A novel Streptococcus pneumoniae human challenge model demonstrates Treg lymphocyte recruitment to the infection site

To investigate local tissue responses to infection we have developed a human model of killed Streptococcus pneumoniae challenge by intradermal injection into the forearm. S. pneumoniae intradermal challenge caused an initial local influx of granulocytes and increases in TNF, IL6 and CXCL8. However, by 48 h lymphocytes were the dominant cell population, mainly consisting of CD4 and CD8 T cells. Increases in local levels of IL17 and IL22 and the high proportion of CD4 cells that were CCR6+ suggested a significant Th17 response. Furthermore, at 48 h the CD4 population contained a surprisingly high proportion of likely memory Treg cells (CCR6 positive and CD45RA negative CD4+CD25highCD127low cells) at 39%. These results demonstrate that the intradermal challenge model can provide novel insights into the human response to S. pneumoniae and that Tregs form a substantial contribution of the normal human lymphocyte response to infection with this important pathogen.


To investigate local tissue responses to infection we have developed a human model of killed
Streptococcus pneumoniae challenge by intradermal injection into the forearm. S. pneumoniae intradermal challenge caused an initial local influx of granulocytes and increases in TNF, IL6 and CXCL8. However, by 48 h lymphocytes were the dominant cell population, mainly consisting of CD4 and CD8 T cells. Increases in local levels of IL17 and IL22 and the high proportion of CD4 cells that were CCR6 + suggested a significant Th17 response. Furthermore, at 48 h the CD4 population contained a surprisingly high proportion of likely memory Treg cells (CCR6 positive and CD45RA negative CD4 + CD25 high CD127 low cells) at 39%. These results demonstrate that the intradermal challenge model can provide novel insights into the human response to S. pneumoniae and that Tregs form a substantial contribution of the normal human lymphocyte response to infection with this important pathogen.
Despite the widespread use of antibiotics and vaccination bacterial pathogens remain an important cause of mortality and morbidity globally 1,2 . Bacterial infection will cause a local and sometimes a systemic inflammatory response, recruitment to the site of infection of both antigen-specific and other white cell populations, and systemic effects on innate and humoral soluble immune effectors, including antibody. A detailed understanding of these pathogen/host interactions is essential to define how individual pathogens establish serious infections and why certain subjects are at higher risk. However, the available methodologies for investigating host pathogen interactions have significant weaknesses. For example, the effects of anatomy, cell to cell interactions, and the large range of different structural and immune cells involved in bacterial infection cannot be accurately replicated using in vitro experiments with human cells, and data obtained from animal models may not reflect bacterial interactions with human cells 3 . Furthermore, animal models usually do not replicate the complex immune background found in humans, who may have adaptive immune responses to the pathogen under investigation due to previous exposure (often on multiple occasions), and in whom co-morbidities or the effects of age can substantially alter immune responses.
Human infection challenge models have been used to overcome some of these limitations of cell and animal infection models of infection. For example, a nasopharyngeal human colonisation model has been established for the important Gram positive pathogen Streptococcus pneumoniae, a common nasopharyngeal commensal species that is also a common cause of severe bacterial infections such as pneumonia, septicaemia and meningitis and causes approximately 1.3 million deaths per year of death 4,5 . Live S. pneumoniae human infection models have made important novel findings about innate and adaptive immune responses to nasopharyngeal colonisation, as well as bacterial/epithelial interactions [6][7][8][9][10] . In addition, the model can be used to test the efficacy of vaccines 11,12 . However, the human challenge colonisation model only reflects bacterial/host interactions at a mucosal site during colonisation, and there is a need for an additional human model that can be used to characterise the inflammatory and adaptive immune response to systemic S. pneumoniae infection, including the effects of different host factors (eg age, immunosuppression) on these responses. The inflammatory and immune response to S. pneumoniae have been defined for animal models, which show that S. pneumoniae pneumonia initially causes a rapid influx of neutrophils to the sites of infection associated with high levels of pro-inflammatory cytokines

Results
An intradermal killed S. pneumoniae challenge model causes a local inflammatory response. Healthy volunteers were given intradermal injections of 7.5 × 10 5 UV-killed S. pneumoniae TIGR4 strain suspended in 100 μl of saline into each forearm. A variable sized patch of erythema developed at the site of infection, fading by 48 h (Fig. 1A). There were no systemic side effects, skin breakdown, or persisting skin changes. Blood flow at the injection site at the injection site was measured using Laser Doppler scanning (Fig. 1B), and was increased at 4 h post-injection before peaking at 24 h (Fig. 1B,C). By 48 h post-challenge, blood flow and visible inflammation had returned to levels close to baseline. These data confirm intradermal injection of killed S. pneumoniae caused an inflammatory response that was detectable by 4 h post-challenge, peaked at approximately 24 h, and was resolving by 48 h.
