Serotype 1 S. pneumoniae (WHO reference laboratory strain SSISP, Statens Institute, Denmark), serotype 4 (TIGR4, American Type Culture Collection CA, USA) or serotype 14 (NCTC11902, Public Health England, UK) were grown as previously described (Dockrell et al., 2003). Prior to their use in vivo, all bacterial strains were passaged through C57BL/6J mice (Harlan, UK) by intraperitoneal (i.p.) injection and recovery of bacteria from blood taken at 16 hours post infection and incubated on Columbia blood agar (Canvin et al., 1995).

All mice were used in compliance with the UK Animals (Scientific Procedures) Act 1986 and Directive 2010/63/EU and all experiments were approved by the animal project review committee of the University of Sheffield and conducted under a Home Office approved animal project license. C57BL/6J mice and ApoE^-/- mice (on a C57BL/6J background) were bred in house. All mice were housed in a controlled environment at 22°C with a 12 h light/dark cycle. Prior to and after periods of high fat diet feeding, mice were fed standard chow diet. Mice which were fed a high fat diet were given either 5g/mouse/day of Western diet (20% casein, 16% fat (15% cacao butter, 1% corn oil), 0.25% cholesterol; Abdiets, The Netherlands) or 5g/mouse/day of Paigen diet (18.5% fat, 0.9% cholesterol and 0.5% cholate; Special Diet Services, UK) for 8 weeks from 11-12 wks of age, and then put back on to chow diet. All experiments were conducted at the same time of day to minimize the impact of circadian rhythm and mice were co-housed to minimize microbiome associated variation.

Intratracheal instillation of pneumococci, or mock infection with phosphate-buffered saline (PBS), was performed as previously described (Dockrell et al., 2003). Prior to intratracheal instillation, each mouse was anaesthetised using ketamine (100 mg/kg i.p.; Willows Francis, UK) and acepromazine (5mg/kg i.p; C-Vet Veterinary Products, UK). Once the pedal reflex was lost, the neck of each anaesthetised mouse was shaved and a midline incision of approximately 1cm was made on the ventral aspect. The trachea was exposed by dissection of the adjacent tissues with blunt forceps. A 24-gauge catheter (Smiths Medical, OH, USA) was then inserted into the lumen of the trachea, and 20 µl solutions containing either pneumococci or PBS alone were instilled into the lungs using a pipette. Following instillation, the catheter was removed from the trachea and the surrounding tissue pinched together to close the incision. The mice were then left on their backs in an incubator at 33°C and observed closely for any signs of distress until they had fully regained consciousness. Mice undergoing intranasal instillation of pneumococci were anaesthetised using inhaled 4% isoflurane (Zoetis, UK) delivered in 2 l/min O₂ inside a 1 litre induction chamber until an areflexic state was reached. Instillation was then performed by placing the animal in a supine position and pipetting 50 µl of a solution of pneumococci (or mock infection with 50 µl PBS) onto the nostrils, allowing the mouse to inhale each droplet before introducing the next one. Following either intranasal or intratracheal instillation of bacteria or PBS, mice were continuously observed until full consciousness was regained. They were then inspected every 12 hours during the first 48 hours following instillation and daily thereafter. Any mouse which displayed signs of severe illness following instillation of bacteria was culled by humane methods. Severe illness was defined as any of the following features: severely reduced respiratory rate (0.5s between breaths), severely reduced mobility (mouse only moved when provoked and then only a few steps) or hunched posture. Blood, bronchoalveolar lavage (BAL) fluid and lungs were collected and total and differential cell counts in BAL fluid and viable bacterial count in blood and lung homogenates were measured as previously described (Miles et al., 1938; Dockrell et al., 2003). During experiments assessing atherosclerosis outcomes, only mice surviving to pre specified time points were included in the analyses.

A veterinary preparation of amoxicillin was used (Amoxypen 150 mg/ml suspension for injection, Intervet, UK). Both infected and mock infected mice received amoxicillin. The mice were given 3 doses of 100mg/kg amoxicillin s.c. at 12 hourly intervals. The first amoxicillin dose was administered 24 hours post infection or mock infection, immediately after the tail vein snip procedure, if performed.

