cells. The results from three different experiments were averaged and graphed as arbitrary units (AU).

Array tomography of human brains.

Tissue collection: To preserve antigenicity of proteins and to prevent freezing-induced destruction of membranes, we obtained fresh adult female, 60-y-old, brain tissue from the human temporal cortex, that had to be removed to gain access for surgical treatment of a deep-brain tumor (deidentified resected cortex: courtesy of Dr. Gábor Tamás, University of Szeged, Hungary). Brain tissue from the superior frontal gyrus of a female 70-y-old Alzheimer’s disease patient was obtained from the Neurodegenerative Disease Brain Bank at the University of California, San Francisco (SI Appendix, Table S1). It was fixed at 9.2 h postmortem. In both cases, the tissue was fixed with 4% paraformaldehyde in PBS for 24 h at 4 °C.

Tissue embedding: The tissue was dehydrated and embedded using standard array tomography protocol (36). Briefly, small tissue blocks from the neocortex (3 × 2 × 1 mm, H × W × L) were dissected out and dehydrated in a graded series of ethanol to 100% ethanol, then infiltrated in LRWhite resin, transferred to gelatin capsules and polymerized at 55 °C for 24 h. Orientation of the block ensured that all the neocortical layers were present in the block facing the cutting surface.

Tissue sectioning: To prepare ribbons of serial sections, the blocks were trimmed around the tissue to the shape of a trapezoid, and glue (Weldwood Contact Cement diluted with xylene) was applied with a thin paint brush to the leading and trailing edges of the block pyramid. The embedded plastic block was cut using a Histo Jumbo diamond knife (Diatome) on an ultramicrotome (Leica Ultracut EM UC6) into 70 nm-thick serial sections, which were mounted on gelatin-coated coverslips. To quantify proteins present at synapses, ribbons of 20 to 50 sections were used.

Immunostaining and Imaging: Sections were stained for LilrB2, complement protein component C4, and its cleavage product C4d, following the protocol in ref. 36. To assess the synaptic localization of LilrB2 and complement proteins, ribbons were costained for excitatory synaptic markers VGlut1, synapsin, synaptophysin, PSD95, inhibitory markers GAD and gephyrin, as well as the astroglial marker glutamine synthetase, and oligomeric Abeta by Alexa-Fluor dye-conjugated secondary antibodies, highly cross-adsorbed (Life Technologies).

Up to three antibodies from different host-species were applied together and imaged, followed by an antibody elution. Ribbons were then restained and reimaged with additional antibodies. This allowed for localization comparison of six or more antibody signals in the same ROI. Sections were mounted using SlowFade Gold antifade reagent with DAPI (Invitrogen). Imaging was done on a Zeiss AxioImager.Z1 fluorescence microscope with AxioCam HRm CCD camera, using a Zeiss 63×/1.4 NA Plan Apochromat objective. Images were aligned using the MultiStackReg plugin in Fiji (78). Colocalization analysis was done in Fiji using van Steensel analysis (79).

Alzheimer brain lysis, Western analysis and synaptosome preparations.

Fresh frozen deidentified specimens of the human frontal cortex were obtained from the Human Brain and Spinal Fluid Resource Center, VA West Los Angeles Healthcare center (Los Angeles, CA), sponsored by NINDS/NIMH, National Multiple Sclerosis Society, Department of Veterans Affairs (SI Appendix, Table S1). Cortices from age-matched controls and Alzheimer’s disease were dounced in 10 volumes of Tris-buffered saline pH 7.5 (TBS) containing protease inhibitors (Pefabloc, Invitrogen, PMSF, Sigma). Homogenates were sonicated for 3 min and then centrifuged at 15,000×g for 30 min. Pellets were solubilized with 2% SDS in TBS with protease inhibitors, and then centrifuged at 15,000×g for 30 min. Supernatants were then denatured in LDS NuPAGE sample buffer (ThermoFisher) with 1% 2-beta-mercaptoethanol, heated at 100 °C for 3 min, and subjected to SDS-PAGE on 4-20% precast gels (BioRad 456-1098). 1 ng purified human C4 (Complement Technology A105) or C4d (Abcam AB198640) was also electropheresed on the gels, as a marker for correct molecular weight of these proteins. Protein levels in lysates were normalized and quantified using REVERT 700 Total Protein Stain (LiCOR 926-11010) and LiCOR ImageStudioLite software (80, 81).

