A single-dose Qβ VLP vaccine against S100A9 protein reduces atherosclerosis in a preclinical model

Abstract

The standard therapy for cardiovascular disease (CVD) is the administration of statins to reduce plasma cholesterol levels, but this requires lifelong treatment. We developed a CVD vaccine candidate that targets the pro-inflammatory mediator calprotectin by eliciting antibodies against the S100A9 protein. The vaccine, based on bacteriophage Qβ virus-like particles (VLPs) displaying S100A9 peptide epitopes, was formulated as a slow-release PLGA:VLP implant by hot-melt extrusion. The single-dose implant elicited S100A9-specific antibody titers comparable to a three-dose injection schedule with soluble VLPs. In an animal model of CVD (ApoE^−/− mice fed on a high-fat diet), the implant reduced serum levels of calprotectin, IL-1β, IL-6 and MCP-1, resulting in less severe aortic lesions. This novel implant was therefore able to attenuate atherosclerosis over a sustained period and offers a novel and promising strategy to replace the repetitive administration of statins for the treatment of CVD.

Graphical Abstract

We developed a cardiovascular disease vaccine candidate that targets the pro-inflammatory mediator calprotectin by eliciting antibodies against the S100A9 protein. The vaccine, virus-like particles displaying S100A9 peptide epitopes, was formulated as a slow-release PLGA:VLP implant. Using a preclinical model, we demonstrate that vaccination results in long-lasting and specific anti-S100A9 antibodies – lowering S100A9 levels as well as pro-atherogenic cytokines levels, and therefore limiting atherosclerosis development.

INTRODUCTION

Cardiovascular disease (CVD) is the leading cause of mortality worldwide, with ~18 million CVD-related fatalities in 2019 representing 31% of all deaths¹. The major underlying cause of cardiovascular events is atherosclerosis, a chronic inflammatory disease that develops in response to lipid accumulation, with an autoimmune component^2,3. The gold standard treatment for atherosclerosis is the administration of statins to reduce plasma levels of low-density lipoprotein-cholesterol (LDL-C)³. However, many atherosclerosis patients remain at risk due to chronic inflammation^4,5. This reflects the activity of pro-atherogenic cytokines such as interleukin (IL)-1β and IL-6, and chemokines such as monocyte chemoattractant protein-1 (MCP-1), which are secreted by macrophages, lymphocytes, natural killer cells, and vascular smooth muscle cells during the development of atherosclerosis^6,7. Several pro-inflammatory and pro-atherogenic cytokines have been targeted in CVD clinical trials, highlighting the relevance of anti-inflammatory therapy for the management of atherosclerosis⁶.

Another promising therapeutic target is the inflammatory biomarker calprotectin, which promotes cytokine secretion in multiple inflammatory diseases⁸. Calprotectin is a heterodimer of the calcium-modulated proteins S100A8 and S100A9, also known as myeloid-related protein (MRP)-8 and MRP-14, and is secreted mainly by neutrophils⁹. S100 family proteins regulate myeloid cell function, and control inflammation in part by activating Toll-like receptor-4 (TLR-4)¹⁰, the receptor for advanced glycation end products (RAGE)¹¹ and the extracellular matrix metalloprotease inducer (EMMPRIN)¹². Elevated levels of calprotectin are found systemically in patients with acute coronary syndromes (ACS), and also at the sites of coronary occlusion and in atherosclerotic plaques^13,14,15,16. In longitudinal studies, calprotectin is a positive marker of CVD and is independent of established risk factors¹⁷.

Transcriptional profiling has shown that S100A9, one of the components of the calprotectin heterodimer, is an important regulator of vascular inflammation and atherothrombosis¹⁴. The corresponding gene was therefore knocked out in the ApoE^−/− mouse model of atherosclerosis, resulting in attenuated atherosclerotic lesion formation and plaque inflammation when the ApoE^−/− S100A9^−/− double knockouts were fed on a high-fat diet, predominantly due to the regulation of vascular inflammation and vascular injury promoted by leukocyte recruitment¹⁸. Calprotectin-associated damage during atherosclerosis can therefore be ameliorated by targeting S100A9. Furthermore, the S100A9-null mice developed normal organs and tissues, and had a normal lifespan¹⁹, suggesting that S100A9 is a suitable target for CVD interventional therapy.

Passive immunization (monoclonal antibody therapy) against pro-atherogenic cytokines such as IL-1β²⁰ and TNF-α²¹ was shown to reduce the risk of cardiovascular events. Furthermore, the US Food and Drug Administration recently approved the monoclonal antibodies alirocumab and evolocumab targeting the proprotein convertase subtilisin/kexin-9 (PCSK9), a cholesterol metabolism checkpoint protein^22,23,24. However, passive immunotherapy requires repeated dosing and lifelong treatment, which is too expensive for many patients²⁵. Active immunotherapy (vaccination) is a cost-effective approach that should achieve similar protective effects. Atherosclerosis vaccines should therefore focus on one of two major objectives: (1) lowering serum LDL-C levels, or (2) attenuating the chronic inflammatory response²⁶.

