Abstract
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.
Introduction
Colonisation of upper respiratory tract (URT) mucosa by a range of bacteria is a necessary precursor to transmission and disease. Streptococcus pneumoniae is a common coloniser of the human nasopharynx and is estimated to be responsible for >500,000 deaths due to pneumonia, meningitis and bacteraemia in children aged 1–59 months worldwide1. In comparison to gut pathogen-mucosal interactions2, the control of pneumococcal colonisation is far less well understood, particularly in humans3.
In Europe and North America, there has been a dramatic impact of pneumococcal conjugate vaccine (PCV) on vaccine serotype (VT) invasive disease and carriage4,5. Indeed, more than 50% of PCV impact has been due to a reduction in VT colonisation resulting in reduced transmission and therefore disease. This is the basis of herd protection4,5. However, the emergence of non-VT pneumococcal disease across the world and the more modest impact of PCV on colonisation in high transmission settings threaten this success6-10. As a first step towards further optimising vaccine impact on pneumococcal colonisation, it is critically important to define the mechanistic basis of the control of S. pneumoniae at the mucosal surface.
We and others have previously demonstrated that antigen-specific URT mucosal T cell immune memory to subcapsular pneumococcal protein antigens in humans is acquired with age in humans, is predominately pro-inflammatory and is heavily regulated by Treg11-13. Antibodies to subcapsular protein antigens rather than to the polysaccharide capsule (the target of currently licenced vaccines), also appear important for the natural control of colonisation and clearance14. We and others15,16, suggest that the URT epithelium is at the centre of this process, orchestrating both innate/inflammatory and adaptive immune mechanisms17-19, promoting bacterial clearance. The epithelium senses bacteria colonising the mucosal surface, rapidly transducing inflammatory signals and recruiting immune cells. However, murine models20 and epidemiological studies of viral-coinfection21,22 suggest that the resulting inflammation also leads to onward transmission to susceptible individuals23. This inflammation-driven transmission is crucial for the continued success of the pneumococcus.
Nasal colonisation by S. pneumoniae in murine models is proinflammatory24 and is associated with epithelial paracellular transmigration and tight junction modulation25. Mediated though pneumococcal protein C (PspC)–polymeric immunoglobulin receptor (pIgR) interactions26,27, S. pneumoniae invasion of immortalised epithelial cell monolayers has also been shown to occur by endocytosis27. The relative importance of epithelial endocytosis and paracellular migration in microinvasion remains uncertain27but may influence epithelial sensing of this otherwise extracellular pathogen through multiple pathogen-associated molecular patterns (PAMPS). These may include TLR2 signalling via lipoteichoic acid28, Nod1 signalling via peptidoglycan29 and TLR4 signalling via pneumolysin23,30, a pore-forming toxin that mediates transmission in an infant mouse model23,30. Indeed, microinvasion of the epithelium may overcome the sequestering of pattern recognition receptors either at the basolateral surface or intracellularly31-33.
Much of what we understand of the control of pneumococcal colonisation is derived from epidemiological studies and murine carriage models. Experimental human pneumococcal challenge (EHPC) model provides a well-controlled, reproducible tool to characterise the cellular and molecular mechanisms that underlie pneumococcal colonisation in humans34. We have therefore explored the hypothesis that epithelial microinvasion S. pneumoniae by enhances the innate immune responses associated with colonisation and have characterised the underlying cellular and molecular mechanism. Here, we show that pneumococcal colonisation in humans is characterised by microcolony formation and junctional protein association, epithelial sensing that is indeed enhanced by microinvasion. This occurs both by epithelial endocytosis and paracellular migration resulting in epithelial innate responses that are not entirely pneumolysin dependent and that is associated with bacterial clearance. These data implicate epithelial microinvasion in the initiation of bacterial clearance which to the benefit of the colonising pathogen may also enhance transmission.
Methods
Bacteria
S. pneumoniae clinical strains used were 6B (BHN 41835), 23F (P112136) and TIGR4 (P167237), together with a pneumolysin deficient TIGR4 mutant strain (kind gift from Prof T Mitchell, University of Birmingham). Stocks of bacterial aliquots grown to O.D 0.3 were stored at −80°C, defrosted, resuspended in cell culture media and used once. Colony forming units were counted on horse blood agar plates (EO Labs).
Experimental Human Pneumococcal Carriage Model (EHPC)
Following written informed consent, healthy non-smoking adults between the ages of 18 – 59 were inoculated with 80,000 CFU live 6B S. pneumoniae (BHN418), grown to mid-log phase in vegetone broth as previously described38. All volunteers were negative for the pneumococcus at baseline. Nasal washes and mucosal cells (curette biopsy) from the inferior turbinate were obtained by PBS syringe and curettage using a plastic Rhino-probeTM (Arlington Scientific, Springville, UT), respectively before pneumococcal inoculation. These were then repeated on days 2, 6, 9, 14 - 27 post inoculation39. Bacteria collected from nasal washes were quantified by CFU counts. Two curettage samples were obtained and processed for confocal immunofluorescence, flow cytometry, primary cell culture and /or transcriptomic analysis by RNAseq.