For synaptosome preparations, frozen human prefrontal cortices were homogenized by 12 strokes of a Dounce homogenizer in Homogenization buffer (10 mM HEPES pH 7.3, 0.5 mM EGTA, 33% sucrose, 4 mM Pefabloc SC PLUS (Roche), and 0.2 mM Phenylmethanesulfonyl fluoride (PMSF), and centrifuged for 10 min at 2,000×g. An aliquot was then isolated for whole cell lysate (WCL). Supernatants were passed through three layered 100 µm pore nylon membranes and then passed through 5 µm nitrocellulose filters. For further purification to synaptosomes, filtrates were centrifuged at 10,000×g for 10 min, and pellets were resuspended in Cell lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1% NP-40, and protease inhibitors Pefabloc, Invitrogen, PMSF, Sigma). 10 micrograms anti-LilrB2 antibody (R&D Systems, AP2078) or goat IgG (R&D Systems, AB-108-C) were added to lysates for 2 h, and then Protein G agarose beads (Invitrogen, 15920010) were added for 1 h. Precipitated beads were washed three times with Cell lysis buffer, and then immunoprecipitates were denatured in LDS NuPAGE sample buffer (ThermoFisher) with 1% 2-beta-mercaptoethanol, heated at 100 °C for 3 min. Samples were subjected to SDS-PAGE, along with a sample lysate from HEK293 cells transfected with recombinant LilrB2, as a marker for LilrB2 molecular weight.

Osmotic minipump infusion of C4d into the visual cortex in vivo.

Craniotomies were performed and minipumps (Alzet Model 1002, 0.25 μL/h, 100 μL capacity) were implanted as previously described (15). Minipumps were loaded with BSA (0.5 μg/μL) (VWR EM-2930) or C4d (0.5 μg/μL) (Abcam 198640) in 1× PBS and implanted subcutaneously and connected to a cannula (Plastics One, 3300PM/SPC) that was inserted just anterior to primary visual cortex (2.5 mm-lateral and 3-mm posterior to bregma), The concentration used in the minipumps (0.5 μg/μL) and the rate of infusion (0.25 μL/h) was chosen to expose the brain parenchyma to 0.125 μg at steady state in a volume of approximately 1 μL (125 μg/mL). This is about 10-fold higher than levels reported to be present in CSF during the neuroinflammation generated by meningitis (26, 48).To determine the extent of C4d diffusion from the infusion site, a subset of brains were rapidly removed and frozen in M-1 embedding matrix (Thermo Scientific Shandon). Cryostat sections cut in the coronal plane and containing visual cortex were postfixed in 4% PFA for 20 min, blocked for 1 h at RT (3% donkey serum, 0.3% Triton-X 100), and placed in primary antibodies overnight anti-T7 (Cell Signaling, 13246). Next day, sections were placed in secondary antibodies for 2 h at RT and mounted onto slides in Prolong Gold. Images were acquired at 10x using a Leica SP8 confocal microscope.

Dendritic spine imaging and analysis in the mouse cerebral cortex.

Brains from animals infused with BSA or C4d for 4 d were fixed by transcardial perfusion of 4% paraformaldehyde (PFA) in PBS. Brains were postfixed overnight at 4 °C in PFA and then sectioned coronally on a vibratome. Sections were cut at a thickness of 150 µm mounted onto slides in Prolong Gold. Basolateral dendrites of YFP-positive layer 5 neurons in the visual cortex were imaged using a 63x objective on a Leica SP8 confocal microscope. Images were acquired at high resolution (2,480 × 2,480) and as done previously (15) and spines along a dendritic branch up to 150 µm from the cell soma were analyzed using ImageJ (Fiji, Simple neurite tracer plugin). Acquisition and analysis were performed in a blinded manner.

Quantification and statistical analysis.

Statistics for each experiment are reported in the accompanying Figure Legends. Data analysis and statistical analyses were performed using SigmaPlot 10.0 (Systat Software Inc.), IBM SPSS Statistics 23 (IBM Corporation), and Prism (GraphPad). Data are reported as mean ± s.e.m, with sample size given as number of cells along with the number of mice (e.g. n = x cells/y mice). One-Way ANOVA with post hoc Tukey was used to control for multiple comparisons. Datasets were comparable in variance (Levene’s test). The pairwise Mann–Whitney (U) test (normal distribution not assumed) was also used to compare means where appropriate. All statistical tests were two-tailed. Statistical significance was reached when probability due to chance fell below 5%; i.e. P-value P < 0.05. Outcomes of statistical tests are reported as exact P values, unless the probability due to chance fell below 0.1%, in which case it is always denoted as P < 0.001.

Additional Materials and Methods, including dot blots, Surface Plasmon Resonance (SPR), Extracellular domain analysis, mouse C4d immunoprecipitations, RNA extraction and Q-PCR, and Basescope PirB mRNA assay are located in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.