To address the second objective, we developed a CVD vaccine targeting the S100A9 protein to reduce serum levels of calprotectin. The success of any vaccine requires the use of a suitable antigen combined with a strong adjuvant and an effective delivery strategy. We therefore used virus-like particles (VLPs) from bacteriophage Qβ to display a B-cell epitope from the mouse S100A9 protein (Figure 1). VLP vaccines have shown efficacy against other self-antigen epitopes (ApoB, CETP and PCSK9)²⁷ as well as pathogen epitopes from human papillomavirus (HPV)²⁸ and SARS-CoV-2²⁹. Qβ VLPs are macromolecular display platforms for foreign epitopes attached by conjugation or genetic fusion, and they act as a built-in adjuvant to enhance the stability of antigens and enable lymphatic trafficking and processing by antigen presenting cells^30,31. Qβ VLPs can be produced in large quantities by the expression of a recombinant Qβ 14-kDa capsid protein in Escherichia coli (in vivo assembly), and scalable production according to current good manufacturing practices (cGMP) has been demonstrated for several Qβ-based VLP vaccine candidates in clinical trials³¹. To prolong the delivery of the vaccine, we opted for a slow-release poly(lactic-co-glycolic acid) (PLGA) implant, which was mixed with the VLP vaccines using hot melt-extrusion. This process does not compromise the structural properties or immunogenicity of the vaccine^{28,29,32,33,34}. We tested the efficacy and safety of the vaccine in a traditional prime-boost-boost schedule vs the single-dose slow-release injectable implant in healthy mice and in the ApoE^−/− model of atherosclerosis by measuring antibody titers and immune responses, plasma levels of calprotectin, IL-1β, IL-6, and MCP-1, and the severity of aortic lesions in the aortic arch and thoracic aorta.

Figure 1.

Similarity between the human and mouse S100A9 proteins. A) Structural models (PDB 6DS2 and 6ZDY). The C-terminal target epitope (residues 101–110) is highlighted in magenta. The structures were rendered using UCSF Chimera³⁷. B) Alignment of the human (orange, Uniprot ID P06702) and mouse (black, Uniprot ID P31725) S100A9 amino acid sequences, revealing 58% identity and 74% similarity. Blue boxes indicate fully conserved positions and green boxes indicate chemically similar residues. The two metal-binding sites (S1 and S2) are shown as red and black arrows, respectively³⁸. The sequences were aligned using Clustal Omega³⁹.

RESULTS AND DISCUSSION

Characterization of QβS100A9 particles

We chose S100A9 epitope ₁₀₁RGHGHSHGKG₁₁₀ (Figure 1) based on its score (0.5 threshold) as a B-cell epitope as determined using the BepiPred-2.0 Sequential B-Cell Epitope Predictor³⁵. In a previous study using a similar approach to treat thrombosis³⁶, the authors used epitope ₁₀₄GHSHGKGCGK₁₁₃, which includes amino acids 111–113 (CGK), but we found that those residues did not pass the score threshold and we excluded them. The epitope ₁₀₁RGHGHSHGKG₁₁₀ was fused to the C-terminus of the Qβ coat protein (CP) with an intervening GSG linker, and was expressed in a vector also containing the unmodified Qβ CP gene to allow the assembly of hybrid particles (Figure 2A) as previously described^27,35. The QβS100A9 vaccine candidate was expressed in E. coli and purified by sucrose gradient ultracentrifugation, with yields of ~360 mg unmodified Qβ and ~195 mg QβS100A9 per liter of culture. The lower yield of hybrid particles has been observed before, and probably reflects the negative impact of the additional peptide on VLP assembly²⁷.

Figure 2.

Production and characterization of QβS100A9 VLPs. A) Schematic (not to scale) of the unmodified Qβ capsid protein (CP) and the QβS100A9 fusion CP with S100A9 epitope ₁₀₁RGHGHSHGKG₁₁₀ and an intervening GSG linker, and a map of expression vector pCOLA_Qβ_QβS100A9 with both genes driven by independent T7 promoters and a single T7 terminator (lacI = lactose repressor gene, ColA ori = ColA origin of replication, KanR = kanamycin resistance gene). B) SDS-PAGE analysis of unmodified Qβ (left) and QβS100A9 VLPs (right) under reducing conditions. Unmodified Qβ CP (CP ~14 kDa) and modified QβS100A9 (S100A9-CP ~15.2 kDa) were visualized after staining with GelCode Blue Safe protein stain. C) The number of peptides displayed per hybrid QβS100A9 VLP was determined by densitometry. QβS100A9 VLPs were then used in two delivery formulations (soluble VLPs and slow-release implants). D) DLS of unmodified Qβ and QβS100A9 VLPs shows monodisperse particles. Z-average (dm) = 32.4 ± 0.5 for unmodified Qβ and 39.2 ±0.2 for QβS100A9. PDI = 0.11 for unmodified Qβ and 0.01 for QβS100A9. E) FPLC analysis of unmodified Qβ and QβS100A9 VLPs display similar elution curves. F) TEM images of negatively-stained unmodified Qβ and QβS100A9 VLPs (scale bar = 100 nm). All samples were tested in triplicate (n = 3) for all characterization methods, and the pictures and data shown are representatives of those triplicates.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis (Figure 2B) confirmed that only wild-type CP (~14 kDa) was present in the unmodified Qβ VLPs, whereas the hybrid particles contained both the wild-type CP and the larger QβS100A9 CP (~15.2 kDa). Densitometric comparison revealed that the hybrid particles were composed of 42–50% QβS100A9 CPs, indicating that 77–90 peptide epitopes are displayed per hybrid VLP (180 CP subunits in total) (Figure 2C). Dynamic light scattering (DLS) revealed the particles were monodisperse with a hydrodynamic diameter (Z-average) of ~32 nm for unmodified Qβ VLPs and ~39 nm for QβS100A9 VLPs (Figure 2D). Fast protein liquid chromatography (FPLC) revealed slightly overlapping major single peaks for unmodified Qβ and QβS100A9, with a similar elution volume of 10–15 mL indicating that both types of VLP were intact and did not form significant or apparent aggregates (Figure 2E). The unmodified Qβ VLPs had a higher polydispersity index (PDI) of 0.11 compared to 0.01 for the QβS100A9 VLPs, indicating that the QβS100A9 particles were larger but more uniformly distributed, as reported for other Qβ-based vaccines²⁷. Transmission electron microscopy (TEM) of negatively-stained particles indicated that the Qβ and QβS100A9 VLPs were intact (Figure 2F). These results confirm that the QβS100A9 VLPs are stable and the peptide does not affect particle integrity.