Ethical approval was given by NHS Research and Ethics Committee (REC)/Liverpool School of Tropical Medicine (LSTM) REC, reference numbers: 15/NW/0146 and 14/NW/1460 and Human Tissue Authority licensing number 12548.
Human Respiratory Tract Epithelial Cells
Human pharyngeal carcinoma Detroit 562 epithelial cells (ATCC_CCL-138) and human bronchial carcinoma Calu3 epithelial cells (ATCC_HTB-55) were grown in 10% FCS in alpha MEM media (Gibco). Human alveolar epithelial carcinoma A549 epithelial cells (ATCC_CCL-185) were grown in 10% FCS with 1% L-glutamine in Hams/F-12 media (Gibco).
Pneumococcal-epithelial cell co-culture
Association and Invasion assays-confluent Detroit 562 (day 8 post plating), Calu3 (day 10 post plating) and A549 (day 4 post plating) monolayers were cultured on 12 well plates (Corning) were exposed to S. pneumoniae for three hours in 1% FCS alpha MEM. The medium was removed and cells washed three times in HBSS+/+. Cells were incubated in 1% saponin for 10 minutes at 37°C and lysed by repetitive pipetting. Dilutions of bacteria were plated on blood agar and colonies counted after 16 hours. To quantify internalised bacteria, 100μ g/ml gentamicin was added for 1 hour to the cells, which were then washed another three times, before incubating with Saponin and plating on blood agar plates. Colony forming units (CFU) were counted after 16 hours incubation at 37°C, 5% CO2. There were no differences in pneumococcal pre-or post-inoculum, or density between the strains, in the cell supernatant three hours post-infection.
Transmigration assay-Detroit 562 cells were cultured on 3μ m pore, PET Transwell Inserts (ThermoFisher) for 10 days to achieve confluent, polarised monolayers. Calu3 cells were plated onto Transwell inserts for 12 days and A549 cells for 6 days. Cell culture media was changed 1 hour prior to addition of bacteria to 1% FCS (250μl apical chamber, 1ml basal chamber). Resistance was recorded before and after S. pneumoniae were added using an EVOM2 (World Precision Instruments). 1mg/ml FITC-dextran (Sigma Aldrich) was added to the apical chamber of selected inserts to assess permeability. Approximately 12 million (± 5.7 × 106) bacteria were added to the cells (~MOI 1 cell : 25 bacteria). During the time course, 50μ l was removed, diluted and plated, from the basal chamber to measure bacterial load by counting CFU/well. Permeability was recorded using a FLUOstar Omega (BMG Labtech) at 488nm.
Inhibition assays - Detroit 562 cells cultured on 12 well plates were treated with 80μ M Dynasore (Cambridge Biosciences) and 7.5μ g/ml Nystatin (Sigma Aldrich) to block endocytosis; or 1μ M Cytochalasin D (Bio Techne Ltd) to block actin polymerisation, for 30 minutes prior to, and for the duration of pneumococcal infection incubation period. DMSO was used as a control. Cells were washed and treated with gentamicin and lysed in saponin as described above.
Confocal Microscopy
For the in vivo analysis, mucosal cells derived by curettage from the EHPC model were placed directly into 4% PFA for 1 hour. Cells were cytospun onto microscope slides and allowed to air dry. For the in vitro analysis, epithelial cell lines on transwell membranes were fixed in either 4% PFA (Pierce, Methanol Free) or 1:1 mix of methanol:acetone for 20 minutes. Cells were permeabilised with 0.2% Triton X-100 for 10 minutes and blocked for 1 hour in blocking buffer (3% goat serum and 3% BSA in PBS) before incubation with anti-6B pneumococcal antisera, JAM-A, Claudin 4 or β catenin primary antibodies (see Supplementary Information) for one hour and then secondary and/or conjugated antibodies for 45 minutes. DAPI solution was added for 5 minutes. After washing, the stained samples were mounted using Aqua PolyMount (VWR International) with a coverslip onto a microslide. The entire cytospin for each sample was manually viewed by microscopy for detection of pneumococci. Multiple fields of view were imaged for each transwell insert, for each condition. Images were captured using either an inverted LSM 700, LSM 880, or TissueFAXS Zeiss Confocal Microscope. Z stacks were recorded at 1μ m intervals at either 40x oil or 63x oil objectives.
Flow cytometry
For the in vivo analysis, two nasal scrapes were used per sample. Cells on rhinoprobes incubated in cold PBS++ (PBS supplemented with 5mM EDTA and 0.5% FCS) were dislodged by pipetting and centrifuged at 440g for 5 mins at 4°C. Supernatant was removed and cells resuspended in 25ul of PBS with Live/DeadTM Fixable Violet Dead Cell Stain (ThermoFisher). After 15 minutes incubation on ice, antibody cocktail (see Supplementary Information) was added and incubated for another 15 minutes. 500μ l of PBS++ was added to a 70μ m filter before vortexing the samples and adding 3.5mls of PBS and filtering over the wet filter. Samples were transferred to a 5ml FACS tube, centrifuged and resuspended in 200μ l Cell Fix (BD Biosciences). Samples were acquired on LSRII Flow Cytometer (BD Biosciences). Analyses of data was performed on the gated epithelial cell population and only samples containing 500 or more cells were considered for interpretation.