Mice receiving atorvastatin were administered powdered Atorvastatin Calcium (Pfizer) in 0.5% methylcellulose by daily gavage.

Mice were killed by an overdose of i.p. sodium pentobarbital. Blood was collected by cardiac puncture and mice were then perfusion fixed with 10% v/v of formalin buffered saline. The aortic sinus and brachiocephalic artery were dissected, embedded in paraffin and serially sectioned (5 µm). Aortic sinus sections were stained with Miller’s Elastin/Modified Van Gieson (MVG), to enable analysis of atheromatous lesions including lesion area, or Picrosirius Red to measure plaque collagen content. Aortic sinus sections were immunostained for macrophage marker MAC-3, α-smooth muscle actin or cellular proliferation marker Ki67. A detailed description of the immunostaining methods used can be found in the supplementary file ( Supplementary Table 1 ). Images of stained histological sections were analysed using NIS-Elements Image Analysis Software (Nikon Instruments, UK). MVG stained aortic sinus sections were used to measure atherosclerotic lesion area which was expressed as a percentage of vessel cross-sectional area (CSA). Four consecutive sections from each aortic sinus were analysed to generate a mean atherosclerotic lesion area/CSA measurement for each mouse. MVG stained sections were also used to measure necrotic core area, as defined by acellular areas within atherosclerotic lesions, as a proportion of total lesion area. Collagen content of atherosclerotic lesions was expressed as area of positive Picrosirius Red staining as a proportion of total lesion area. Similarly, area of positive immunostaining for MAC-3 and smooth muscle actin were expressed as a proportion of total lesion area, while Ki67 immunostaining was expressed as the proportion of atherosclerotic lesion cells staining positive.

Aortic sinus sections were stained with anti-serotype 4 pneumococcal antiserum using immunofluorescence to detect bacterial invasion into tissue. Following incubation with the pneumococcal antiserum, sections were stained with Alexa Fluor 568 anti-rabbit IgG secondary antibody and analysed using a wide field fluorescence microscope. A detailed immunofluorescence staining protocol can be found in the Supplementary file.

Mice were killed by an overdose of i.p. sodium pentobarbital. The aortic sinus was dissected, embedded in optimal cutting temperature compound (OCT) (Thermo Fisher Scientific, UK) and frozen. Serial sections (7µm) through the aortic sinus were prepared using a cryostat (Leica CM3050 S, Leica, Germany). Every 5^th frozen aortic sinus slide was immunostained for the macrophage marker MAC-3 as a ‘guide slide’ to direct laser capture microdissection (LCM) of macrophages on the remaining cryo-sections (Feig and Fisher, 2013). A detailed protocol for MAC-3 immunostaining of cryo-sections can be found in the supplementary file. All aortic sinus cryo-sections which were not immunostained for MAC-3 underwent LCM and were stained with 0.1% w/v Toluidine blue. Sections were fixed in 70% ethanol for 30s, rinsed in DEPC water before staining for 30s in Toluidine blue and rinsing again in DEPC water. Tissue sections were then dehydrated through increasing concentrations of ethanol for 30 seconds each step (70%, 95%, 2 changes of 100%) and xylene for 5 minutes. Slides were then left to air dry before proceeding immediately to LCM. LCM on aortic sinus cryo-sections was performed using the Pixcell II laser-capture microdissection system (Arcturus Engineering, CA, USA) and cells were collected using Capsure Macro caps (Arcturus Engineering). The Pixcell system was set to the following parameters: 40 mW power and 7.5 µm laser spot size. Approximately 2000 MAC-3 positive cells were collected from each aortic sinus. The PicoPure RNA isolation kit (Thermo Fisher Scientific, UK) was used to recover total RNA from cells isolated during LCM, according to the manufacturer’s instructions.