Immunogenicity of soluble QβS100A9 VLPs and slow-release implants

We previously reported slow-release formulations for vaccines against HPV, COVID-19 and cholesterol-related proteins associated with CVD, and we developed protocols for hot-melt extrusion yielding degradable PLGA:Qβ implants with a ~1 month release timeframe that achieved the same immunogenicity as a multi-dose regimen of soluble vaccines^27,28,29. To confirm the immunogenicity of the VLPs before formulation, we compared the soluble QβS100A9 VLPs to the free peptide epitope in a traditional subcutaneous (s.c.) vaccination schedule consisting of a prime plus two boosts, 2 weeks apart. Each injection consisted of 100 μg QβS100A9 VLPs or 5 μg of the free peptide, resulting in a molar equivalent dosage of the epitope in each case (Figure 3A). As expected, the QβS100A9 VLPs elicited significantly higher specific IgG titers than the free peptide over a period of 12 weeks (Figure 3B), confirming the adjuvant activity of the VLPs⁴⁰.

Figure 3.

Immunization of C57Bl6J mice with QβS100A9 vaccines. A) Mice were immunized with 100 μg per injection of the soluble QβS100A9 vaccine or 5 μg of the free peptide (control) in a traditional prime plus two boosts schedule. B) ELISA against S100A9 peptide from the soluble vaccine showing endpoint IgG titers. C) ELISA against S100A9 peptide from the soluble vaccine showing absorbance of IgG subclasses and Ig isotypes over time. D) T-cell helper (Th)-biased profile based on the IgG1/IgG2b ratio of the soluble vaccine. E) Mice were immunized with a PLGA implant containing 300 μg QβS100A9 VLPs (equivalent to the total amount of soluble VLPs) or non-modified Qβ VLPs (control). F) ELISA against S100A9 peptide from implant vaccine showing endpoint IgG titers. G) ELISA against S100A9 peptide from implant vaccine showing absorbance of IgG subclasses and Ig isotypes over time. H) T-cell helper (Th)-biased profile based on the IgG1/IgG2b ratio of the implant vaccine. Data are means ± SEM (n = 5 biological samples, unpaired two-tailed t-test with 95% confidence value, *p < 0.05)

Next, we produced slow-release implants (80% PLGA, 10% VLPs and 10% PEG8000) containing either unmodified Qβ (control) or QβS100A9 VLPs. Hot-melt extrusion yielded injectable implant rods. The implants contained 300 μg of VLPs to match the combined dose of the prime plus two boosts schedule (Figure 3E). As expected, the IgG titers against the S100A9 epitope were significantly higher in the PLGA/QβS100A9 VLP group than the control VLP group over a period of 20 weeks (Figure 3F). The IgG titers in the PLGA/QβS100A9 implant VLP group were similar to those in the soluble QβS100A9 VLP group (~10⁴), confirming that a single dose of the PLGA/QβS100A9 implant is effective (Figure 3B,F). From a clinical perspective, the single-dose implant favors patient compliance and reduces the healthcare burden by offering the potential for long-acting vaccines whose release rate can be tailored based on the polymer composition⁴¹.

We next evaluated the immunoglobulin subtypes elicited by the vaccine formulations. IgG2b is a subclass of IgG induced primarily by Th1-type cytokines such as interferon gamma (IFN-γ), whereas IgG1 antibodies are induced by Th2-type cytokines such as IL-4^40,42. The soluble VLPs resulted in an IgG1/IgG2b ratio < 1 that persisted over 12 weeks, indicating a Th1-biased profile (Figure 3D). In contrast, the implant resulted in a marked Th1 bias initially, but this trended toward a balanced Th1/Th2 response by week 15 (Figure 3H). We also observed that IgM titers declined over time following vaccination with soluble VLPs but increased over time following vaccination with the implant, indicating that the slow release maintains a constant boosting effect that influences the levels of IgM and other immunoglobulins (Figure 3C,G). We have previously observed sustained high levels of IgM with other vaccines formulated as implants²⁹.

Given that S100A9 belongs to a large family of S100 proteins with similar structures and shared domains³⁶, we selected an epitope unique to S100A9 to avoid off-target effects. To confirm the specificity of the antibodies elicited by the QβS100A9 vaccine, we tested their ability to bind S100A8, S100A9 and S100A8/9 heterodimers in dot blot assays. As expected, plasma from QβS100A9-vaccinated mice only recognized S100A9 and the heterodimer S100A8/9, but not the S100A8 protein (Figure 4). Antibodies from mice injected with the free peptide generated weaker signals but we observed the same overall result (Figure 4).

Figure 4.

Specificity of antibodies elicited against the S100A9 epitope displayed on VLPs or as a free peptide. A) Schematic of dot blot immunodetection of S100A8, S100A9, and calprotectin (S100A8/9). B) Pooled plasma (dilution 1:100) from mice vaccinated with QβS100A9 VLPs (right column) or the free peptide (left column) was tested against mouse recombinant S100A8, S100A9 and heterodimer S100A8/9 (calprotectin) by dot blot (1 μg/dot and 3 dots per membrane). Plasma from mice vaccinated with QβS100A9 recognized the S100A9 protein and its heterodimer S100A8/9, but not the S100A8 protein. Black arrows indicate positive staining. No or barely detectable signal was observed from free-peptide plasma group, as expected.