For the in vitro analysis, confluent monolayers of Detroit 562 cells on 6 well plates were incubated with S. pneumoniae for 6 hours in 1% FCS phenol free alpha MEM (base media, Life Technologies). Cells were washed three times in PBS and gently lifted from the plate using a cell scraper in 300μ l of base media supplemented with 1mM EDTA. Samples were transferred to 5ml FACS tubes and placed on ice for the duration of the protocol. Each cell sample was incubated with an antibody cocktail (see supplemental information) were added to the cells for 30 minutes before rinsing in 1ml base media and centrifuging at 300g for 5 minutes at 4°C. Cells were fixed in 600μ l of 4% PFA and run through LSR II Flow Cytometer (BD Biosciences). Compensation was run and applied for each experimental replicate and voltages consistent throughout. Isotype controls (BD Biosciences), FL-1 and single stains were also run for each experiment. Samples were acquired until 300,000 events had been collected. Analyses was performed using FlowJo version 10 software.
ELISAs
Supernatent from Detroit 562 cells that had been incubated with S. pneumoniae for 6 hours, was collected for cytokine analysis. IL-1beta, IL-6, IL-8, IFNg, TNFa, ICAM-1 DuoSet® ELISA kits were purchased from R&D Systems and protocol followed according to manufacturers2019; instructions.
RNA samples and sequencing (RNASeq)
Mucosal curettage samples and epithelial cell cultures (incubated with or without S. pneumoniae for 3 hours) were collected in RNALater (ThermoFisher) at −80C until extraction. Extraction was performed using the RNEasy micro kit (Qiagen) with on column DNA digestion. RNA was treated for DNA using Turbo DNA-free Kit (Qiagen) and cleaned using RNEasy Micro kit (Qiagen). Extracted RNA quality was assessed and quantified using a BioAnalyser (Agilent 2100). Library preparation and RNA-sequencing (Illumina Hiseq4000, 20M reads, 100 paired-end reads) were performed at the Beijing Genome Institute (China) or the Sanger Institute for mucosal curettage samples. In vitro samples used the KAPA Stranded mRNA-Seq Kit (Roche Diagnostics) to construct stranded mRNA-seq libraries from 500 ng intact total RNA after which paired-end sequencing was carried out using a 75-cycle high-output kit on the NextSeq 500 desktop sequencer (Illumina Platform, performed by the PGU, UCL).
Paired end reads were mapped to the Ensembl human transcriptome reference sequence (homo sapiens GRCh38, latest version). Mapping and generation of read counts per transcript were performed using Kallisto40, based on pseudoalignment. R/Bioconductor package tximport was used to import the mapped counts data and summarise the transcripts-level data into gene level41. Further analyses were run using DESeq2 and the SARTools packages42. Normalisation and differential analyses were run using DESeq2 by use of a negative binomial generalised linear model. The estimates of dispersion and logarithmic fold changes incorporate data-driven prior distributions. SARTools, which is an R pipeline based on DESeq2, was used to generate lists of differentially expressed genes and diagnostic plots for quality control. Using these techniques, cells exposed to different strains were compared against non-infected control cells, and extracted a result table with log2fold changes, Wald test p values and adjusted p values (according to false discovery rate, FDR).
Pathways and networks analyses were performed using XGR R package43. For each strain genes that were upregulated compared to the non-infected samples with an adjusted p-value (FDR) < 0.05 were selected. Network analyses were performed using as nodes the upregulated genes labelled with significance (FDR). Four gene subnetworks were generated using the Pathway Common database which contains directed interactions from a physical and pathways aspect. These new lists were then used for enrichment analysis (hypergeometric test) to identify enriched pathways from the REACTOME database. Enriched pathways were then represented in a heat map using log2 z-scores. REACTOME database has a non-structured list of terms, therefore terms were clustered based on overlapping genes. All heat maps were produced with a heat map R package using Euclidean distances and hierarchical clustering. The same gene lists were used to test for enrichment in Gene Ontology cellular components and membrane-related terms were selected. Upstream regulator analysis was performed in Ingenuity Pathway Analysis (IPA). Venn diagrams were generated using http://bioinformatics.psb.ugent.be/webtools/Venn/. in vivo data were processed with the same pipeline used for the in vitro experiments. Mapped reads ranged between 16M to 66M. Upregulated gene lists were produced and only genes with a log2 FC>1 were used for further pathway analysis. Pathway analysis with REACTOME database was performed with InnateDB. TPM for all genes were obtained and transformed into log2 scale. Quality control for 75 samples showed a batch effect due to two different labs sequencing the data. Combat function in the SVA R package was used44to reduce this effect. Principal component analysis identified an outlier that was removed for further analysis. Using the gene interactome lists for each strain from the in vitro data, a pan signature or module was obtained which included 200 genes that were upregulated in at least one strain. 16 genes were shared among all strains. Module scores for each group were derived by calculating the log2 average gene expression for each module. A non-parametric (Mann-Whitney) test was performed to compare carriers to non-carriers for each time point. Violin plots were produced with in house script in R and ggplot243.
