Epithelial control of colonisation by Streptococcus pneumoniae at the human mucosal surface

Control of Streptococcus pneumoniae colonisation at human mucosal surfaces is critical to reducing the burden of pneumonia and invasive disease, interrupting onward transmission, and in achieving herd protection. We hypothesised that the pattern of pneumococcal-epithelial engagement dictates the inflammatory response to colonisation, and that this epithelial sensing is linked to bacterial clearance. Here we have used nasal curette biopsies from a serotype 6B Experimental Human Pneumococcal Carriage Model (EHPC) to visualize S. pneumoniae colonisation and relate these interactions to epithelial surface marker expression and transcriptomic profile upregulation. We have used a Detroit 562 cell co-culture model to further understand these processes and develop an integrated epithelial transcriptomic module to interrogate gene expression in the EHPC model. We have shown for the first time that pneumococcal colonisation in humans is characterised by microcolony formation at the epithelial surface, microinvasion, cell junction protein association, epithelial sensing, and both epithelial endocytosis and paracellular transmigration. Comparisons with other clinical strains in vitro has revealed that the degree of pneumococcal epithelial surface adherence and microinvasion determines the host cell surface marker expression (ICAM-1 and CD107), cytokine production (IL-6, IL-8 and ICAM-1) and the transcriptomic response. In the context of retained barrier function, epithelial microinvasion is associated with the upregulation of a wide range of epithelial innate signalling and regulatory pathways, inflammatory mediators, adhesion molecules, cellular metabolism and stress response genes. The prominence of epithelial TLR4R signalling pathways implicates pneumolysin, a key virulence factor, but although pneumolysin gene deletion partially ameliorates the inflammatory transcriptional response in vitro, critical inflammatory pathways persist in association with enhanced epithelial adhesion and microinvasion. Importantly, the pattern of the host-bacterial interaction seen with the 6B strain in vitro is also reflected in the EHPC model, with evidence of microinvasion and a relatively silent epithelial transcriptomic profile that becomes most prominent around the time of bacterial clearance. Together these data suggest that epithelial sensing of the pneumococcus during colonisation in humans is enhanced by microinvasion, resulting in innate epithelial responses that are associated with bacterial clearance. Highlights Colonisation of the human mucosa by Streptococcus pneumoniae is associated with microcolony formation, microinvasion, epithelial sensing and an epithelial innate response. Following adherence to the epithelial cell surface, microinvasion of the epithelium may occur by endocytosis and/or lateral migration between cells without necessarily compromising barrier integrity. The pattern of pneumococcal epithelial surface adherence and microinvasion determines the host cell response through a range of innate signaling and regulatory pathways, inflammatory mediators, adhesion molecules, cellular metabolism and stress response genes. Epithelial sensing is triggered by, but not wholly dependent on pneumolysin, a key virulence factor of S. pneumoniae.

mechanism. Here, we show that pneumococcal colonisation in humans is characterised by 136 microcolony formation and junctional protein association, epithelial sensing that is indeed enhanced 137 by microinvasion. This occurs both by epithelial endocytosis and paracellular migration resulting in 138 epithelial innate responses that are not entirely pneumolysin dependent and that is associated with 139 bacterial clearance. These data implicate epithelial microinvasion in the initiation of bacterial 140 clearance which to the benefit of the colonising pathogen may also enhance transmission.   The entire cytospin for each sample was manually viewed by microscopy for detection of 217 pneumococci. Multiple fields of view were imaged for each transwell insert, for each 218 condition. Images were captured using either an inverted LSM 700, LSM 880, or 9 TissueFAXS Zeiss Confocal Microscope. Z stacks were recorded at 1µm intervals at either 220 40x oil or 63x oil objectives. were also run for each experiment. Samples were acquired until 300,000 events had been 245 collected. Analyses was performed using FlowJo version 10 software.   and diagnostic plots for quality control. Using these techniques, cells exposed to different S.  (hypergeometric test) to identify enriched pathways from the REACTOME database.

288
Enriched pathways were then represented in a heat map using log2 z-scores. REACTOME 289 database has a non-structured list of terms, therefore terms were clustered based on 290 overlapping genes. All heat maps were produced with a heat map R package using 291 Euclidean distances and hierarchical clustering. The same gene lists were used to test for 292 enrichment in Gene Ontology cellular components and membrane-related terms were 293 selected. Upstream regulator analysis was performed in Ingenuity Pathway Analysis (IPA).

12
In vivo data were processed with the same pipeline used for the in vitro experiments. Mapped reads 297 ranged between 16M to 66M. Upregulated gene lists were produced and only genes with a log2 298 FC>1 were used for further pathway analysis. Pathway analysis with REACTOME database was 299 performed with InnateDB. TPM for all genes were obtained and transformed into log2 scale. Quality 300 control for 75 samples showed a batch effect due to two different labs sequencing the data. Combat 301 function in the SVA R package was used 44 to reduce this effect. Principal component analysis 302 identified an outlier that was removed for further analysis. Using the gene interactome lists for each 303 strain from the in vitro data, a pan signature or module was obtained which included 200 genes that 304 were upregulated in at least one strain. 16 genes were shared among all strains. Module scores for 305 each group were derived by calculating the log2 average gene expression for each module. A non-306 parametric (Mann-Whitney) test was performed to compare carriers to non-carriers for each time 307 point. Violin plots were produced with in house script in R and ggplot2 43 .