Safety of soluble QβS100A9 VLPs and slow-release implants

S100A9 is a self-antigen, so immunotoxicity parameters such as the absence of a target-specific T-cell response (Th and cytotoxic T cells) are desirable^43,44,45. We therefore used ELISpot assays to evaluate the activation of primed T cells following vaccination with soluble QβS100A9 VLPs or the implant. We monitored the production of IFN-γ (linked to Th1-biased profiles) and IL-4 (linked to Th2-biased profiles) in splenocytes isolated 4 weeks after the second 100-μg dose of soluble QβS100A9 or a single 200-μg dose of the QβS100A9 implant, and also in splenocytes from naïve mice.

IFN-γ was significantly more abundant in the splenocytes of vaccinated (soluble or implant) and naïve mice following stimulation with recombinant mouse S100A9, but not following stimulation with the free peptide (Figure 5A–C). The detection of this response even in naïve splenocytes indicates that the S100A9 monomer and S100A8/9 heterodimer act as damage associated molecular patterns (DAMPs) via their ability to bind the RAGE, TLR-4 and EMMPRIN receptors, leading to MAPK signaling and the translocation of NF-κβ from the cytosol to the nucleus, triggering the activation of genes that control inflammation^46,47. This result makes it difficult to determine whether the presence of T cells primed with the S100A9 epitope could exacerbate the T cell response because even naïve splenocytes are activated in the presence of S100A9. However, the IFN-γ SFCs were more abundant on splenocytes from mice vaccinated with the soluble VLPs, which agrees with the persistent Th1-biased profile revealed by the IgG1/IgG2b ratio (Figure 3C,D). On the other hand, the IFN-γ SFCs observed on splenocytes from mice vaccinated with the implant (Figure 5B) showed a response comparable to that of naïve splenocytes, reflecting the balanced Th1/Th2 profile (Figure 3G,H).

Figure 5.

Plasma biomarkers and ELISpot assays for the analysis of vaccine safety. (A-C) ELISpot assays using splenocytes from mice vaccinated with (A) soluble QβS100A9 VLPs or (B) PLGA-based QβS100A9 implants compared to (C) splenocytes from naïve mice. The splenocytes from each vaccinated (n = 3) or naïve (n = 2) group were cultivated at least in duplicate with medium (negative control), free peptide, whole S100A9 protein, unmodified Qβ VLPs, or PMA/ionomycin (positive control). D) Three plasma biomarkers related to liver injury (AST and ALT) and kidney injury (KIM-1) were evaluated at week 0 (baseline) and week 12 after vaccination (n = 4 from pooled plasma). Dashed lines in the AST and ALT plots indicate the threshold for normal healthy mice. Data are means ± SEM (unpaired two-tailed t-test with 95% confidence value, *p < 0.05).

The stimulation of splenocytes with unmodified Qβ VLPs resulted in a significant increase in IFN-γ compared to the non-stimulated control (medium only), indicating a strong response to the VLP scaffold^27,29. The phorbol 12-myristate 13-acetate (PMA)/ionomycin positive control was the only stimulant that triggered the significant production of both IFN-γ and IL-4 by splenocytes from all vaccinated and naïve mice. Accordingly, we decided to study the PLGA-based QβS100A9 implant in more detail, given its more balanced Th1/Th2 response and its IgG titers comparable to the soluble VLPs. Most Qβ VLP vaccines developed by us²⁷ and others⁴⁴ triggered a Th2-biased response, but this is likely to be epitope-dependent given the Th1-biased response reported for a Qβ VLP self-antigen vaccine against epitopes derived from different loops of the C3 domain of IgE that binds to the high-affinity FcεRI receptor⁴⁸. The Th1-biased response was dependent on TLR-7 activation because TLR-7 knockout mice switched to a Th2-biased profile⁴⁸. We found that the Th response can be modulated by delivering the vaccine as a slow-release implant, probably reflecting the constant release of small quantities of VLPs during implant biodegradation.

We also assessed the safety of the vaccines by measuring the concentration of kidney injury molecule 1 (KIM-1), a transmembrane glycoprotein and biomarker of kidney injury⁴⁹, as well as the plasma activity of aspartate aminotransferase (AST) and alanine transaminase (ALT), which are biomarkers of liver injury⁵⁰. We compared mice vaccinated with QβS100A9 VLPs vs controls after 12 weeks, but observed no significant changes on the normal range⁵¹ of KIM-1 or AST/ALT activity (Figure 5D). We can therefore confirm that the QβS100A9 VLPs neither altered kidney physiology nor caused liver injury.