Statistics
All experiments were conducted with replicates in three or more independent experiments unless stated otherwise. Error bars represent SEM unless stated otherwise. GraphPad Prism Version 10 was used to perform parametric (t-tests or ANOVA) or non-parametric (Mann-Whitney or Kruskal-Wallis tests) analysis, which was based on the Shapiro-Wilk normality test. P values lower than 0.05 were considered significant.
RESULTS
Streptococcus pneumoniae colonisation of the human nasal mucosal is associated with adhesion, microcolony formation and microinvasion
We have used an Experimental Human Pneumococcal Carriage Model45 to characterise pneumococcal-epithelial interactions in vivo. Colonisation was detected in 9/13 healthy volunteers by culture, 11/13 by microscopy and 9/11 by LytA PCR (Table 1). The carriage status of each volunteer in the study was blinded until sample collection was completed. Differences in the results obtained with each detection method may reflect methodological threshold detection, or the location and therefore the accessibility of the colonising pneumococci (e.g. in the mucus escalator vs. adherence to the epithelial cell surface). Nonetheless, all three methods demonstrated that colonisation was established and that clearance largely occurred between day 9 and 27 (Table 1 and Figure 1B).
Curette biopsy samples yielded intact sheets of epithelial cell associated with immune cells visualised by confocal microscopy (Figure 1A and Supplementary Figure 1A). Pneumococcal surface adhesion increased over time and was associated with microcolony formation (Figure 1E, middle and right panels). This provides evidence that the EHPC model represents true carriage and colonisation of the pneumococci. There was also evidence of pneumococcal microinvasion through the epithelial monolayer (Figure 1C and 1D) which comprised both endocytosis (Figure 1E left and middle panels) and paracellular migration (Figure 1E, left panel). Internalised pneumococci were also observed in immune cells (Supplementary Figure 1A).
Co-association between pneumococci and the junctional protein JAM-A was also observed (Figure 1F). JAM-A is a tight junction protein which is important for the regulation of barrier function in the respiratory epithelium46. These bacteria were either located at junctions between cells (left panel) or internalised inside cells (right panel). Junctional association of S. pneumoniae were also observed with nasal epithelial cells grown in culture ex-vivo, differentiated on an air-liquid interface for 30 days and then co-cultured with either 6B or 23F S. pneumoniae (Supplementary Figure 1B). Microcolony formation and internal pneumococci were also observed in these cell cultures.
Epithelial surface marker expression in response to Streptococcus pneumoniae in vivo
We stained the nasal curette biopsies for surface expression of IL-22Ra1, HLADR, CD40, CD54 or CD107a. Epithelial cells were identified by EpCAM expression (Supplementary Figure 2A-D for Flow Cytometry parameters). IL-22Ra1 is expressed exclusively on epithelial cells47 and is considered to protect the epithelial barrier and promote anti-microbial product secretion during infection, in response to IL-22 secretion by immune cells48,49. In the context of pneumococcal infection in mice, IL-22 appears to play an important role in carriage and clearance50,51. In the EHPC model, there was a trend towards increased expression of IL-22Ra1 at day 9, (when clearance starts to occur) in the carriage positive, compared to the carriage negative volunteers, although this did not reach statistical significance (Figure 2A, 2B top). We did not detect any change in the relative expression of inflammatory marker HLADR52 on the nasal epithelium in carriers vs. non-carriers over time (Figure 2A, 2B, second row). Similarly, epithelial surface expression of CD40, a costimulatory protein which binds CD154 (CD40L) and CD54 (Intercellular Adhesion Molecule 1, ICAM-1) a key leukocyte adhesion molecule which is also upregulated by CD4053-55, and plays a role in neutrophil migration and recruitment56-59, did not change over time. Finally, we assessed the expression of CD107a, also known as lysosomal associated membrane protein 1 (LAMP-1) which is a marker for natural killer cell activity60 and in the epithelium, forms the membrane glycoprotein of lysosomes and endosomes27. CD107a has been shown to be cleaved during infection with Neisseria species which are also extracellular mucosal pathogens61. Although the number of epithelial cells expressing CD107a did not change over time (Figure 2A, bottom), we did observe an increase intensity of expression at day 2 post inoculation in carriage positive volunteers vs. carriage negative volunteers, which was maintained throughout the remainder of the time course.