The Affymetrix GeneChip Mouse Gene 2.0 ST Array (Affymetrix, UK) was used to measure whole-transcriptome gene expression in a single hybridisation. The WT Pico Reagent Kit (Affymetrix) was used according to the manufacturer’s instructions to process RNA samples (500pg of RNA from each aortic sinus) obtained following LCM and convert them to double stranded cDNA to act as a template for in vitro transcription (IVT). During the IVT step, complimentary RNA (cRNA) was synthesised and also amplified using T7 RNA polymerase. 20 µg of cRNA from each sample was then used to synthesise fragmented and labelled cDNA ready for array hybridisation. The Affymetrix Gene Chip Mouse Gene 2.0 ST Array cartridge was used according to manufacturer’s instructions. Following hybridisation, each microarray cartridge was washed and stained in the GeneChip Fluidics Station 400 (Affymetrix) according to manufacturer’s instructions. The arrays were stained with light sensitive streptavidin phycoerythrin (SAPE), followed by the addition of a biotinylated anti-streptavidin antibody. Finally, the arrays were scanned in the GeneChip 3000 scanner (Affymetrix, UK).

Microarray data were normalised using the Robust Multichip Average method (Bolstad et al., 2003) and then analysed using the Linear Models for Microarray (LIMMA) package in R. Differentially expressed genes between infected and mock infected groups were identified through unpaired t-tests, using the parameters p<0.05 and gene expression log2 fold change >1. The analysis was conducted twice, firstly with False Discovery Rate (FDR <0.05) adjusted p value and the second time with unadjusted p value. The list of differentially expressed genes was then used for pathway analysis to identify the biological pathways that had been significantly perturbed by pneumococcal infection. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed using the clusterProfiler in R (Yu et al., 2012), with statistical significance set at p<0.05.

Quantitative real-time PCR (qPCR) was used to validate microarray differential gene expression results. Firstly, the qScript cDNA Supermix (Quanta Bio, MA, USA) was used according to the manufacturer’s instructions to synthesise single stranded cDNA from each RNA sample obtained following LCM of plaque macrophages. The PCR reaction was run on a 7900T fast real time PCR system (Applied Biosystems, UK) using the GoTaq Probe qPCR Master Mix (Promega) in accordance with the manufacturer’s instructions. Data were normalised to cyclophilin A or 18S. Fold change was calculated using Delta Delta Ct values. The PCR primers used are listed in Supplementary Table 2 .

Statistical analysis was performed using R for the microarray or for all other analyses in Graphpad Prism version 7.0. Data were expressed as mean +/- standard error of the mean (SEM) unless otherwise stated. Comparison between groups employed unpaired t-test unless otherwise stated in figure legends. Statistical significance was defined as p<0.05.

Serotype 1 S. pneumoniae was initially used due to its high invasive potential (Sandgren et al., 2004; Sandgren et al., 2005) but preliminary studies showed only 3 out of 8 Paigen diet fed mice (38%) survived to 14 days post infection making this model unsuitable to investigate long term effects of pneumococcal disease. Two mice died within 6 h of infection, 2 were culled 24-72 h post infection and 1 was culled at 5 days due to signs of severe illness ( Figure 1A ). All but 1 of the 6 infected mice which had a blood sample taken at 24 h post infection by tail vein snip were bacteremic, with mean blood viable bacterial count of 5.3 x 10⁵ cfu/ml (range 3.3 x 10² – 1.6 x 10⁶).

There was no significant difference in aortic sinus lesion area between the 2 groups (22.78% ± 4.82 mock infected vs. 23.99% ± 1.15 infected; n=4-5, ns) ( Figure 1B ) in mice which had survived for greater than 72h after infection/mock infection.