Efficacy of QβS100A9 slow-release implants against atherosclerosis

Having confirmed the immunogenicity and safety of the QβS100A9 VLPs and implants in healthy animals, we tested the efficacy of the QβS100A9 implant in ApoE^−/− mice fed on a high-fat western diet as a model of atherosclerosis. We selected the implant because it can be administered as a single dose but it elicits antibody titers similar to the soluble VLPs while generating a more balanced Th1/Th2 profile. Mice were immunized once with the QβS100A9 implant or the unmodified Qβ implant as a control and were monitored for 24 weeks, the first 4 weeks on a regular diet before switching at the beginning of week 5 to the high-fat diet (Figure 6A). S100A9-specific IgG appeared during week 2, and titers remained stable until week 8 and then increased slightly between weeks 8 and 12 (Figure 6B). As expected, the QβS100A9 group elicited significantly higher titers than the control group at all time points. We assessed the IgG1/IgG2b ratio and observed a similar profile to that observed for healthy mice. In more detail, we observed a Th1-biased profile (IgG1/IgG2b < 1) for all except one of the mice at week 4. Most profiles remained the same at week 8, except one mouse with a balanced Th1/Th2 profile and two trending toward a Th2-biased response. Finally, three of the mice showed a Th2-biased response by week 24, two showed a balanced Th1/Th2 response, and five retained their Th1-biased profile (Figure 6C). Additionally, as observed in the healthy animals, IgM levels remained high throughout the experiment (Figure 6D). These data confirmed that the QβS100A9 implant behaves similarly in healthy and atherosclerotic mice.

Figure 6.

Immunogenicity of the QβS100A9 vaccine implant in a mouse model of atherosclerosis fed on a high-fat western diet. A) Vaccination and diet schedule for ApoE^−/− male mice (n = 10 per group). Two groups were tested, one vaccinated with the QβS100A9 implant (300 μg total dose) and a control group vaccinated with an equivalent dose of unmodified Qβ VLPs. B) Anti-S100A9 peptide antibody titers at different time points after implantation. C) IgG1/IgG2b ratios at four time points to characterize the Th profile (ratio < 1 = Th1-biased, ratio > 1 = Th2-biased). The Th profile shifted from Th1-biased to balanced Th1/Th2 over time. D) Complete immunoglobulin profile over time. Data are means ± SEM (n = 10, unpaired two-tailed t-test with 95% confidence value, *p < 0.0001).

The QβS100A9 implant ameliorates aortic lesions by reducing the plasma levels of calprotectin and pro-inflammatory cytokines

We measured the degree of plaque formation on the aortas of mice fed on the high-fat western diet by standard oil red O staining, which detects neutral fat, fatty acids and triglycerides⁵² that accumulate in atherosclerotic plaques but not in the healthy endothelium. The positive staining is directly correlated to atherosclerosis and can be quantified as the percentage of lesion in the aortas (Figure 7C and Supp. Figure 1). We found that the percentage of lesion in the QβS100A9 vaccine implant group (21.5 ± 2%) was 32% lower than that in the control group (31.6 ± 3%). This is noteworthy because double knockout animals (ApoE^−/− S100A9^−/−) fed on a high-fat diet showed a 25–30% reduction in lesion area in the thoracic and abdominal aorta¹⁸, suggesting that blocking S100A9 activity either by gene knockout or vaccination may reduce the level of S100A9 in the plasma, thus preventing calprotectin from signaling via TLR-4⁵³ and modulating calcium signaling¹⁹, reorganizing the cytoskeleton⁵⁴ and triggering the release of pro-inflammatory cytokines.

Figure 7.

Effect of QβS100A9 vaccine implants on a mouse model of atherosclerosis fed on a high-fat western diet. A) Body weight (grams) determined weekly. B) Total cholesterol in plasma determined at four time points. C) The presence of atherosclerotic plaques shown as the percentage of lesion determined by oil red O staining. Representative aortic arch and thoracic aorta images from QβS100A9 vaccinated and control groups are shown below, with red areas indicating the atherosclerotic plaques. D-G) Levels of key biomarkers in plasma before (week 0) and after vaccination (weeks 8, 12 and 24): D) calprotectin, E) monocyte chemoattractant protein-1 (MCP-1), F) interleukin-6 (IL-6), and G) interleukin-1β (IL-1β). Data are means ± SEM (n= 10), unpaired two-tailed t-test, 95% confidence value, p < 0.05 was considered the threshold for statistical significance vs control group.

To determine the mechanism of action, we measured the levels of calprotectin in the plasma. We found significantly lower levels in the QβS100A9 group compared to controls at all time points, reaching a maximum differential of 3.3-fold during week 12 (Figure 7D). In addition to atherosclerosis and CVD^13,14,17,55, calprotectin is an important biomarker for the diagnosis and monitoring of many other inflammatory diseases⁸ because it stimulates the production of pro-inflammatory cytokines/chemokines such as IL-1β⁵⁶, IL-6 and MCP-1⁵⁷. IL-1β is key mediator of the pro-inflammatory response. Higher levels of this cytokine are found in atherosclerotic human arteries compared to healthy controls and its abundance positively correlates with disease severity⁵⁸. IL-1β activates secondary inflammatory mediators such as IL-6 and triggers pro-coagulant activity, the expression of adhesion molecules required for leukocyte recruitment, and the production of MCP-1^59,60. Together these changes promote the recruitment of monocytic phagocytes, which are strongly implicated in atherogenesis⁶¹. IL-1β deficiency attenuated the spontaneous development of atherosclerotic lesions in ApoE^−/− mice⁶². We found that the lower levels of calprotectin in the QβS100A9 group correlated with lower levels of IL-1β, IL-6 and MCP-1 (Figure 7E–G). These cytokines/chemokines were also depleted in ApoE^−/− S100A9^−/− double knockout mice, correlating with the less severe aortic lesions¹⁸.