Epithelial adherence, endocytosis and transmigration by varies by Streptococcus pneumoniae varies by pneumococcal strain and is modulated by pneumolysin
To further investigate our observations from the EPHC, we undertook epithelial co-culture experiments with the cell line Detroit 562, derived from a nasal pharyngeal carcinoma. We used the EHPC 6B strain and two other representative clinical isolates serotype 4 (TIGR4, the original sequences strain) and 23F. Any differences between these strains are not simply explained by differential growth during colonisation (data not shown). Although they may be partially explained by capsule serotype37,62, TIGR4, 23F and 6B genome comparisons have revealed ~15,000 single nucleotide polymorphisms and insertion/deletion mutations (SNPs and INDELS) not related to capsule (data not shown). We have also used a TIGR4 strain where pneumolysin, a key virulence factor that is associated with pore-forming induced inflammation15,23 has been knocked out (kind gift from Prof. TJ Mitchell, University of Birmingham, UK).
Strikingly, the number of TIGR4 pneumococci associated with the Detroit 562 cells was ten-fold higher than 6B or 23F strains (Figure 3A). This pattern was also observed by immunofluorescence (Figure 3D and 3E, and Supplementary Figure 4). Epithelial adhesion was associated with internalisation within the cells (Figure 3B) into what appeared to be intracellular vesicles that were coated in host proteins, in this case JAM-A (Figure 3G and Figure 5A), which indicates vesicular endocytosis and using bacteria that are pre-stained with FAMSE, we were able to distinguish extracellular bacteria (blue) from those below the apical surface prior to permeabilization of the cells (green, Figure 3F). Interestingly, co-association with another tight junction protein, Claudin 4, was not readily observed, while occasional co-association with the adherens junction protein β catenin was observed (Supplementary Figure 4A and 4B).
To assess transmigration, pneumococci that had penetrated the basal chamber of cells cultured on transwell inserts were counted. Although only statistically significant at 1hr between 23F and TIGR4, 23F was more readily detected compared to the other strains (Figure 3C). By microscopy, we observed laterally located bacteria, and pneumococci zip-wiring between cell junctions (Figure 3H). We observed pneumococci at the level of the nuclei and below the basal membrane (Figure 3I). This was more readily, but not exclusively, seen with the 23F strain. These data demonstrate a similar pattern of interaction between the S. pneumoniae 6B strain and human epithelium in vivo and in vitro; and show that the relative prominence of adhesion, endocytosis and paracellular transmigration varies by genotype.
Pneumolysin deletion in the TIGR4 mutant showed a significant increase in internalisation (Figure 3B) and an increase in transmigration capacity (Figure 3C). These data suggest that interactions within epithelial cells are in part, regulated by pneumolysin.
Loss of epithelial cell barrier function is not a pre-requisite for microinvasion by S. pneumoniae
Several mucosal pathogens including S. pneumoniae, are known to directly, and indirectly affect the integrity of epithelial barriers and tight junction function25,27,28,61,63-67. It was not possible to directly assess epithelial barrier function during colonisation in the EHPC model. We have therefore explored the possibility that over the same time frame where we have observed pneumococcal adhesion and microinvasion, epithelial surface molecule upregulation and cytokine production in vitro, there is epithelial barrier function disruption. Trans-epithelial electrical resistance (TEER) is not high in Detroit 562 cells but nevertheless 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 quantified from the basal chamber. With the Detroit 562 cells, a significant reduction in permeability was seen with 6B pneumococci, 23F and TIGR4 (23 – 34%), compared to non-infected cells (Figure 3K). This implicates a role for pneumolysin in epithelial integrity.. These data suggest that loss of epithelial cell barrier function is not a pre-requisite for pneumococcal adhesion and microinvasion, and that as described in murine models23,25, changes to barrier function appear pneumolysin dependent.
To explore the possibility that our findings were not cell line dependent, we also used A549 cells (Supplementary Figure 3A-E), which are undifferentiated alveolar Type II pneumocytes, and Calu3 cells (Supplementary Figure 3F-J), which represent a more polarised and differentiated cell originally derived from bronchial submucosa. Pneumococcal behaviour with both cell lines was similar, although absolute intensity of adhesion, microinvasion and transmigration differed (Supplementary Figure 3A and 3F, Supplementary Figure 3B and 3G, and Supplementary Figure 3C and 3H, respectively). This may in part be due to the high expression of polymeric Immunoglobulin Receptor found on Detroit cells63, and differential barrier function with A549 cells having the least trans-epithelial electrical resistance and the most permeability (Supplementary Figure 3D and 3E, respectively), and Calu3 having the greatest trans-epithelial electrical resistance and the least permeability (Supplementary Figure 3I and 3J, respectively). Importantly, no change in barrier function was seen with these cells. As with Detroit 562 cells, for A549 and Calu3 cells TEER (Supplementary Figure DA and 3I) and permeability (Supplementary Figure 3E and 3J) was preserved following exposure to pneumococci for three hours. Indeed, if anything, epithelial co-culture resulted in an enhanced barrier function as shown via an increase in TEER (Supplementary Figure 3D) and a decrease in permeability to TIGR4 (Supplementary Figure 3E) in A549 cells.
