Heparan sulfate-dependent RAGE oligomerization is indispensable for pathophysiological functions of RAGE

RAGE, a druggable inflammatory receptor, is known to function as an oligomer but the exact oligomerization mechanism remains poorly understood. Previously we have shown that heparan sulfate (HS) plays an active role in RAGE oligomerization. To understand the physiological significance of HS-induced RAGE oligomerization in vivo, we generated RAGE knock-in mice (AgerAHA/AHA) by introducing point mutations to specifically disrupt HS-RAGE interaction. The RAGE mutant demonstrated normal ligand-binding but impaired capacity of HS-binding and oligomerization. Remarkably, AgerAHA/AHA mice phenocopied Ager−/− mice in two different pathophysiological processes, namely bone remodeling and neutrophil-mediated liver injury, which demonstrates that HS-induced RAGE oligomerization is essential for RAGE signaling. Our findings suggest that it should be possible to block RAGE signaling by inhibiting HS-RAGE interaction. To test this, we generated a monoclonal antibody that targets the HS-binding site of RAGE. This antibody blocks RAGE signaling in vitro and in vivo, recapitulating the phenotype of AgerAHA/AHA mice. By inhibiting HS-RAGE interaction genetically and pharmacologically, our work validated an alternative strategy to antagonize RAGE. Finally, we have performed RNA-seq analysis of neutrophils and lungs and found that while Ager−/− mice had a broad alteration of transcriptome in both tissues compared to wild-type mice, the changes of transcriptome in AgerAHA/AHA mice were much more restricted. This unexpected finding suggests that by preserving the expression of RAGE protein (in a dominant-negative form), AgerAHA/AHA mouse might represent a cleaner genetic model to study physiological roles of RAGE in vivo compared to Ager−/− mice.


Introduction
Receptor for Advanced Glycation End-Products (Ager gene) encodes RAGE, a transmembrane protein belonging to the immunoglobulin (Ig) superfamily of receptors. RAGE has attracted tremendous interest in the last two decades given its involvement in many disease conditions including diabetes, Alzheimer's, atherosclerosis, cancer, and many other inflammatory disorders (Hudson and Lippman, 2018;Yan et al., 2010). RAGE binds a wide array of ligands including advanced glycation end products, amyloid β-peptide, S100 family proteins, and high mobility group Box 1 (HMGB1). In most pathological conditions, diseased tissues show a dramatically enhanced expression and activation of RAGE. Ablation of RAGE signaling in these conditions, either by neutralizing RAGE ligands or by using RAGE knockout mice, has resulted in greatly improved outcome in murine models, which strongly suggests that RAGE can be pursued as a promising therapeutic target (Arnold et al., 2020;Harja et al., 2008;Hofmann et al., 1999;Huebener et al., 2015;Park et al., 1998;Taguchi et al., 2000).
While many pathological roles of RAGE have been well established, the homeostatic role of RAGE remained a mystery until 2006, when it was found that RAGE plays an essential role in regulating osteoclast differentiation and maturation (Zhou et al., 2006). While unchallenged RAGE null mice display normal homeostasis in all other aspects, they demonstrated increased bone mineral density. Mechanistically, it was revealed that HMGB1 secreted by preosteoclasts activates RAGE in an autocrine fashion, and HMGB1-RAGE interaction plays a critical role in establishing the sealing zone, a structural feature of osteoclasts essential for bone resorption (Zhou et al., 2008).
Recently, HMGB1-RAGE signaling has been shown to also play a key role in drug-induced liver injury (Huebener et al., 2015). Following acetaminophen (APAP) overdose, hepatocytes undergoing necrosis release HMGB1 as a danger signal to promote tissue repair. However, when extensive tissue necrosis occurs, the exaggerated inflammation, mediated by HMGB1-driven, RAGE-dependent neutrophil infiltration, often leads to secondary liver damage with worsened disease outcome. In a mouse of APAP-induced liver injury, we and others demonstrated that Ager −/− mice were greatly protected from liver damage chiefly due to suppression of neutrophil infiltration (Arnold et al., 2020;Huebener et al., 2015).
Like many cell surface receptors, RAGE signaling depends on receptor oligomerization Popa et al., 2014;Zong et al., 2010). However, the exact mechanism by which RAGE oligomerizes remains controversial (Moysa et al., 2019). Historically, RAGE has been hypothesized to undergo ligand-dependent oligomerization (Xue et al., 2016). We have proposed an alternative hypothesis: that prior to ligand binding, RAGE preassembles into an oligomeric complex by interacting with heparan sulfate (HS), a negatively charged polysaccharide widely expressed at the cell surface. This alternative hypothesis is supported by in-depth structural analyses, where we showed by both crystal and solution structures that HS oligosaccharides induce RAGE to form a stable hexamer in the absence of ligand (Xu et al., 2013). Furthermore, in the same study, we showed that in endothelial cells, activation of RAGE signaling by all tested RAGE ligands, including HMGB1, S100A8/ A9, S100A12, S100B, and AGE, depended upon the presence of HS at the cell surface. Based upon these findings, we propose that HS functions as a main driver of RAGE oligomerization, assembling RAGE into an oligomeric complex at cell surface without activating it. Only this oligomeric complex can serve as a functional receptor, which can then be activated by RAGE ligands to initiate signaling. As virtually all mammalian cells express HS, HS-dependent RAGE oligomerization is likely a universal requirement for RAGE signaling regardless of cell types and ligands involved.
To investigate the physiological significance of HS-RAGE interaction and HS-dependent RAGE oligomerization, we generated a novel RAGE knock-in (Ager AHA/AHA ) mouse model by introducing point mutations into residues that are involved in HS binding. Remarkably, when bone morphometric indexes and susceptibility to APAP-induced liver injury were examined, we found that the Ager AHA/ AHA mice phenocopy Ager −/− mice precisely. Our findings strongly suggest that HS-RAGE interaction is required for normal function of RAGE in multiple cellular systems and that targeting HS-RAGE interaction could represent a new opportunity to curb RAGE activation. By developing a monoclonal antibody (mAb) that specifically targets part of the HS-binding site of RAGE, we demonstrated that this novel targeting strategy was indeed effective in inhibiting osteoclastogenesis and APAP-induced liver injury. Finally, we discovered that compared to Ager −/− mice, alterations in global transcriptome were much more restricted in Ager AHA/AHA mice, which suggest that Ager AHA/AHA mice might be a better murine model to study RAGE-dependent biological processes.