The precise mechanism by which the S100A9-specific antibodies reduce the severity of lesions is unclear. When a similar epitope was used to treat thrombosis in mice (₁₀₄GHSHGKGCGK₁₁₃, from mouse S100A9) and monkeys (₁₀₂GHHHKPGLGE₁₁₁, from monkey S100A9), the promising results were attributed to the inhibition of S100A9/CD36 signaling in platelets^36,63. Many pro-inflammatory cytokines are triggered by S100A9/TLR-4 signaling, but a recent study showed that 73–85 amino acids in S100A9 specifically interact with TLR-4 and trigger TNF-α expression⁵³. Our data show that the depletion of calprotectin is sufficient to reduce the secretion of pro-inflammatory cytokines/chemokines and is therefore protective against aortic lesions, without knowing which S100A9-dependent signaling pathway is blocked by the antibodies.

Interestingly, the QβS100A9 implant did not affect plasma levels of calprotectin in healthy animals on a regular diet (Supp. Figure 2) but prevented the levels from increasing when these animals were fed on the high-fat diet (Figure 7D). This is important because calprotectin also protects against pathogens⁶⁴, and it would be undesirable to completely remove a beneficial defense mechanism. Accordingly, we measured other health-related parameters beyond the severity of aortic lesions and cytokine levels. We observed no significant differences in body weight between the treatment groups at any point during the experiment, which was anticipated because the vaccine does not affect food intake (Figure 7A). However, around the middle of the study (week 12) we observed a significantly higher amount of total cholesterol in the QβS100A9 vaccine group (1130 ± 44 mg/dL) compared to the control (996 ± 31 mg/dL), but this relationship switched by week 24, with the QβS100A9 vaccine group (878 ± 60 mg/dL) showing a significantly lower amount of total cholesterol than the controls (1158 ± 94 mg/dL) (Figure 7B). The relationship between systemic S100A9 and cholesterol homeostasis is unclear, and would be an important subject for further investigation. We found no significant difference in the weights of individual organs between the QβS100A9 vaccine group and controls (Supp. Figure 3).

Conclusion

In summary, our single-dose QβS100A9 vaccine formulated as a PLGA:VLP implant successfully reduced the extent of aortic lesions in a model of atherosclerosis, most likely by depleting calprotectin and thus preventing the secretion of pro-inflammatory cytokines/chemokines such as IL-1β, IL-6 and MCP-1. This was achieved without apparent systemic damage or adverse autoimmune responses. Calprotectin remained at normal basal levels in healthy mice. The effect of the vaccine was to deplete calprotectin only in the disease state, when unvaccinated animals experienced elevated levels. This promising anti-atherosclerosis vaccine offers a novel approach for the management of CVD and other inflammatory diseases and should now be tested in larger mammals and human clinical trials.

METHODS

Production of Qβ VLPs.

Bacteriophage Qβ VLPs were expressed as previously reported^27,29,65. Genes encoding the wild-type Qβ CP (NCBI accession: P03615) and CP fused to the mouse S100A9 epitope ₁₀₁RGHGHSHGKG₁₁₀ (NCBI accession: P31725) were codon optimized for E. coli and inserted into the expression vector pCOLA-DUET1 by GenScript Biotech. A linker (GSG) was placed between the C-terminus of the CP and the N-terminus of the peptide. The final vector was named pCOLA_Qβ_QβS100A9. As a control for some experiments, we used vector pCDF_Qβ carrying only the wild-type Qβ CP gene²⁷. E. coli Bl21 (DE3) cells (New England BioLabs) transformed with pCDF_Qβ or pCOLA_Qβ_QβS100A9 were grown at 37 °C for 16 h shaking at 250 rpm in 10 mL MagicMedia (Invitrogen) with the appropriate antibiotics: 25 μg/mL streptomycin (Sigma-Aldrich) for pCDF_Qβ and 50 μg/mL kanamycin (Sigma-Aldrich) for pCOLA_Qβ_QβS100A9. The culture was scaled up to 200 mL in the same medium and incubated at 37 °C for 20 h, shaking at 300 rpm. The cells were pelleted by centrifugation (5000 × g, 20 min, 4 °C) and frozen at −80 °C overnight. The pellet was then lysed by resuspending it in 10 mL lysis buffer (GoldBio) per gram of wet mass, adding a lysis cocktail comprising 1 mg/mL lysozyme (GoldBio), 2 μg/mL DNase (Promega) and 2 mM MgCl₂, and incubating at 37 °C for 1 h before sonicating at 30% amplitude for 10 min on ice, with 5-s pulses interspersed with 5-s gaps. The lysate was centrifuged (5000 × g, 30 min, 4 °C) and the clear supernatant was set aside. Unmodified Qβ VLPs and hybrid QβS100A9 VLPs were precipitated by adding 10% (w/v) PEG8000 (Thermo Fisher Scientific) at 4 °C for 12 h. The precipitated fraction was pelleted by centrifugation (5000 × g, 10 min, 4 °C) and dissolved in phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH 7.4) and then extracted with 0.5 volumes of 1:1 (v/v) butanol/chloroform. The aqueous fraction containing VLPs was separated by centrifugation (5000 × g, 10 min, 4 °C) and pure VLPs were recovered by 10−40% sucrose velocity gradient ultracentrifugation (9,6281 × g, 2.5 h, 4 °C). The light-scattering VLP band was collected and pelleted by ultracentrifugation (16,0326 × g, 2 h, 4 °C) and the pure VLPs were resuspended in PBS and stored at 4 °C until further use.

Characterization of Qβ VLPs.