Streptococcus pneumonia upregulates epithelial surface CD54 and CD107a in vitro
It is uncertain whether the observed, at best, modest surface changes seen in the 6B EHPC model reflect a relatively silent host response to colonization by this strain, or was confounded by inter-volunteer variation. We have therefore compared the impact of the pneumococcal strains on Detroit 562 cell expression of the same range of surface markers. (See Supplementary Figure 2E-G for Flow Cytometry parameters). There was no significant change in epithelial markers IL-22Ra1, HLADR or CD40 (Figure 4A and 4B). However, although not seen with the 6B or 23F strains, CD54high expression was significantly greater for TIGR4 and dPLY strains, compared to non-infected cells. Epithelial CD107a, which has previously been implicated in pneumococcal endocytosis27 was upregulated in response to the 6B, 23F and TIGR4 strains (Figure 4A and 4B, bottom graphs), but was not seen with the pneumolysin TIGR4 mutant. These data again demonstrate a similar pattern of interaction between the S. pneumoniae 6B strain and human epithelium in vivo and in vitro implicating pneumolysin in the induction of CD54 but not CD107a surface expression.
S. pneumonia upregulates epithelial inflammatory cytokines and soluble CD54 in vitro
To further investigate the inflammatory potential of the epithelium, measured for IL-6, IL-8 and CD54 secretion in the supernatants of Detroit 562 cells following incubation with pneumococci (Figure 4C). We have detected a significant increase in IL-6 and IL-8 (P = <0.0001), which was not entirely pneumolysin dependent. In line with the surface marker observations, only TIGR4 significantly upregulated the secretion of soluble CD54 (P = 0.0013), which was dependent on the presence of pneumolysin (Figure 4C). Other cytokine responses to S. pneumoniae, such as IFNΓ, IL-1β and TNFα were below the limits of detection (Data not shown).
Internalised pneumococci do not replicate within the epithelium
S. pneumoniae is generally considered to be an extracellular bacterium68. However, since we and others have observed intracellular bacteria27, we wanted to test whether they remain viable, can replicate, and egress from the epithelial cell. Pneumococci that adhere on the epithelial cell surface are capable of replicating, as demonstrated by epithelial surface microcolony formation in the 6B EHPC model (Figure 1E). In contrast, the pneumococci that were identified by confocal microscopy to be intracellular, were often single bacterial cells, co-localised with host proteins (Figure 5A) and did not appear to increase in number over time (Figure 5B). Bacteria that had transmigrated across the epithelial monolayer in vitro did replicate and remained viable for at least three hours post removal of the transwell insert (Figure 5C). To test the hypothesis that intracellular migration is not permissive for bacterial growth, we co-cultured pneumococci with epithelial cells for three hours, treated with gentamicin for one hour, replenished the media and recorded CFUs over time from the apical and basal chamber of transwell inserts (Figure 5D and E, respectively). Although bacteria were detected at low levels, replication was not readily apparent. To test whether these bacteria transmigrated across the cells in a transcellular or paracellular manner, we inhibited endocytosis by cellular treatment with Dynasore and Nystatin, or actin polymerisation by cellular treatment with Cytochalasin D. We found that the inhibition of endocytosis prevented transmigration but the inhibition of actin polymerisation enhanced transmigration in Detroit 562 cells with 23F pneumococci (data not shown).
Streptococcus pneumonia induces epithelial innate transcriptomic responses that is influenced by the pattern of epithelial adhesion and microinvasion
To further explore the hypothesis that the pattern of epithelial adhesion and microinvasion results in differential epithelial sensing and therefore epithelial inflammatory-response genes, we performed RNAseq and obtained transcriptomic data from our pneumococci infected Detroit 562 cells. As shown in Figure 6A, we found that TIGR4 upregulated 1127 genes (550 unique genes), 23F upregulated 650 genes (69 unique genes), and 6B upregulated only 153 genes (10 unique genes) compared to non-infected cells. The pneumolysin mutant upregulated 220 genes (14 unique genes). 93 genes were upregulated by all strains compared to non-infected cells. These findings appeared to reflect the invasive and inflammatory nature of these bacteria in this in vitro epithelial model. To further explore the nature of these differences, we performed pathway analyses using the REACTOME database and performed with XGR(Figure 6B). Again, we found that the upregulated pathways for TIGR4 and 23F were pro-inflammatory, but that the 6B profile was relatively silent. For example, TIGR4 upregulated pathways involved innate immunity, such as TLR signalling, cytokine signalling and stress responses. In comparison, 6B increased pathways involved in NOD and NRL signalling, and gene regulation. The TIGR4 pneumolysin mutant transcriptomic profile suggested that pneumolysin modulates epithelial cell RIG-I/MDA5 mediated induction of IFN-α/β activation of IRF and NFκB pathways. In line with the cytokine profiles that we observed at the protein level, we detected upregulation of IL-6, IRAK2, TNFAIP3 and CD54 genes (Figure 6C) within the innate immune pathways selected (Supplementary Figure 5A). In line with our previous observation, 6B elicited the least and TIGR4 the greatest transcriptomic response. Given the pneumococcus interacts with the epithelial cell surface, analysis of genes associated with host cell membrane components were analysed at the transcriptomic level (Figure 6D) using Gene Ontology database (Cellular component terms only, Supplementary Figure 5B). Analysis of genes associated with host cell membrane components showed that the tight junction protein Claudin 4 was upregulated in response to 23F and TIGR4. Claudin 4 is normally associated with a tight barrier in epithelial cells69,70, which would support our hypothesis that the epithelium responds to preserve barrier function during co-culture.