Results
Characterization of RAGE-AHA mutant and generation of Ager AHA/AHA knock-in mice RAGE is a multidomain protein consisting of three extracellular immunoglobulin domains (V-C1-C2, also known as sRAGE), a single transmembrane domain, and a short cytoplasmic tail (Dattilo et al., 2007;Koch et al., 2010). We previously reported that HS can induce RAGE to form a hexamer (trimer of dimers) with 1:2 stoichiometry, and that the interaction involves five basic residues in the V domain and two residues (R216 and R218) in the C1 domain ( Figure 1A; Xu et al., 2013). Here, we prepared a triple mutant of mouse RAGE VC1 domain bearing R216A, R217H, and R218A mutations (mVC1-AHA). As expected, the binding capacity of mVC1-AHA to the heparin Sepharose column was reduced significantly compared with wild-type mouse RAGE VC1 domain (WT-mVC1) ( Figure 1B). Furthermore, mVC1-AHA was unable to form stable hexamer in the presence of HS dodecasaccharides ( Figure 1C). Importantly, mVC1-AHA maintains WT-like binding affinity to RAGE ligand HMGB1 (apparent binding affinity 9.9 nM vs. 11.7 nM, Figure 1D), which is expected because binding of RAGE ligands predominantly occurs at the V domain. To generate a murine model to specifically disrupt HS-RAGE interaction, we introduced R216A-R217H-R218A triple mutations into the ager (synonymous with rage) locus by CRISPR-Cas9 mediated homologous recombination ( Figure 1E). Ager AHA/AHA mice were born with normal litter size and grew normally without gross morphological abnormalities. The RAGE expression level in Ager AHA/AHA mice lungs was similar to WT mice, which suggested that the mutation did not negatively affect the expression of RAGE ( Figure 1F). Ager AHA/AHA and Ager AHA/+ mice develop an osteopetrotic phenotype As RAGE was reported to be associated with osteoclast maturation (Zhou et al., 2006), we first examined the bone phenotype of Ager AHA/AHA mice. Micro-computed tomography (micro-CT) analysis of the femoral trabecular bone morphology and microarchitecture showed that 10-week-old male Ager AHA/ AHA mice exhibited increased trabecular bone mass ( Figure 2A) and displayed a 49% increase in bone volume over tissue volume (BV/TV) ( Figure 2B), 17% increase in trabecular thickness (Tb.Th), and 15% decrease in trabecular separation (Tb.Sp) (Figure 2-figure supplement 1). In addition, these Ager AHA/AHA mice also displayed a significant increase in the cortical bone thickness (from 0.182 to 0.205 mm) ( Figure 2C). Similar phenotype was also observed in 10-week-old female Ager AHA/AHA mice, which displayed a 70% increase in bone volume compared to WT mice ( Figure 2-figure supplement  2). Notably, all these parameters are highly similar to those of 10-week-old Ager −/− mice ( Interestingly, when they were 4 weeks old, the Ager AHA/AHA mice had normal bone mass, while Ager −/− mice had slightly lower bone mass compared to WT mice. By 4 months old, while the bone mass of Ager AHA/AHA mice was 26% higher than WT mice, the bone mass of Ager −/− mice was significantly higher (75% over WT). Taken together, these data indicate that the interaction between HS and RAGE, and HS-induced oligomer formation, are indispensable for RAGE to function in bone remodeling.
To understand the potential contribution of haploinsufficiency and RAGE gene dosage in bone phenotype, we examined both strains of heterozygous mice. To our surprise, male Ager AHA/+ mice already manifested significant increase in BV/TV (30%) and in cortical bone thickness (from 0.182 to 0.198 mm) (Figure 2A-C). In contrast, although an upward trend was observed in male Ager +/− mice, they did not display significant increase in either parameter ( Figure 2B-C). However, 10-week-old female Ager +/− mice displayed a significant increase in BV/TV (49%) which was almost identical to the increase observed in female Ager AHA/+ mice (51% increase) (Figure 2-figure supplement 2). Of note, while the BV/TV value was significantly different between Ager +/− and Ager −/− mice in both male and female mice, it was not significantly different between Ager AHA/+ and Ager AHA/AHA mice. Our findings suggest that RAGE-AHA mutation led to a haploinsufficient bone phenotype, and that gene dosage of RAGE also plays a role in determining the bone phenotype, especially in female mice.
Osteoclastogenesis is impaired in Ager AHA/AHA mice Tartrate-resistant acid phosphatase (TRAP) staining of tibia sections showed the number of osteoclasts in Ager AHA/AHA mice was reduced to a similar level as in Ager −/− mice ( Figure 3A-B), which indicates a defect of osteoclast differentiation in Ager AHA/AHA mice. To confirm this defect, we performed an osteoblast/bone marrow macrophage (BMM) co-culture osteoclastogenesis assay in vitro. In this model, WT Figure 1. Characterization of HS binding-deficient RAGE mutant (R 216 R 217 R 218 to A 216 H 217 A 218 ) and generation of Ager AHA/AHA knock-in mice. (A) Cartoon diagram of HS-induced hexamer of RAGE V-C1 domains (PDB 4IMB). The hexamer is organized as trimer of dimers, with each dimer stabilized by one molecule of HS oligosaccharide. Oligosaccharides are shown in space-filling models. One dimer (salmon and green) is enlarged to show HS binding residues (in stick representation). R216 and R218 from both monomers (V-C1 and V'-C1') are marked. (B) Binding of wild-type murine RAGE V-C1 domain (mVC1-WT) and RAGE-AHA mutant (mVC1-AHA) to heparin Sepharose column. RAGE-AHA mutant had reduced HS-binding capacity. (C) mVC1-WT or mVC1-AHA were incubated with HS dodecasaccharide (H12) and the mixtures were resolved on a Superdex 200 (10/300 mm) gel filtration column. H12 was unable to induce RAGE-AHA to form a stable RAGE hexamer. (D) Binding affinity of mVC1-WT and mVC1-AHA to immobilized HMGB1 was determined by enzyme-linked immunosorbent assay (ELISA). RAGE-AHA displayed WT-like binding affinity to ligand. (E) Targeting strategy for generating Ager AHA/AHA knock-in mice. Sequences of the targeting single guide RNA (sgRNA), the mutation sites in Ager exon 6, and the repairing template single-stranded donor oligonucleotides (ssODN) are shown. (F) Western blotting analysis of RAGE expression in lung lysate. Top panel, lung lysates from WT, Ager AHA/AHA , and Ager −/− mice were blotted with a rat anti-RAGE mAb (R&D system), which showed RAGE was expressed at normal level in Ager AHA/AHA mice. Bottom panel, as a protein loading control, the blot was reprobed with anti-HMGB1, which is universally expressed by all cells.
The online version of this article includes the following source data for figure 1: Source data 1. Binding affinity of mVC1-WT and mVC1-AHA to immobilized HMGB1 was determined by enzyme-linked immunosorbent assay (ELISA).
Source data 2. Western blotting analysis of RAGE expression in lung lysate.  Figure 3C). To better quantify the differences in osteoclast numbers among the three genotypes, we sub-grouped the osteoclasts based on the number of nuclei number they contain ( Figure 3D). Interestingly, while the number smaller osteoclasts (3-10 nuclei) were not significantly reduced in Ager AHA/ AHA and Ager −/− , the number of larger osteoclasts (>11 nuclei) were substantially reduced ( Figure 3D). Collectively, these results demonstrate that impaired HS-RAGE interaction leads to abnormal maturation of osteoclasts, which suggests a key role for RAGE-HS interaction in regulating osteoclastic differentiation and function. In addition, the mutagenesis in osteoblast lineage cells did not seem to have an impact on bone formation as examined by fluorochrome double labeling ( Ager AHA/AHA mice were protected from liver injury after APAP overdose We and others have previously reported that RAGE was responsible for neutrophil-mediated secondary injury (in an HMGB1-dependent manner) following acetaminophen (APAP)-induced liver necrosis (Arnold et al., 2020;Huebener et al., 2015). When subjected to a sublethal dose of APAP, Ager −/− mice displayed a reduction of neutrophil infiltration and liver damage compared with WT mice. We thus similarly used this APAP overdose model to examine the role of HS-RAGE interaction in mediating neutrophilic liver injury. As expected, we found that like Ager −/− mice, Ager AHA/AHA mice were also protected from APAP-induced liver injury. The plasma level of alanine aminotransferase (ALT), a biomarker of liver damage, was reduced by nearly 45% in Ager AHA/AHA mice ( Figure 4A). Similarly, neutrophil infiltration into the liver was also significantly reduced in Ager AHA/AHA mice ( Figure 4B). Histologic examination also confirmed that livers from Ager AHA/AHA mice had significantly reduced necrotic area compared to the injured WT livers ( Figure 4C-D). Notably, these results were very similar to what we previously observed in Ager −/− mice (Arnold et al., 2020). Our data strongly suggest that in this drug-induced liver injury model, HS-RAGE interaction plays an essential role in mediating HMGB1-dependent neutrophil infiltration.
RNA-seq transcriptome analysis of neutrophils from Ager AHA/AHA mice and Ager −/− mice To understand whether ablation of RAGE signaling has an impact on the general mobility of neutrophils, we performed an air pouch assay to examine LPS-induced neutrophil infiltration. Interestingly, in this assay, neutrophil infiltration was indistinguishable between WT, Ager AHA/AHA , and Ager −/− mice ( Figure 5A). This result confirmed that Ager AHA/AHA and Ager −/− neutrophils have normal mobility when a different chemoattractant (other than HMGB1) was responsible for the infiltration. During this assay, we unexpectedly discovered that several common surface markers were dramatically upregulated on Ager −/− neutrophils with 100% penetrance. Compared to WT neutrophils, the expression level of Ly6C and CD45 on Ager −/− neutrophils were 10-to 15-fold higher ( Figure 5B). Surprisingly, the expression levels of both markers were completely normal on Ager AHA/AHA neutrophils. The abnormal expression of Ly6C and CD45 was not restricted to neutrophils as we also found they are overexpressed in T and B cells from Ager −/− mice ( Figure 5-figure supplement 1). The fact that two out of eight common leukocyte markers were highly elevated in Ager −/− (the other six markers were normal, including CD11b shown in Figure 5B), but were normal in Ager AHA/AHA neutrophils strongly suggests that complete 0.01, and 0.001, respectively.
The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Ager AHA/AHA and Ager AHA/+ mice develop osteopetrotic phenotype.     As expected, the overall number of differentially expressed genes (DEGs) in Ager −/− neutrophils were almost 2.5 times higher than in Ager AHA/AHA contain. n=3 wells. * and ** represent p<0.05 and 0.01, respectively. Data are representative of at least three separate assays. TRAP, tartrate-resistant acid phosphatase; WT, wild-type.
The online version of this article includes the following source data and figure supplement(s) for figure 3: Source data 1. Ager AHA/AHA mice display impaired osteoclastogenesis.    The online version of this article includes the following source data for figure 4: Source data 1. Ager AHA/AHA mice were protected from liver injury after APAP overdose. DEGs between Ager AHA/AHA and WT PMNs were determined based on Benjamin-Hochberg multicomparison correction p values. 178 genes were found to be significantly upregulated (p adj value≤0.05, log 2 fold change≥1) while 69 genes were found to be significantly downregulated (p adj value≤0.05, log 2 fold change≤-1). DEGs between Ager −/− and WT PMNs were determined similarly. 476 genes were found to be significantly upregulated (p adj value≤0.05, log 2 fold change≥1) while 107 genes were found to be significantly downregulated (p adj value≤0.05, log 2 fold change≤-1). (D) Volcano plot of DEGs between Ager AHA/AHA PMN and WT PMNs. Genes with Figure 5 continued on next page neutrophils (603 vs. 247). Significant overlap was found in upregulated genes, with 80% of upregulated genes in Ager AHA/AHA (143/178) also being upregulated in Ager −/− neutrophils ( Figure 5C and Supplementary file 1). In contrast, the overlap was rather limited in downregulated genes, with merely 29% (20/69) of downregulated genes in Ager AHA/AHA also being downregulated in Ager −/− . Gene ontology (GO) analysis of upregulated genes found that genes related to regulation of leukocyte activation (43 genes), differentiation (27 genes), proliferation (29 genes), adhesion (34 genes), and regulation of cytokine production (30 genes) were highly enriched in Ager −/− neutrophils (Supplementary file 2). The same biological processes were also enriched in Ager AHA/AHA neutrophils, except that the number of genes involved were roughly half of the number of genes found in Ager −/− neutrophils (Supplementary file 3). These findings suggest that RAGE signaling plays a profound role in leukocyte biology. GO analysis of downregulated genes did not find any biological processes that are significantly differentially regulated in either Ager −/− and Ager AHA/AHA neutrophils. Curiously, we did identify several highly downregulated genes in Ager AHA/AHA neutrophils that were not found in Ager −/− neutrophils, including Gbp2b, Ifi44, and Ifi44l ( Figure 5D and Figure 5-figure supplement 3A-C), all of which have been shown to be interferon γ-induced genes and play important roles in defense against infection (Santos and Broz, 2018;Kim et al., 2011;Busse et al., 2020). Many top DEGs in Ager −/− neutrophils were not found in Ager AHA/AHA neutrophils ( Figure 5E), which include highly upregulated genes: Rps3a2, Rps3a3, Rps13_ps1, Rps15a_ps7, Tpm3_rs7, and Rnf5; and highly downregulated genes: Cxcl1, Muc1, Trim46, Ifi208, and BC147527. In addition, we also examined expression levels of seven most highly expressed RAGE ligands and found none of those were altered in Ager AHA/AHA neutrophils ( Figure 5-figure supplement 4). In Ager −/− neutrophils, while we found expression levels of most RAGE ligands were unaltered, the expression levels of S100a8 and S100a11 were reduced by ~40%. To our surprise, Ly6C and CD45 were not among the DEGs in Ager −/− neutrophils, which suggest that either the posttranscriptional regulation or the secretion of Ly6C and CD45 were altered in Ager −/− neutrophils. If that were the case, we would expect that the global protein expression profile of Ager −/− neutrophils would perhaps be even more dramatically altered than what is apparent on RNA sequencing.
RNA-seq transcriptome analysis of lung tissues from Ager AHA/AHA mice and Ager −/− mice To gain further insight into the impact of impaired RAGE signaling on gene transcription, we performed RNA-seq analysis of lung tissue, an organ in which RAGE is abundantly expressed. Surprisingly, the number of DEGs was much less in lungs compared to neutrophils in both strains. We found 202 DEGs in Ager −/− lungs and merely 31 DEGs in Ager AHA/AHA lungs ( Figure 5F and Supplementary file 4). The difference between the two genotypes was however even more dramatic in lungs (7-fold) than we observed in neutrophils (2.5-fold). Similar to what we found in neutrophils, in Ager AHA/AHA lung significant overlap was found in upregulated genes, with 50% of upregulated genes in Ager AHA/AHA (7/14) were also upregulated in Ager −/− ( Figure 5F and Supplementary file 4). In contrast, the overlap was most significant changes were marked in the plot. (E) Volcano plot of DEGs between Ager −/− PMN and WT PMNs. (F-H) RNA-seq transcriptome analysis of whole lungs from Ager AHA/AHA and Ager −/− mice. (F) Plot of up-and downregulated genes in Ager AHA/AHA and Ager −/− lungs. 14 genes were found to be significantly upregulated while 17 genes were found to be significantly downregulated in Ager AHA/AHA lungs. In Ager −/− lungs, 91 genes were found to be significantly upregulated while 111 genes were found to be significantly downregulated. (G) Volcano plot of DEGs between Ager AHA/AHA and WT lungs. Genes with most significant changes were marked in the plot. (H) Volcano plot of DEGs between Ager −/− and WT lungs. DEG, differentially expressed gene; PMN, polymorphonuclear; WT, wild-type.
The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. LPS-induced neutrophil infiltration into air pouch in WT, Ager AHA/AHA , and Ager −/− mice.       very limited in downregulated genes, with merely 23.5% (4/17) of downregulated genes in Ager AHA/ AHA were similarly downregulated in Ager −/− . Comparing the DEGs between lungs and neutrophils, we found that interestingly, many top DEGs appeared in both tissues in a genotype-specific manner. In Ager AHA/AHA , Gbp2b, Ifi44, and Gm28438 were highly downregulated in both lungs and neutrophils ( Figure 5D and G, Figure 5-figure supplement 3). In Ager −/− , Rps3a2, Rps3a3, Rps13_ps1, Rps15a_ ps7, Tpm3_rs7, Rf5, and BC147527 were differently regulated in both lungs and neutrophils to similar extent ( Figure 5E and H). Furthermore, two genes, Rpl34_ps1 and Gm42031, were found to be highly downregulated in both genotypes and in both tissues. As expected, in Ager −/− lung, Ager (the gene that expresses RAGE) is one of the top downregulated genes ( Figure 5H). Also consistent with our Western blot analysis ( Figure 1F), in Ager AHA/AHA lung, Ager was expressed at a similar level as in WT lung on RNA sequencing ( Figure 5-figure supplement 3F). GO analysis revealed genes related to regulation of cell killing (11 genes), canonical Wnt signaling (7 genes), and epithelial cell proliferation (11 genes), among others, was highly enriched in Ager −/− lung (Supplementary file 5). In contrast, none of these biological processes were enriched in Ager AHA/AHA lung due to limited number of DEGs.
Development of a rabbit mAb that binds specifically to the HS-binding site of RAGE As Ager AHA/AHA mice phenocopy Ager −/− mice in both osteoclastogenesis and response to APAPinduced liver injury, inhibition of HS-RAGE interaction could be an effective means to block RAGE signaling. To this end, we developed a rabbit monoclonal antibody (mAb B2) to specifically target HS-RAGE interaction. The epitope of B2 includes the HS-binding residues R216 and R218. Compared to the apparent binding affinity between B2 and WT RAGE (0.1 nM), the affinity between B2 and R216A-R218A mutant was dramatically reduced to 4 nM ( Figure 6A). Using a heparin Sepharose column, we confirmed that B2 was indeed able to significantly inhibit the binding of RAGE to heparin ( Figure 6B). Importantly, while a polyclonal anti-RAGE antibody significantly inhibited the binding between RAGE and its ligands HMGB1 and S100B, B2 did not interfere with the binding between RAGE and its ligands ( Figure 6C-D). The specificity of B2 was determined by immunostaining of lung sections from WT and Ager −/− mice. RAGE-specific staining was observed in WT lung section while no staining was shown on Ager −/− ( Figure 6E). Western blot analysis of whole lung lysate from WT, Ager AHA/AHA , and Ager −/− mice further showed that B2 was highly specific for RAGE but failed to detect the RAGE-AHA mutant, which again confirmed the epitope of B2 includes R216 and R218 ( Figure 6F).
Using a human breast cancer MDA-MB-453 cell line, which was known to overexpress RAGE (Nasser et al., 2015), we examined B2 binding to cell surface RAGE by FACS. As expected, B2 stains 453 cells extremely well, displayed a two-log shift compared to IgG control ( Figure 6G). Interestingly, when 453 cells were pretreated with heparin lyase III (HL-III) to remove cell surface HS, binding of B2 IgG was dramatically reduced by 20-fold ( Figure 6G). Because removal of HS would dissociate cell surface oligomeric RAGE back to monomers, we suspect that B2 preferably binds to dimeric RAGE in a bivalent fashion. Indeed, when Fab fragment of B2 was tested on 453 cells, we found that its cell surface binding was much weaker (due to monovalent interaction), and the binding showed minimal sensitivity to HL-III treatment ( Figure 6H).