We then adjusted the diet to a Western diet to develop a model with less extensive (sub-maximal) atherosclerotic plaque formation, thus enabling more potential for enhancement by infection, and also with the aim of reducing early infection-related mortality that was potentially attributable to the Paigen diet. In addition to atherosclerosis, high-fat diet fed ApoE^-/- mice have been shown to be a model of non-alcoholic steatohepatitis (NASH) with hepatic fibrosis, a condition which is associated with an increased risk of postoperative death following major surgery in humans (Zarzavadjian Le Bian et al., 2012; Schierwagen et al., 2015). Ketamine and acepromazine are metabolized primarily by the liver, and therefore liver dysfunction may have potentiated their effects in this model, possibly contributing to the early post-procedure mortality (Soleimanpour et al., 2015). One of the reasons therefore to switch to the Western diet, which does not contain any cholate and which has a lower cholesterol content than the Paigen diet, was to mitigate the potentially increased susceptibility to anaesthesia. The model was further modified by the replacement of intratracheal with intranasal instillation, thus modifying the general anaesthetic requirements which may also have contributed to the early mortality demonstrated following intratracheal instillation. We next screened pneumococcal strains with relatively high invasive potential to identify one that resulted in high rates of bacteremia following intranasal instillation. Of the serotype 4 infected mice, 3 out of 4 were bacteremic whereas there were no bacteremic mice amongst those infected with serotypes 1 and 14 ( Figure 2A ). Three out of 4 mice infected with serotype 4 S. pneumoniae had detectable pneumococci in the lungs compared to 1 out of 4 of those infected with serotypes 1 and 14 ( Figure 2B ).

Having demonstrated relatively high rates of bacteremia following intranasal instillation of serotype 4 S. pneumoniae, we then sought to optimize the infecting dose in Western diet fed ApoE^-/- mice to maximize the percentage of mice with bacteremia. The percentage of mice with bacteremia following the 10⁵ cfu dose was 100%, compared with 40% (5 x 10⁴ cfu) and 20% (10⁴ cfu) for the other doses (n=3-5) ( Supplementary Figure 1 ).

The survival rate of Western diet fed ApoE^-/- mice infected with the 10⁵ cfu dose was then assessed. Within 72 h of infection, all mice (n=8) either died or had to be culled (2 died, 6 culled) due to developing signs of severe illness. Two of these mice were not bacteremic at 24 hours post infection. Survival assessments in smaller cohorts (n=3) of Western diet-fed ApoE^-/- mice infected with 5 x10⁴ cfu and 10⁴ cfu also demonstrated high mortality rates within 72 hours of infection (3 and 2 were culled respectively).

As doses lower than 10⁵ cfu resulted in relatively low rates of bacteremia, the model was further modified to combine invasive disease with subsequent antimicrobial treatment to enhance survival despite establishment of pneumonia and early bacteremia, thus replicating the clinical situation. The use of antibiotics was also predicted to allow for the instillation of a higher dose of pneumococci, and 5 x10⁵ cfu was chosen with the aim of generating greater blood bacterial counts than those seen with the 10⁵ cfu dose.

Eight out of 9 mice (88.9%) were bacteremic and eight out of 9 (88.9%) had detectable pneumococci in the lungs at 24 h post infection ( Figures 3A, B ). The single mouse in this cohort which was not bacteremic did have detectable bacteria in the lungs and 9% BAL fluid differential neutrophil count, suggestive of lung infection with associated inflammatory response. Compared to mock infected mice, the infected mice demonstrated a significant increase in total and differential (% neutrophil) BAL fluid cell counts, consistent with the development of pneumonia ( Figures 3C, D ).

The introduction of antimicrobial therapy was the final modification to the model. Infected mice were given 3 x 100mg/kg doses of amoxicillin subcutaneously (s.c.) at 12 hourly intervals with the first dose given at 24 h post infection. Twenty-six mice were infected using this model, three were culled at 36, 48 and 72 h post-infection respectively due to signs of severe illness, with the rest surviving throughout the 2 wk study. The model therefore resulted in a 2 wk post-infection survival rate of 88.5%.

Atherosclerotic lesion development was assessed in the aortic sinus of Western diet-fed ApoE^-/- mice at the end of the 8 week high fat-diet feeding period ( Supplementary Figure 2 ). All 5 mice developed at least one large advanced atherosclerotic plaque at the aortic root with extracellular lipid and areas of fibrous cap formation.