Further bioinformatics analysis of upstream regulators revealed that RELA, or the nuclear factor NFκB p65 subunit is likely to be a key mediator of these pneumococcal-epithelial interactions (Figure 6E, Supplementary Figure 5C). Comparisons between the strains again reveal a more silent upstream profile with 6B compared to TIGR4 or 23F.
Epithelial transcriptomic responses to Streptococcus pneumoniae in vivo are most marked around the time of bacterial clearance
To test whether the relatively silent transcriptomic profile seen with the S. pneumoniae 6B strain during in vitro co-culture, was also present in vivo, we have first had to design an approach to focus on the epithelial response in the curette biopsy tissue. Using the epithelial transcriptomic data, we have derived an integrated transcriptome signature that allows us to interrogate the epithelial RNAseq transcriptomic response obtained from the 6B EHPC model (Supplementary Figure 6A). Within the cohort (Supplementary Figure 6B), the number of significantly upregulated genes was low (Figure 7A). However, looking at the genes average of the epithelial signatures’ genes (200 genes for the Integrated and 16 genes for the Core signature), we observed a shift in gene expression following 6B inoculation that was maximal at day 9, coinciding with the time of maximal bacterial clearance (Figure 7B, and Figure 7C, Integrated and Core signatures). Qualitatively, there was a shift from generic homeostasis at baseline towards a metabolic and innate defence profile at day 2, and surface receptor upregulation and inflammatory signalling pathways by day 9 (Figure 7A).
DISCUSSION
The upper respiratory tract is at the centre of the control of colonisation by a wide range of commensal bacteria. For some more pathogenic members of this commensal community, epithelial sensing and the triggering of inflammation may result in bacterial clearance but may also promote onward transmission. By combining in vitro cell culture systems and the EHPC model, we have shown that human epithelial sensing of the pneumococcus is enhanced by microinvasion, resulting in an epithelial inflammatory/innate immune response that is temporally associated with clearance.
We have demonstrated that the pneumococcus interacts with the human respiratory epithelium and that the innate epithelial cell response is dependent on the association of the bacteria (Figure 8). We show that colonisation leads to adherence, microcolony formation and microinvasion within the epithelium, which results in activation of signalling pathways that lead to cytokine and chemokine upregulation, biochemical and metabolic pathway enrichment. However, although microinvasion does not support bacterial growth, co-association with junctional proteins provides a possible mechanism for migration across the barrier, that could ultimately affect transmission or cause invasive disease. We provide evidence of epithelial sensing of the pneumococcus that coincides with clearance in the EHPC model.
The occurrence of microinvasion during colonisation in healthy individuals is supported by murine colonisation experiments71 and the detection of pneumococcal DNA in the blood of healthy colonised children72. Using primary and immortalised epithelial cell line models that mirror this process, and in line with other cell culture and murine models64,66,71,73, we have demonstrated that pneumococcal microinvasion occurs by endocytosis and the formation of cytoplasmic vacuoles, and by paracellular transcytosis. Transcriptomic analysis of the epithelial response in vitro and in vivo has revealed that the pattern of pneumococcal epithelial surface adherence and microinvasion determines the host cell response through a range of innate signalling and regulatory pathways, inflammatory mediators, adhesion molecules, cellular metabolism and stress response genes. These data support the view that beyond forming a physical barrier, secreting mucus, and modulating the transport of immunoglobulins, the epithelium plays a critical role in the regulation of these complex host-pathogen interactions19,74,75.
Nasal colonisation in murine models is proinflammatory24,76 and is associated with epithelial microinvasion and tight junction modulation25. Our in vitro epithelial model co-culture with a serotype 6B strain suggests that this is not always the case with only modest pneumococcal-host cell adherence, endocytosis and paracellular migration, and a relatively silent epithelial inflammatory profile. Indeed, volunteers who undergo EHPC generally remain clinically asymptomatic and this silent transcriptomic pattern of epithelial response is mirrored in the 6B EHPC model, where we have observed surface adherence, microcolony formation and some microinvasion.