B2 inhibits RAGE-dependent biological processes in cell and animal models
We first tested the inhibitory activity of B2 on RAGE-dependent osteoclastogenesis. BMMs were co-cultured with primary calvarial osteoblasts in the presence of 20 µg/ml of B2 or isotype-matched control antibody. As expected, cells treated with B2 formed fewer and smaller TRAP-positive cells, especially for cells with more than 10 nuclei ( Figure 7A-B). Remarkably, the osteoclastogenic phenotype of WT BMM treated with B2 was essentially identical to that of Ager AHA/AHA BMMs ( Figure 3C-D).
In monoculture osteoclastogenesis assay, where BMMs were induced with m-CSF and RANKL, B2 was equally effective in inhibiting osteoclast formation ( Figure 7C). The hepatoprotective effects of B2 were also examined in an APAP-induced acute liver failure model. In this experiment, mice were pretreated with B2 at 10 mg/kg before injection of sublethal dose of APAP. Compared to control IgG treated mice, B2 was able to reduce plasma ALT concentration by 40% and neutrophil infiltration by 50% ( Figure 7D-E). Again, the level of protection provided by B2 was highly similar to what Figure 6. Development of a rabbit mAb that targets specifically to the HS-binding site of RAGE. (A) Binding of rabbit mAb B2 to immobilized wildtype (WT) sRAGE or R216A-R218A mutant was determined by ELISA. The dramatically reduced binding affinity of B2 to R216A-R218A mutant (K d =4 nM) compared to its affinity to WT sRAGE (K d =0.1 nM) indicates that R216 and R218 are part of the epitope for B2. (B) B2 directly inhibits binding of RAGE to heparin. WT mouse RAGE VC1 domain (mVC1) were either directly loaded onto heparin Sepharose column, or loaded after 30 min incubation with Fab fragment of B2 at 1:1 molar ratio. B2-Fab bound mVC1 displayed greatly reduced binding to heparin column. (C) Binding of biotinylated sRAGE to immobilized HMGB1 was measured in the presence of anti-RAGE rabbit polyclonal Ab or B2 at 5 µg/ml. Binding of sRAGE to HMGB1 in the absence of antibodies was set to 100%. (D) Binding of biotinylated sRAGE to immobilized S100B was measured in the presence of anti-RAGE rabbit polyclonal Ab or B2 at 5 µg/ml. (E) Determine the specificity of B2 by immunostaining of lung sections from WT, Ager AHA/AHA , and Ager −/− mice with 1 µg/ml B2. (F) Western blot analysis of the specificity of B2 using lung lysate from WT, Ager AHA/AHA , and Ager −/− mice. Please note B2 failed to detect RAGE-AHA mutant, again confirming the epitope of B2. The loading ctrl is the same as shown in Figure 1F. The online version of this article includes the following source data for figure 6: Source data 1. Development of a rabbit mAb that targets specifically to the HS-binding site of RAGE.
Source data 2. Western blot analysis of the specificity of B2 using lung lysate from WT, Ager AHA/AHA , and Ager −/− mice.   Figure 4A-B). Collectively, these results validated our strategy of targeting the HS-binding site of RAGE as a means to inhibit RAGE signaling.