The final optimized infection plus atherosclerosis model can be summarized as follows:

1. Male ApoE^-/- mice aged 11-12 weeks fed a Western diet for 8 weeks prior to infection.

2. Intranasal instillation of 5x10⁵ cfu serotype 4 (TIGR4) S. pneumoniae.

3. Three doses of s.c. 100 mg/kg amoxicillin 12 hourly commenced 24 h post infection.

4. Mice maintained on chow diet for the required observation period post infection.

Using this optimized model, we then investigated the effect of pneumococcal pneumonia on aortic sinus atherosclerotic plaque burden and markers of plaque vulnerability to rupture at 2 weeks and 8 weeks post infection. Pneumococcal pneumonia had no effect on aortic sinus plaque area at either time point, but did lead to increased plaque macrophage content 2 weeks post infection ( Figures 4A, B , 5 , 6 ). No difference in plaque macrophage content was observed at the 8 week time point, and no difference in aortic sinus plaque smooth muscle, collagen content or necrotic core size were found at either time point ( Figures 4C–E , 5 , 7 ). Plaque cell proliferation, quantified by the proportion of Ki67-positive plaque cells, was also unaffected by pneumococcal pneumonia at 2 weeks post infection ( Figure 4F ). Brachiocephalic arteries did not develop significant atherosclerotic plaques in our model, nor was there any evidence of inflammatory infiltration in the arterial wall at this site following infection ( Supplementary Figure 3 ). Immunofluorescence assays using anti-serotype 4 pneumococcal antiserum showed no positive staining in atherosclerotic plaques or elsewhere in the aortic sinus 2 weeks post infection using our optimized model, and therefore we found no evidence of pneumococcal invasion into atherosclerotic plaques at this time point ( Supplementary Figure 4 ).

Having demonstrated increased plaque macrophage content 2 weeks post pneumococcal infection, we studied the transcriptional profile of these macrophages. First, to confirm that LCM leads to enrichment of macrophage specific transcripts, RNA was extracted from LCM-derived MAC-3 immuno-positive plaque cells and from whole aortic sinus sections of ApoE^-/- mice fed a Western diet for 8 weeks. Using qPCR to compare gene expression, LCM obtained samples demonstrated enrichment of the macrophage specific transcript CD68 and conversely expression of the smooth muscle cell marker ACTA2 was reduced in samples obtained by LCM ( Supplementary Figure 5 ).

Next, plaque macrophages were collected at 2 weeks post infection (n=9) or mock infection (n=11) using LCM and RNA extracted from these cells to enable comparison of gene expression profiles between infected and mock infected mice. Analysis of microarray data using FDR adjusted p value identified only 1 differentially expressed probe which was non coding and therefore not assigned to a particular gene. The analysis was repeated using FDR unadjusted p value and identified 36 differentially expressed coding probes between infected and mock infected groups ( Figure 8 ). These genes are listed in Supplementary Table 3 .

In order to characterise the biological relevance of the 36 genes, pathway analysis was carried out using clusterProfiler to identify significantly perturbed KEGG pathways. Three pathways were significantly affected by pneumococcal infection: ubiquitin mediated proteolysis, endocytosis and HTLV-1 infection ( Supplementary Table 4 ). All genes associated with these pathways were downregulated in plaque macrophages from the infected group. Both ubiquitin mediated proteolysis and endocytosis pathways were identified as significantly perturbed due to downregulation of 3 genes which coded for E3 ubiquitin ligases: anaphase promoting complex subunit 1 (Anapc1), HECT, UBA and WWE domain containing 1 (Huwe1) and Itch.

To validate the microarray findings demonstrating downregulation of these E3 ubiquitin ligases, we performed qPCR analysis of LCM derived plaque macrophages from independent biological replicates 2 weeks post infection or mock infection. Down-regulation of Huwe1 and Itch was confirmed in infected mice using qPCR, but downregulation of Anapc1 was not demonstrated ( Figure 9 ).

We next investigated whether the plaque stabilising capability of statins could be used to mitigate the previously demonstrated increase in plaque macrophage content following pneumonia. In the optimised model, mice were administered powdered atorvastatin calcium 40mg/kg in 0.5% methylcellulose or control (methylcellulose alone) once daily by gavage for 10 days, with the first statin dose administered 24 hours post infection. We observed no difference in aortic sinus plaque burden, plaque macrophage content or other markers of plaque vulnerability at 2 weeks post infection following statin therapy ( Figure 10 ).