317
Streptococcus pneumoniae colonisation of the human nasal mucosal is associated 318 with adhesion, microcolony formation and microinvasion 319 We have used an Experimental Human Pneumococcal Carriage Model 45 to characterise 320 pneumococcal-epithelial interactions in vivo. Colonisation was detected in 9/13 healthy 321 volunteers by culture, 11/13 by microscopy and 9/11 by LytA PCR ( Table 1). The carriage 322 status of each volunteer in the study was blinded until sample collection was completed.

13
Differences in the results obtained with each detection method may reflect methodological 324 threshold detection, or the location and therefore the accessibility of the colonising 325 pneumococci (e.g. in the mucus escalator vs. adherence to the epithelial cell surface). 326 Nonetheless, all three methods demonstrated that colonisation was established and that 327 clearance largely occurred between day 9 and 27 (Table 1 and Figure 1B).   Strikingly, the number of TIGR4 pneumococci associated with the Detroit 562 cells was ten-389 fold higher than 6B or 23F strains ( Figure 3A). This pattern was also observed by  Figure 5A), which indicates vesicular endocytosis and using bacteria that are pre-stained 394 with FAMSE, we were able to distinguish extracellular bacteria (blue) from those below the 395 apical surface prior to permeabilization of the cells (green, Figure 3F). Interestingly, co-  To assess transmigration, pneumococci that had penetrated the basal chamber of cells 400 cultured on transwell inserts were counted. Although only statistically significant at 1hr 401 between 23F and TIGR4, 23F was more readily detected compared to the other strains 402 ( Figure 3C). By microscopy, we observed laterally located bacteria, and pneumococci zip-403 wiring between cell junctions ( Figure 3H). We observed pneumococci at the level of the 404 nuclei and below the basal membrane ( Figure 3I). This was more readily, but not exclusively,    Trans-epithelial electrical resistance (TEER) is not high in Detroit 562 cells but nevertheless 423 TEER was not affected by pneumococcal co-culture ( Figure 3J). To assess permeability, 4kDa FITC-dextran was applied to the apical chamber of transwells and epithelial leak 425 quantified from the basal chamber. With the Detroit 562 cells, a significant reduction in 426 permeability was seen with 6B pneumococci, 23F and TIGR4 (23 -34%), compared to non-427 infected cells ( Figure 3K). This implicates a role for pneumolysin in epithelial integrity.. These 428 data suggest that loss of epithelial cell barrier function is not a pre-requisite for 429 pneumococcal adhesion and microinvasion, and that as described in murine models 23,25 , 430 changes to barrier function appear pneumolysin dependent.

432
To explore the possibility that our findings were not cell line dependent, we also used A549    To further explore the hypothesis that the pattern of epithelial adhesion and microinvasion 499 results in differential epithelial sensing and therefore epithelial inflammatory-response 500 genes, we performed RNAseq and obtained transcriptomic data from our pneumococci 501 infected Detroit 562 cells. As shown in Figure 6A, we found that TIGR4 upregulated 1127 502 genes (550 unique genes), 23F upregulated 650 genes (69 unique genes), and 6B 503 upregulated only 153 genes (10 unique genes) compared to non-infected cells. The 504 pneumolysin mutant upregulated 220 genes (14 unique genes). 93 genes were upregulated 505 by all strains compared to non-infected cells. These findings appeared to reflect the invasive 506 and inflammatory nature of these bacteria in this in vitro epithelial model. To further explore 507 the nature of these differences, we performed pathway analyses using the REACTOME 508 database and performed with XGR( Figure 6B). Again, we found that the upregulated 509 pathways for TIGR4 and 23F were pro-inflammatory, but that the 6B profile was relatively Further bioinformatics analysis of upstream regulators revealed that RELA, or the nuclear 527 factor NFκB p65 subunit is likely to be a key mediator of these pneumococcal-epithelial 528 interactions ( Figure 6E, Supplementary Figure 5C). Comparisons between the strains again 529 reveal a more silent upstream profile with 6B compared to TIGR4 or 23F.               6B  23F  TIGR4  dPLY  ECSIT  TCF4  WT1  ETS1  PSMD10  NFKBIA  EIF4E  NCOA3  ATF7  IRF5  IFI16  ZBTB17  FOXO3  FLI1  NFKBIB  JUNB  CCND1  RUNX2  NCOA1  IRF7  ATF4  NOTCH1  SMARCA4  SMARCB1  STAT3  SREBF1  FOXL2  ETV4  FOXO1  MED1  NUPR1  TARDBP  RELA  ELF3  HIF1A  SREBF2  NFATC1  SMAD2  TGIF1  XBP1  TP53  Gm21596/Hmgb1  STAT1  MYC  REL  ETV1  POU5F1  ARNT2  TP63  TEAD1  EGR1  MYB  CTNNB1  NFATC2  NFKB1  BCL10  JUN  STAT4  HMGB1  PPRC1  EZH2