Discussion
While our previous work has shown that HS is a potent inducer of RAGE oligomerization and that HS-RAGE interaction is essential for RAGE signaling in endothelial cells (Xu et al., 2013), the physiological significance of HS-induced RAGE oligomerization remains unexplored due to a lack of genetic models. To address this gap, we generated a Ager AHA/AHA knock-in mice to specifically disrupt HS-RAGE interactions. As the HS-binding site and the ligand-binding site of RAGE are spatially separated (Xu et al., 2013), we were able to create a RAGE variant (R216A-R217H-R218A) with normal ligand binding, but impaired HS-binding capacity (Figure 1). This novel mouse model allowed us to dissect the specific contribution of HS-RAGE interaction to RAGE signaling in both physiological and pathological conditions for the first time. Zhou et al., 2006 first reported that Ager −/− mice display enhanced bone mineral density due to impaired osteoclastogenesis. Examination of the bone morphometric indexes of Ager AHA/AHA mice found that they also display an osteopetrotic bone phenotype in both trabecular and cortical bones, with a severity that was equivalent to Ager −/− mice ( Figure 2B). Natural history study of the bone phenotype found that for both strains, the bone was largely normal at 4 weeks old ( Figure 2D), suggesting the osteoclastogenesis during the bone modeling phase is not dependent on RAGE signaling. The osteopetrotic phenotype became full-fledged when mice reach 10 weeks old (for both sexes), and the phenotype remained when mice reach 4 months of age. It is interesting that while the bone phenotypes of Ager AHA/AHA and Ager −/− mice were indistinguishable at 10 weeks for both males and females, their phenotypes at 4 weeks and 4 months were somewhat different. It is unclear what factors contributed to this, but we suspect that the extensive transcriptomic changes experienced by Ager −/− mice ( Figure 5) might have modified the bone phenotype in an age-dependent manner. The fact that RAGE signaling regulates bone density in mature bones only, and that it regulates both trabecular and cortical bone homeostasis suggests that RAGE would be an attractive therapeutic target for treating diseases involving abnormal bone remodeling.
The fact that the Ager AHA/+ mice are haploinsufficient suggests that RAGE AHA mutant protein plays a dominant effect on WT RAGE. The dominant negative role of RAGE AHA mutant can only be explained in the light of RAGE oligomerization. In heterozygous Ager AHA/+ mice, the chance of assembling a functional dimer with intact HS-binding site would be merely 25% (50%×50%). This represents a 75% loss of functional RAGE signaling complex compared to WT mice, which apparently was sufficiently severe to impair normal osteoclastogenesis. In contrast, in Ager +/− mice, because all expressed RAGE are WT, they have a 100% of chance to form functional RAGE signaling complex. The fact that male Ager +/− mice displayed slightly elevated bone density (although not significant) could be attributed to somewhat reduced overall expression level of RAGE, which can be reasonably expected from heterozygous mice. Interestingly, this gene dosage effect was in full play in 10-week-old female mice, where the Ager +/− mice did show significantly increased bone volume (49%), which was identical to the increase observed in Ager AHA/+ mice (51%). In summary, our analysis has shown for the first time that mice are highly sensitive to the quantity of functional RAGE oligomeric complex, and that female mice appeared to be more sensitive to this regulation than males.
Our genetic manipulation of HS-RAGE interaction has provided clear evidence that disrupting HS-RAGE interaction pharmacologically would be an effective means to block RAGE signaling. To validate this novel concept, we developed a mAb (B2) that specifically targets the HS-binding site of RAGE. The epitope of B2 includes R216 and R218, the exact residues that we mutated in Ager AHA/AHA mice. Accordingly, B2 treatment of WT BMMs precisely copied the osteoclastogenesis phenotype of Ager AHA/AHA BMMs (Figures 7A-C , and 3C-D). Similarly, B2 treatment also provided (300 mg/kg). Plasma ALT concentrations (D) and neutrophils recruitment into to liver, n=8-9 (E) were quantified 24 hr post liver injury, n=8-9. TRAP, tartrate-resistant acid phosphatase; WT, wild-type.
The online version of this article includes the following source data for figure 7: Source data 1. B2 inhibits RAGE-dependent biological processes in cell and animal models.