The VLPs were characterized as previously described^27,28,29. The VLP concentration was determined using a Pierce BCA assay kit (Thermo Fisher Scientific). To confirm hybrid QβS100A9 assembly and peptide display, 10 μg of QβS100A9 particles was analyzed by SDS-PAGE under reducing conditions on NuPAGE 12% Bis-Tris protein gels (Thermo Fisher Scientific) stained with GelCode Blue Safe protein stain (Thermo Fisher Scientific). The gel images were acquired using the ProteinSimple FluorChem R imaging system, and densitometry was used to determine the number of peptides displayed per hybrid QβS100A9 VLP. The integrity of VLPs was confirmed by TEM using a FEI Tecnai Spirit G2 BioTWIN instrument to examine samples stained with 2% uranyl acetate. FPLC was carried out using an AKTA-FPLC 900 system fitted with Superose 6 Increase 10/300 GL columns (GE Healthcare) using PBS as the mobile phase. Particle size was confirmed by DLS using a Malvern Instruments Zetasizer Nano at 25 °C and plastic disposable cuvettes.

Hot-melt extrusion of VLP-loaded implants.

Implants were prepared from PLGA (Akina; LG ratio = 50:50, molecular weight = 10−15 kDa) using our previously reported desktop melt-processing system^{27,28,29,32,33,34}. Briefly, lyophilized QβS100A9 or Qβ VLPs were formulated with the ratio 80% PLGA, 10% VLP and 10% PEG8000 (by weight). The dry components were vortexed and loaded into the hot melt-processing system. The barrel was heated to 70 °C for 90 s, and the piston was set to 10 psi (69 kPa) for extrusion. Implants were stored at room temperature with a desiccant until use.

Immunization of mice.

All animal experiments were approved by the UC San Diego Institutional Animal Care and Use Committee (assurance number D16-00020, protocol number S18021). Eight-week-old female C57BL/6J mice (Jackson Laboratory, #000664) were kept under controlled conditions with food and water provided ad libitum. Five animals were assigned to the following four groups: (1) soluble QβS100A9, (2) soluble free peptide (control), (3) QβS100A9 implant, and (4) Qβ implant (control). For soluble formulations, we injected 100 μg of QβS100A9 VLPs or 5 μg of free peptide epitope (synthesized by GenScript Biotech) s.c. in 100 μl of PBS every 2 weeks, making three doses in total (prime + two boosts). The PLGA-based implants were cut into lengths of 0.3–0.5 cm according to their weight to provide 300 μg of QβS100A9 vaccine (to match the 300 μg total dose of the soluble formulation) or unmodified Qβ as control. Implants were placed using an 18G needle (BD Biosciences) s.c. behind the neck. Blood samples were taken by tail bleeding at week 0 (before vaccination) and at multiple time points thereafter. Plasma was separated in lithium/heparin-treated tubes (Thomas Scientific) by centrifugation (2000 × g, 10 min, room temperature) and the plasma was stored at −80 °C.

Peptide-specific IgG titers.

Endpoint total IgG titers against the S100A9 peptide epitope were determined by enzyme-linked immunosorbent assay (ELISA)^27,29,34. We coated 96-well, maleimide-activated plates (Thermo Fisher Scientific) with 2.5 μg cysteine-modified S100A9 peptide (CGSGRGHGHSHGKG) in 100 μL coating buffer per well (0.1 M sodium phosphate, 0.15 M sodium chloride, 10 mM EDTA, pH 7.2) overnight at 4 °C. After washing (3 × 5 min) with 200 μL/well PBS + 0.05% (v/v) Tween-20 (PBST), the plates were blocked for 1 h with 1% (w/v) l-cysteine (Sigma-Aldrich). After washing as above, plasma samples from vaccinated animals (two-fold serial dilutions in coating buffer) were added to the plates and incubated for 1 h at room temperature. After washing as above, we added a horseradish peroxidase (HRP)-labeled goat anti-mouse IgG (H+L) secondary antibody (Thermo Fisher Scientific, A16072) diluted 1:5000 in PBST (100 μL/well) and incubated as above. Finally, we added 1-Step Ultra TMB substrate (Thermo Fisher Scientific, 100 μL/well) and stopped the reaction after 5 mins with 100 μL/well 2 M H₂SO₄. The endpoint IgG titers were defined as the reciprocal plasma dilution at which the absorbance at 450 nm exceeded twice the background value (blank wells without a plasma sample).

IgG subclasses and immunoglobulin isotypes.

Pooled samples from weeks 2, 8 and 12 (soluble vaccine) or 5, 10 and 15 (implant vaccine) were diluted 1:1000 in coating buffer and tested as described for the endpoint IgG titers above, but the secondary antibodies were HRP-labeled goat anti-mouse IgG1 (Invitrogen PA174421, diluted 1:5000), IgG2b (Abcam ab97250, diluted 1:5000), IgA (Abcam ab98708, diluted 1:5000), IgE (Invitrogen PA184764, diluted 1:1000), and IgM (Abcam ab97230, diluted 1:5000). The IgG1/IgG2b ratio was used to define the response as Th2-biased (IgG1/IgG2b > 1) or Th1-biased (IgG1/IgG2b < 1).

ELISpot assay.