Cell recruitment in response to intradermal S. pneumoniae injection. To analyse inflammatory cell recruitment at the site of intradermal S. pneumoniae challenge a blister was raised at the site of injection using a suction chamber at 4 h and 48 h post-injection (Fig. 1D). Fluid was aspirated from the blister, and its cellular content analysed using flow cytometry. Cells were gated for granulocytes, agranulocytes, CD3 − agranulocytes, and CD3 + T cells, which were further separated into CD8 + , CD4 + and CD4 + CD25 high CD127 low subsets (Supplementary Fig. 1). The full gating strategy was validated using blood samples ( Fig. 2A,C,E left hand panels). Initially granulocytes were the majority cell type present representing 77.7% ± 11.4% of all cells at 4 h post-injection but their absolute numbers decreased by 48 h (4 h 1.24 × 10 5 ± 2.08 × 10 5 cells per ml, 48 h 1.48 × 10 4 ± 1.58 × 10 4 cells per ml), and the proportion of granulocytes had decreased to 17.6% ± 14.6% of total cells by 48 h (Fig. 2A,B). Although the data were variable between volunteers, overall the cell numbers for the CD3 − agranulocyte population increased from 1.31 × 10 4 ± 1.11 × 10 4 cells per ml at 4 h to 4.52 × 10 4 ± 2.79 × 10 4 cells per ml at 48 h, suggesting increased monocyte recruitment over 48 h (Fig. 2C

Discussion
Accurately defining human immune and inflammatory responses to challenge with S. pneumoniae is essential for our understanding of disease pathogenesis. An effective immune and inflammatory response is necessary for bacterial clearance, but can also have negative consequences for the host causing consolidation of the lung during pneumonia, neuronal damage in meningitis, and severe complications of infection such as septic shock or acute respiratory distress syndrome (ARDS) 33,34 . We have described a new model of intradermal injection of killed S. pneumoniae human challenge and demonstrated that the model can define the inflammatory and immunological responses at the site of infection.
Using the model, we have shown that after intradermal injection of killed S. pneumoniae there is a large increase in local blood flow to the site of injection, an initial influx of granulocytes (likely neutrophils), followed by a more delayed influx of monocytes and lymphocytes. The early 4 h cytokine/chemokine response was dominated by TNF, IL6 and CXCL8; of these only the IL6 response persisted at 48 h. This pattern of the initial innate response to S. pneumoniae is very similar to that described using mouse models of infection 13,17,18 , and compatible with the data obtained using samples from human infection which shows a marked neutrophilic infiltrate into the lungs during pneumonia associated with high levels of inflammatory cytokines 35,36 . The marked dynamic changes over time in white cell populations and cytokines demonstrate the blister fluid data reflects a strong dermal inflammatory response to injection of killed S. pneumoniae. For comparison, previously published data show the response to intradermal injection of PBS alone resulted in very low blister fluid cytokine levels (eg under 10 pg/ml for TNFα, IL-1β, and IL-6), and neutrophil and T cell numbers that were approximately 2 log 10 and 1 log 10 lower respectively than seen in response to S. pneumoniae 37 . Our data show that the intradermal challenge model can be used to assess cellular recruitment and cytokine responses to S. pneumoniae at specified timepoints, and could be used to investigate how these responses are affected by host factors such as treatment with immunomodulators, age, or disease. However, future studies should include PBS treated data as controls to ensure differences between different subjects reflects the response to S. pneumoniae rather than to the injection and blister formation. Responses could also be compared between different S. pneumoniae strains including genetically modified strains missing specific virulence factors or pro-inflammatory pathogen associated molecular patterns (eg lipoproteins) 38 . The model is based on a previously described model using intradermal injection of killed Escherichia coli that described innate immune defects in patients with inflammatory bowel disease [39][40][41] , demonstrating the potential for the model to identify clinically significant differences between patient groups. As the model recovers significant numbers of immune and inflammatory cells from the site of infection, it has the potential to allow a very detailed analysis of immune responses to S. pneumoniae using single cell RNAseq or isolation of antigen specific lymphocytes for further molecular characterisation.