Figure 7 continued
significant protection against APAP-induced liver injury, and the extent of protection in liver damage and suppression in neutrophil infiltration was highly similar to what was observed in Ager AHA/AHA mice ( Figures 7D-E , and 4). Combined, these data show that HS-RAGE interaction can be effectively targeted pharmacologically.
From a pharmacological point of view, targeting the HS-binding site of RAGE bears two main advantages over strategies targeting the ligand-binding site of RAGE. First, targeting the HS-binding site of RAGE can in principle block the signaling of all RAGE ligands because HS-dependent RAGE oligomerization is a common mechanism of RAGE signaling irrespective of the ligands involved. This is difficult to achieve for inhibitors that target the ligand-binding site of RAGE because the binding sites for different RAGE ligands are not identical (Koch et al., 2010;Xue et al., 2011;Ostendorp et al., 2007;Leclerc et al., 2007). Second, inhibitors that target the HS-binding site of RAGE leave the ligand-binding site of RAGE unoccupied, which would be free to interact with its multiple ligands that are overexpressed in disease states. Such inhibitors may thus turn RAGE into a decoy receptor, which is able to soak up ligands without transducing signal. In principle, this would greatly lower the risk of undesirable side effects by preventing spillover of unbound RAGE ligands to other inflammatory receptors such as TLR4, with which RAGE share many ligands (Yang et al., 2010;Vogl et al., 2007;Foell et al., 2013;Esposito et al., 2014).
By comparing transcriptomes of neutrophils and lung tissues from Ager AHA/AHA and Ager −/− mice, we present here clear evidence that complete deficiency of RAGE had much broader impact on global gene expression compared to point mutations of RAGE. Under the assumption that RAGE signaling is abolished in Ager AHA/AHA mice, which was true in both animal models we tested, we predict that most of the additional DEGs found in Ager −/− mice were perhaps not associated with RAGE signaling. Presumably, the expression levels of these genes were changed due to certain compensatory mechanisms for the sheer loss of RAGE protein. Because some of the changes in expression level were truly dramatic in both mRNA and protein level, it is highly plausible that some of these changes will manifest as phenotypes in Ager −/− mice in certain disease models, but in reality might have little to do with RAGE signaling. While it can be argued that the additional DEGs found in Ager −/− mice are due to HS-independent RAGE signaling, such position cannot be well defended for the following reason. If RAGE signaling really consists of HS-independent and HS-dependent RAGE signaling (which is abolished in Ager AHA/AHA mice), one would expect that the DEGs in Ager AHA/AHA mice are fully encompassed by the DEGs found in Ager −/− mice. However, this was not the case, especially for the downregulated genes (in both neutrophils and lungs, Figure 5C and F), where many more downregulated genes are found uniquely in Ager AHA/AHA mice than the ones that overlap with the downregulated genes in Ager −/− mice. Based on the scope of the inflammatory genes that are differently regulated in Ager −/− neutrophils (and most likely in other leukocytes as well), we believe caution will need to be taken when using Ager −/− mice in inflammatory models. In many cases, the Ager AHA/AHA mice might be a cleaner model for investigating the role of RAGE signaling.
The fact that most of the downregulated genes in Ager AHA/AHA mice are not found in Ager −/− mice is highly curious. Taking Gbp2b and Ifi44L as examples, the expression levels of both were completely abolished in neutrophils and lungs in Ager AHA/AHA mice, but they maintained WT-like expression levels in Ager −/− mice ( Figure 5-figure supplement 3). If Gbp2b and Ifi44L (both are interferon γ responsive genes) were truly associated with RAGE activation, the only logical explanation would be that the activation of these genes was somehow preserved in Ager −/− mice. Taking into considerations that many RAGE ligands can also cross-activate TLR4, it is plausible that in the absence of RAGE, certain RAGE ligands activate TLR4 instead, resulting in induction of Gbp2b and Ifi44L. In contrast, because RAGE AHA mutant behaves essentially like a decoy receptor, being able to still bind the ligand without transducing signal, spillover of RAGE ligands to TLR4 would be unlikely. Indeed, LPS has been shown to be able to induce Gbp2b (synonymous with Gbp1) expression, which suggests that in theory activation of TLR4 with another ligand would also induce Gbp2b expression (Kim et al., 2011). Ongoing research in our lab seeks to establish the connection between RAGE signaling and Gbp2b-mediated defense signaling.
In summary, by using Ager AHA/AHA mice, we presented strong evidence for an essential role of HS-RAGE interaction in bone remodeling and drug-induced liver injury. We further presented a new strategy in treating RAGE-associated diseases by blocking the HS-binding site of RAGE using mAb. Based on our RNA-seq study, we expect that HS-RAGE interaction would play essential roles in many inflammatory responses that involve other types of leukocytes. The Ager AHA/AHA mice and the anti-RAGE mAb antibody we reported here would certainly be invaluable tools to further our understanding of the pathophysiological roles of RAGE. To evaluate the therapeutic potential of mAb B2, we plan to determine its efficacy in a number of inflammatory and tumor models and is in the process of humanization of B2 for testing in primates. A limitation of our study is that we only focused on a small portion of the HS-binding site located on C1 domain. It is perceivable that mAbs that target other epitopes within the HS-binding site would also effectively block RAGE signaling and we are hoping to identify more such mAbs to fully realize the therapeutic potential of RAGE antagonism.