The ELISpot assay was carried out using a mouse IFN-γ/IL-4 double-color ELISPOT kit (Cellular Technology)^27,29,34. Briefly, 96-well ELISpot plates were coated with anti-mouse IFN-γ and anti-mouse IL-4 antibodies overnight at 4 °C. Splenocyte suspensions (1 × 10⁶ cells/well) collected from n = 3 mice 4 weeks post-immunization or n = 2 naïve mice were cultured with 100 μL medium alone (negative control), 10 μg free S100A9 peptide, 5 μg recombinant mouse S100A9 protein (R&D Systems), 5 μg unmodified Qβ, or 5 ng PMA/100 ng ionomycin (Sigma-Aldrich, positive control) at 37 °C in a 5% CO₂ atmosphere for 24 h. After washing with PBST, the plates were incubated with fluorescein isothiocyanate (FITC)-labeled anti-mouse IFN-γ (1:1000 dilution) and biotin-labeled anti-mouse IL-4 (1:666 dilution) antibodies for 2 h at room temperature. After washing as above, we added streptavidin alkaline phosphatase (AP, 1:1000 dilution) and anti-FITC-HRP (1:1000 dilution) to each well and incubated for 1 h at room temperature. After a final wash with PBST and a rinse in distilled water, we added the AP substrate and incubated for 15 min at room temperature, then rinsed with distilled water, added the HRP substrate, and incubated for 10 min at room temperature. The plates were rinsed five times with distilled water and air-dried at room temperature overnight. Colored spots were quantified using the Immunospot S6 Entry analyzer. Splenocytes were evaluated per animal and tested at least in duplicate for each stimulant. The results were reported as spot-forming cells (SFC) per 1 × 10⁶ cells.

Dot blot assay.

The C-terminal S100A9 epitope is unique among the S100 family⁶⁶ so we used a dot blot assay to confirm the specificity of the antibodies elicited by QβS100A9 particles. We spotted 2 μL (1 μg) of recombinant mouse S100A8, S100A9 or heterodimer S100A8/9 (R&D Systems) onto a nitrocellulose membrane (0.45-μm pore size, GE Healthcare) and blocked with 3% bovine serum albumin (Roche) at room temperature for 1 h, followed by washing with PBST. The membrane was then incubated at room temperature for 1 h with pooled plasma (week 4) from QβS100A9-vaccinated or free-peptide-vaccinated mice (diluted 1:100 in PBS). After another wash with PBST, we incubated the membrane with HRP-labeled goat anti-mouse IgG (diluted 1:5000 in PBST) at room temperature for 1 h. Finally, the membrane was washed as above and incubated with 3,3’-diaminobenzidine (DAB) substrate for 1 min. The development of a brown color indicated specific binding to the S100 proteins.

Liver and kidney biomarkers.

The safety of the QβS100A9 vaccine was determined by detecting plasma biomarkers related to liver and kidney injury. For liver damage, we determined the concentrations of the enzymes AST and ALT using the corresponding activity assay kits (Abcam). For kidney damage, we determined the concentration of KIM-1 using the Mouse KIM-1 ELISA Kit (Abcam). Plasma samples collected and tested at weeks 0 and 12 post-vaccination. The plasma samples were pooled from each group and tested in quadruplicate.

Mouse atherosclerosis model.

To determine the efficacy of our vaccine candidates, they were tested in the ApoE^−/− mouse model of atherosclerosis (Jackson Laboratory, #002052). Eight-week-old male ApoE^−/− mice were fed on the Envigo TD.88137a western purified atherogenic diet (20–23% milkfat/butterfat, 0.2% total cholesterol, 34% sucrose by weight). The animals were vaccinated as described above with PLGA-based QβS100A9 implants containing 300 μg VLPs, and matching implants containing the same dose of unmodified Qβ VLPs as controls. The mice were fed on a regular diet for 4 weeks post-immunization before switching to the high-fat diet for 20 weeks. Mice were fasted for 4 h and blood was sampled at weeks 0, 2, 4, 8, 12 and 24. Plasma was separated to determine endpoint IgG titers and immunoglobulin isotypes as described above. At the end of the study (week 24), the mice were euthanized by CO₂ asphyxiation followed by exsanguination by cardiac puncture. Aortas were perfused with PBS (pH 7.4) before the thoracic aorta and aortic arch were dissected, fixed with 4% formalin, denuded of connective tissue and processed for oil red O staining. Hearts, livers, kidneys, lungs, spleens, and abdominal fat tissues were weighed and fixed with 4% formalin and embedded in paraffin for further analysis.

Oil red O staining.

Aortas were stained as previously reported⁶⁷. Briefly, cleaned and fixed aortas were placed individually in 1.5-mL tubes and equilibrated in 1 mL 78% methanol by gentle motion on a tilted roller (2 × 5 min). The methanol was then replaced with 1 mL fresh 0.2% oil red O solution and the tissue was incubated at room temperature for 1 h. After staining, the tissue was transferred to a clean tube and washed with 1 mL 78% methanol on the tilted roller (2 × 5 min). Finally, the methanol was replaced with 1 mL PBS and aortas were stored at 4 °C. To visualize and quantify the atherosclerotic lesions, aortas were dissected longitudinally and pinned with the lumen siding facing up on a dissecting dish under a stereomicroscope (VistaVision, 10X). The percentage of atherosclerotic lesions was determined by densitometry using ImageJ v1.44o (http://imagej.nih.gov/ij) and was calculated as follows:

$$\%\text{ lesion}=\left(\text{total lesion areas }\left[\text{stained red area}\right]/\text{total aorta area}\right)*100$$

Cytokine quantification.

Commercial sandwich ELISA kits were used to measure the concentrations of different cytokines in plasma from QβS100A9-vaccinated and control mice. We used kits for mouse calprotectin S100A8/9 heterodimer (R&D Systems, DY8596-05), mouse MCP-1 (Abcam, ab208979), mouse IL-1β (Abcam, ab229440) and mouse IL-6 (R&D Systems, M6000B).

Statistical analysis.

Data are presented as means ± standard errors of the mean (SEM), with the number of replicates presented for individual experiments. Differences between groups were analyzed using an unpaired two-tailed t-test. Data were analyzed in GraphPad Prism v6, with p < 0.05 defined as the threshold for statistical significance.