After three hours infection in vitro, we did not observe a breakdown of epithelial barrier function. Previous studies in mice and human lung tissue that have investigated infection over longer periods of incubation, have seen tight junction dysregulation25,64, and we did observe co-association with junctional proteins such as JAM-A and β catenin. IN accordance with studies in human alveolar cells64, Claudin 4 was not affected by the pneumococcus, although we did detect a transcriptomic upregulation in Detroit cells. In mice, changes in claudin regulation was TLR dependent 25,77 and we detected TLR4 and TLR3 activation in our Detroit 562 cells transcriptomic analyses, in response to all strains of the pneumococci we tested. Previous studies have implicated pneumolysin to activate TLR2 and TLR4 stimulated cytokine release, such as IL-6 and IL-8, both of which we detected in infected Detroit 562 cell supernatants30,78. Interestingly, TLR3 is normally associated with double stranded RNA detection of viruses, such as Influenza79. There appears to be a relationship between the outcome of infection between Influenza and S. pneumoniae and 79,80, which may be important for understanding the dynamics of flu vaccination success. TLR3 leads to the activation of IRF3 and the secretion of type 1 interferons81,82. Type 1 interferons have been shown to be stimulated in response to murine pneumococcal infection that leads to bacterial clearance83,84. Here, the authors also show anti-microbial product secretion, and we detected evidence of β defensin gene upregulation in the EHPC model two days post inoculation with 6B (Figure 7A). We also detected upregulation of RIG-I/MDA5 mediated induction of IFN-α/β pathways following in vitro stimulation with the TIGR4 pneumolysin mutant, providing further evidence that sensing by the epithelium may be important. DNA sensing of S. pneumoniae has been demonstrated in alveolar macrophages, where secretion of Type 1 Interferons led to upregulation of STING and the transcription factor IFN regulatory factor 3, augmented by pneumolysin76,84.
Pneumolysin, a pore forming toxin, has been implicated as a major virulence factor contributing to host inflammation and transmission23,85. We found pneumolysin to be a prominent trigger of epithelial surface molecule upregulation, cytokine production and the transcriptomic inflammatory response in vitro. The prominence of TLR4 signalling pathways in the transcriptomic profile observed and the presence of TLR4 on epithelial cells, implicates pneumolysin. Mediated by autolysin, the pneumococcus undergoes autolysis when reaching stationery growth phase, resulting in the release of additional PAMPs including bacterial DNA. We therefore suggest that in the context of microinvasion, pneumococcal DNA may act as an alternative epithelial sensing agonist to induce inflammation. Furthermore, cellular entry of DNA may be enhanced by pneumolysin pore formation84. In mice, pneumococcal DNA triggers inflammation through a DAI/STING/TBK1/IRF3 cascade76,84, a response that contributes to pneumococcal clearance. Indeed, we observed an increased epithelial expression of the lysosomal membrane protein CD107a, following pneumococcal co-infection.
Our findings are limited by the number of pneumococcal strains that can be safely tested in an EHPC model to enable direct comparisons between the in vivo and in vitro data. Nonetheless, the use of different strains in vitro has enabled us and others73,86 to interrogate the impact of different patterns of epithelial adherence and invasion on the host inflammatory/ innate immune response. Transcriptomic analysis has enabled us to postulate the potential epithelial sensing pathways but these will need to be fully defined in more precise model systems. We have placed considerable reliance on the findings from immortalised cell lines from relevant tissue but our findings have been reassuringly paralleled by our findings in primary cell lines derived from the EHPC and the EHPC itself.
Our data highlight the complex interactions between the host epithelium and S. pneumoniae whereby pneumococcal microinvasion may ultimately dictate the outcome of colonisation, altering the delicate balance between inflammation-mediated transmission and clearance (Figure 8). Ultimately, epithelial sensing of pneumococcal-epithelial interaction and its outcome may be dictated by the bacterial strain, the force of infection, or the frequency of co-colonisation of pneumococcal strains, (more important in children and high carriage prevalence populations), viral co-infections and other environmental pressures1,20,87. Measures of human to human transmission are needed to fully understand the critical pathways and a mechanistic insight into the impact of pneumococcal vaccine on epithelial adhesion and invasion is required if we are to improve herd protection.
Author contributions
CMW, SPJ, DMF and RSH conceived and designed the study. CMW, SP, SPJ, JR, EN, CS, CA acquired the data. CMW, CV, SPJ, MN, JSB, DMF, RSH analysed and interpreted the data. CMW wrote the first draft of the manuscript. CMW, CV, SP, SPJ, JR, EN, CS, CA, MN, JSB, DMF, RSH commented on and approved the manuscript.
Conflicts of interest
the authors declare no conflicts of interest.
Acknowledgements
This study was funded by the Wellcome Trust (Grant 106846/Z/15/Z). DF is supported by the Medical Research Council (grant MR/M011569/1), Bill and Melinda Gates Foundation (grant OPP1117728) and the National Institute for Health Research (NIHR) Local Comprehensive Research Network. LytA PCR was performed by Prof. D Bogaert, University of Edinburgh, UK. RNAseq library preparation undertaken at UCL was provided by the Pathogens Genomic Unit. Confocal imaging facilities at LSTM were funded by a Wellcome Trust Multi-User Equipment Grant (104936/Z/14/Z). Flow cytometric acquisition was funded by a Wellcome Trust Multi-User Equipment Grant (104936/Z/14/Z).
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.
- 9.
- 10.↵
- 11.↵
- 12.
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.
- 55.↵
- 56.↵
- 57.
- 58.
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.
- 66.
- 67.
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.
- 75.
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.