Continued
Generation of Ager AHA/AHA knock-in mice The knock-in project was performed at the Gene Targeting and Transgenic Shared Resource of Roswell Park comprehensive cancer center. Briefly, the synthesized sgRNA (which targets exon 6 of mouse rage gene as shown in Figure 1), repair template ssODN, and mRNA encodes for Cas9 were micro-injected into 1-cell stage fertilized egg of C57BL/6J strain, which were then surgically transferred into pseudopregnant foster mothers also of C57BL/6J background. Through CRIPSR-based homology-directed repair, the triple mutations were introduced into the rage locus. The pups were screened by next-generation sequencing, and the mice that bear correct mutations were mated to WT C57BL/6J mice to confirm germline transmission of the mutations. The F1 heterozygous mice were bred together to generate homozygous offspring. Ager −/− mice were originally gifted by A. Bierhaus (University of Heidelberg, Heidelberg, Germany).

MicroCT analysis
Mouse femurs from different ages of WT, Ager AHA/+ , Ager AHA/AHA , Ager +/− , and Ager −/− mice were harvested, fixed for 48 hr in 10% neutral buffered formalin. A quantitative analysis of the gross bone morphology and microarchitecture was performed using ScanCo microCT 100 system (University at Buffalo). 3D reconstruction and bone microarchitecture analysis were performed using AnalyzePro (AnalyzeDirect Inc).

TRAP staining
Mouse tibia from 10-week-old male mice were harvested, fixed for 48 hr in 10% neutral buffered formalin, and decalcified in 10% EDTA for 2 weeks at room temperature (RT). The samples were embedded in paraffin and sectioned at 5 μm for TRAP staining. The sections were counter stained with Fast Green.

Osteoclastogenesis assay
Primary osteoblasts were isolated from calvaria of 5-to 8-day-old WT mice following an established protocol (Taylor et al., 2014). Osteoblasts (5×10 3 cells/well) were seeded in a 96-well plate the day before starting the co-culture. Freshly isolated bone marrow cells (from WT, Ager AHA/AHA , and Ager −/− mice) were suspended in 10 ml of α-MEM containing 10% fetal bovine serum and 1× penicillin/streptomycin, 10 -7 M dexamethasone, and 10 -8 M 1α-and 25-dihydroxyvitamin D3. About 100 µl of bone marrow cells were added to each well. The medium was replaced every 2 days thereafter until the appearance of giant osteoclasts. For B2 inhibition assay, 20 µg/ml B2-IgG or control IgG was included in the medium in selected wells. To visualize osteoclasts, the cells were fixed and stained for TRAP activity using a Leukocyte Acid Phosphatase Kit (Sigma-Aldrich). Osteoclasts were sub-grouped into four categories based on the number of nuclei and quantified to analyze the population of osteoclasts with different number of nuclei.

Mineral apposition rate measurement
Eight-week-old male mice of WT, Ager AHA/AHA , and Ager −/− were labeled at 9 days (20 mg/kg of calcein) and 2 days (30 mg/kg of Alizarin Red) prior to sacrifice via intraperitoneal injection. Tibias were fixed with 10% formalin for 48 hr, treated with 5% KOH for 96 hr, and then processed and sectioned followed an established method (Porter et al., 2017). Images were taken with a Nikon Ci-S fluorescence microscope and merged using ImageJ software. Mineral apposition rate is the distance between the midpoints of the two labels (measured at multiple locations using ImageJ and averaged) divided by the time between the dye injections.

Mouse model of APAP liver injury
WT, Ager AHA/AHA , and Ager −/− mice were fasted overnight (12-15 hr) to deplete GSH stores before APAP (Sigma-Aldrich) administration. Fresh APAP was dissolved in warm (~50°C) sterile saline, cooled to 37°C, and injected intraperitoneally at 300 mg/kg. The plasma samples were collected 24 hr after APAP administrations for measuring ALT concentration. In general, the concentration of ALT reached its highest point at 24 hr after APAP overdose. For B2 protective treatment, WT mice were injected intraperitoneally with 10 mg/kg of B2-IgG 12 hr before APAP overdose (300 mg/kg). Plasma ALT was measured using the ALT Infinity reagent (Thermo Fisher Scientific) following the manufacturer's instructions. Livers were harvested, fixed for 24 hr in 10% neutral buffered formalin. Samples were further embedded in paraffin and sectioned at 4 μm for hematoxylin and eosin staining. The necrosis area was quantified with ImageJ.

Neutrophil infiltration in liver necrosis
The identical liver lobe from different mice was collected, minced, and digested with 0.2% collagenase A for 1 hr at 37° with shaking, and the digested tissue was further dissociated into single-cell suspension by pipetting several times. Cells passed through 70-µm cell strainer were pelleted, resuspended in 1 ml phosphate-buffered saline (PBS) and overlayed onto 4 ml of 33% Percoll gradient. After centrifugation at RT for 15 min at 1400×g, cell pellet was collected as leukocytes which include polymorphonuclear (PMN) cells. Leukocytes were further treated with ACK Lysing Buffer for a short time to remove red blood cells and counted. The number of neutrophils was quantified by FACS after staining with rat anti-mouse Ly-6G-PE, Clone 1A8 (BioLegend), rat anti-mouse CD45-FITC (BioLegend), and rat anti-mouse CD11b-APC (BioLegend) for 15 min at 4°. Ly-6G high cells were counted as neutrophils.

Air pouch model and neutrophil isolation
About 3 ml of sterile air was injected under the dorsal skin of WT, Ager AHA/AHA , and Ager −/− mice. After 3 days, the pouch was refilled with 2 ml of sterile air. On the seventh day, 1 μg of LPS from Escherichia coli O111:B4 (Sigma-Aldrich) in 0.5% solution of carboxymethyl cellulose (sodium salt; Sigma-Aldrich) was injected into the air pouch. After 4 hr, mice were euthanized by isoflurane inhalation and the cells in the pouch were collected by washing with PBS. Cells were counted and stained with Ly-6G-PE to determine the percentage of PMNs.

RNA-seq transcriptome analysis
LPS-induced leucocytes (>95% PMNs) were collected from air pouch as described above and used for RNA extraction. For lung tissue samples, 30 mg thoroughly perfused lungs were homogenized by bullet blender 5E (Next Advance). Total RNA was prepared using RNeasy Mini Kit (QIAGEN) following the protocol. cDNA library construction and RNA sequencing were performed at Novogene Co. Inc (Sacramento, CA). RNA integrity and quantitation were assessed using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, USA). A total amount of 1 μg RNA per sample was used as input material for the RNA sample preparations. Sequencing libraries were generated using NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, USA) following the manufacturer's recommendations. The clustering of the index-coded samples was performed on a cBot Cluster Generation System using PE Cluster Kit cBot-HS (Illumina) according to the manufacturer's instructions. After cluster generation, the library preparations were sequenced on an Illumina platform PE150 and paired-end reads (150 bp) were generated. Reference genome and gene model annotation files were downloaded from genome website browser (NCBI/UCSC/Ensembl, mm10) directly. Paired-end clean reads were aligned to the reference genome using Spliced Transcripts Alignment to a Reference (STAR) (Dobin et al., 2013).
DESeq2 R package (v1.30.1) was used for gene count normalization and differential expression analysis (Love et al., 2014). The list of Ager AHA/AHA and Ager −/− differentially expressed genes compared to WT was filtered by an absolute log 2 fold change≥1 and a multiple testing corrected p value ≤0.05. This list of differentially expressed genes in Ager AHA/AHA and Ager −/− lungs or PMNs were further analyzed and annotated with GO analysis using clusterProfiler R package (v3.18.1) (Yu et al., 2012). A multiple testing corrected p value ≤0.05 cutoff was used for the applied conditional hypergeometric test for GO term overrepresentation of GO terms in molecular function and biological process ontologies. The used GO annotations were obtained from the Bioconductor Mus musculus annotation package. PMN and lung RNA-sequencing data have been deposited into the NCBI Gene Expression Omnibus database (accession number GSE174178).

Development of B2
Rabbit monoclonal antibodies (mAbs) were raised against the murine RAGE V-C1 domain in its native conformation. The rabbit hybridomas were generated at Epitomics Inc (now part of Abcam). The hybridoma line 240 E-W3 was used to fuse with rabbit splenocytes. Out of the 4000 hybridoma clones that were screened, around 100 were able to detect mVC1 by direct ELISA. The supernatants of the positive clones were then screened for mAbs that specifically interacted with the HS-binding site. In this screening, instead of using WT mVC1, various mVC1 mutants (K43A-K44A, K39A-R104A, K107A, and R216A-R218A) were immobilized to ELISA plates. These mutants bear alanine mutations in the HS-binding residues. In this assay, mAbs were sought that show reduced binding to the mutants compared to the WT mVC1. In principle, the reduced binding would only occur if the residues that we mutated belong to the epitope recognized by the mAb. In this assay, the epitopes of B2 were found to include R216 and R218. Rabbit B2 was purified from hybridoma supernatant by Protein A chromatography.

Direct ELISA for screening RAGE-binding mAbs
About 200 ng of human RAGE extracellular domain or R216A-R218A mutant were immobilized onto a 96-well high-binding ELISA plate. Plates were blocked with 5% bovine serum albumin (BSA) in PBS and incubated with 0.01-10 nM of rabbit B2 mAb for 1 hr at RT. Bound B2 mAbs were quantitated with anti-rabbit-HRP (Jackson ImmunoResearch) followed by the addition of HRP substrate (Thermo Fisher Scientific).

Recombinant expression of B2
To produce B2 mAbs in large quantity for in vivo studies, we cloned the variable regions of B2 from hybridoma cells and grafted them onto human IgG1 constant region. The chimeric antibody was expressed in Chinese hamster ovary (CHO) cells. The pcDNA5_FRT plasmid purchased from Invitrogen was modified to insert the variable regions of B2 heavy and light chains in human IgG1 constant regions. Then the plasmid was transfected into CHO_Flp cells using lipofectamine 3000 Reagent (Invitrogen), and the positive clones were screened out over 10 days in the presence of Hygromycin B (50 mg/ml, Gibco). Cells with positive expression of the antibody were grown in CD CHO AGT expression medium (Thermo Fisher Scientific) with supplemental glutamine, D+ glucose, and Pen/Strep for 2 weeks. The media was then harvested, and the secreted antibody (IHH-IgG) was purified using a 5-ml HiTrap protein G column (Cytiva). The antibodies were buffer exchanged into PBS, and the concentration was determined using NanoDrop Spectrophotometer (Thermo Fisher Scientific). The purity of each batch was examined using SDS-PAGE. Purified antibodies were stored at 4°C before experiments. For B2-Fab, human chimeric B2 were digested with immobilized papain using Pierce Fab Preparation Kit (Thermo Fisher Scientific) following the instructions.

Effect of B2 on RAGE binding on heparin
About 30 µg of mouse mVC1 was either directly loaded onto heparin Sepharose column, or loaded after 30 min incubation with 68 µg Fab fragment of B2 (1:1 molar ratio). B2-Fab bound mVC1 displayed greatly reduced binding to heparin column. The bound mVC1 was eluted with a salt gradient from 200 mM to 1.4 M NaCl, pH 7.1 in HEPES buffer.

B2 cell surface binding by FACS
Breast tumor MDA-MB-453 was lifted from culture dish using Accutase (BioLegend) and incubated with human B2-IgG or B2-Fab in 100 μl PBS and 0.1% BSA for 45 min at 4°C. Bound B2-IgG was stained with goat anti-human IgG-Alexa 568 (1:1000; Invitrogen) for 30 min and analyzed by FACS. B2-Fab was stained with biotinylated goat anti-human IgG (1:1000, Invitrogen) followed by streptavidin-PE staining. In some experiments, cells were pretreated with recombinant heparin lyases III (5 milliunits/ml) for 15 min at RT prior to binding experiments. The MDA-MB-453 cell line was obtained from ATCC (HTB-131). Its identity was authenticated by SRT profiling, and it was tested negative for mycoplasma in the lab.

Immunohistochemistry
Lungs from WT and Ager −/− mice were perfused, harvested, and fixed for 24 hr in 10% neutral buffered formalin. Samples were further embedded in paraffin and sectioned at 5 μm for RAGE staining with 1 µg/ml human B2-IgG. After staining with biotinylated rabbit anti-human IgG secondary antibody (1:200), the sections were developed using the ABC system (Vector Laboratories) and the cell nuclei were counter stained with 15% Ehrlich's hematoxylin (Electron Microscopy Sciences).

Effect of antibodies on RAGE-ligand binding
HMGB1 or S100b (200 ng) was immobilized and the plate was blocked with 5% BSA. Biotinylatedmouse B2 V-C1 domain (200 ng/ml) was pre-incubated with rabbit IgG, polyclonal or monoclonal anti-RAGE (all at 5 μg/ml) for 30 min at RT before being added to the plate. The percentage of antibody inhibition was calculated by comparing the absorbance obtained in the presence of specific antibodies to the absorbance obtained in the presence of control rabbit IgG (which was set to 100%).

Expression of recombinant murine and human RAGE
WT mouse RAGE V-C1 domains (mVC1-WT) and human sRAGE (V-C1-C2 domains) were produced in E. coli cells. Purification was carried out using HiTrap SP cation exchange column at pH 7.8, followed by SEC on a Superdex 200 column (GE Healthcare). The recombinant protein was purified to >98% pure as examined by silver staining. Mouse RAGE triple mutant R216A-R217H-R218A (mVC1-AHA) was confirmed by sequencing and purified as described above for WT protein.

Heparin-Sepharose chromatography and analytical size-exclusion chromatography
To characterize the binding of mVC1 and mutant to heparin, 100 µg of purified mVC1-WT or mVC1-AHA were applied to a HiTrap heparin-Sepharose column and eluted with a salt gradient from 200 mM to 1.4 M NaCl, pH 7.1 (HEPES buffer). The conductivity measurements at the peak of the elution were converted to the concentration of NaCl based on a standard curve. For analyses of mVC1 and HS dodecasaccharide (H12) complex, purified mVC1-WT or mVC1-AHA (40 µg) was incubated with HS dodecasaccharide (H12, 4 µg) in 20 mM Tris, 150 mM NaCl, pH 7.4, at 4°C overnight. H12 is chemoenzymatically synthesized HS 12mer with NS, 2S, and 6S modification (gift from Dr. Jian Liu, UNC Chapel Hill). All complexes were resolved on a Superdex 200 (10/300 mm) gel filtration column using 20 mM Tris, 150 mM NaCl, pH 7.4, at 4°C.

HMGB1 binding ELISA
Binding affinity of mRAGE-WT and mRAGE-AHA to HMGB1 was measured by ELISA. Briefly, 96-well plate was coated by 200 ng of mRAGE-WT and mRAGE-AHA protein and blocked by 1% BSA. Biotinylated HMGB1 with concentrations from 50 ng/ml to 1000 ng/ml was added into wells and incubated for 2 hr followed by incubation with streptavidin-HRP for 30 min. About 50 µl of HRP substrate solution was added for developing, and the reaction was stopped by adding 50 µl of 1 M H 2 SO 4 . The absorbance at 450 nm was measured by a plate reader. Apparent K d value was calculated using Prism software.

Statistical analysis
All data are expressed as means ± SEM. Statistical significance of differences between experimental and control groups was analyzed by two-tailed unpaired Student's t-test, between multiple groups by one-way analysis of variance (ANOVA) followed by Dunnett's or Tukey's multiple comparisons test using Prism software (version 7.03; GraphPad Software Inc).