Interpretation of data obtained using the intradermal model needs to take into account that for safety reasons the model uses killed bacteria. Hence, the model will not identify specific effects of bacterial/host interactions that depend on live bacteria such as active cell invasion and modulation of host signalling pathways. The model is limited to investigating the response to a single short lived exposure to S. pneumoniae, with each subject having two samples in total taken (e.g. 4 and 24 h, or 4 and 48 h). The significant variability in results between subjects and the low numbers of cells obtained from the blister fluid will make it hard to identify smaller differences between groups of subjects (eg due to age or sex) or time points, and provides technical limits on the assays that can be easily performed. Significantly larger numbers of subjects are likely to be needed to identiy biological differences between patient groups. The skin is not a natural site of S. pneumoniae infection and although the   www.nature.com/scientificreports/ data will likely reflect what happens at inaccessible infection sites such as the lung and brain, there are also likely to be some differences. However, skin infection has been used successfully to characterise systemic alterations in human responses to infection 40 , and our data indicate that immunological responses to S. pneumoniae in the skin resemble those previously described for the lung. For example, as discussed above the innate responses seen after intradermal challenge with killed S. pneumoniae were very similar to those seen in mouse lung after live S. pneumoniae infection. Furthermore, in contrast to blood CD4 + T cells very few lung resident CD4 + T cells are CD45RA +42 , and this resembles the CD4 + T cell response after intradermal injection of S. pneumoniae www.nature.com/scientificreports/ shown here. Combining data from the established live pneumococcal nasopharyngeal colonisation model and from acute studies in humans with natural infection 35 with the detailed immunological data obtained using the intradermal model will provide a comprehensive overview of human innate and adaptive responses to S. pneumoniae. Animal studies directly comparing immune responses to live and dead S. pneumoniae in mice with adaptive immune responses to S. pneumoniae from previous exposure would be helpful in interpreting the effects of using killed bacteria in the intradermal human challenge model. We have used the intradermal challenge model to define the lymphocyte response to S. pneumoniae challenge in humans. The results show that 4 h post-challenge there are only very low levels of lymphocytes in blister fluid, but by 48 h lymphocytes are the dominant cell population. This influx of lymphocytes was dominated by T cells rather than B cells. There was a slightly greater relative increase in CD8 + T cells compared to CD4 + T cells with these two subsets each contributing about 40% of the recruited lymphocyte population at 48 h. The functional relevance of this high level of recruited CD8 + T cells for immunity to an extracellular bacterial pathogen is not clear. Previous investigation 43 using a mouse model of S. pneumoniae infection in immune naïve animals demonstrated an unexpectedly important protective role for CD8 + T cells; the mechanism(s) involved remain unclear. The increase in blister fluid levels of IL17 and IL22 and the high proportion (83%) of CD4 + T cells that were CCR6 + at 48 h suggest that recruited CD4 + T cells included a significant population of Th17 cells. Corresponding with these findings, mouse models have shown important roles for enhanced Th17 responses to reinfection including from CD4 + T resident memory T cells to naturally acquired protection against re-infection with S. pneumoniae 44,45 . The CD4 + T population also consisted of a surprisingly high number of likely Treg cells (CD4 + CD25 high CD127 low , 39% of all CD4 + T cells at 48 h). Most of these potential Tregs cells were CCR6 positive and CD45RA negative suggesting they are memory cells recruited in response to S. pneumoniae, but their functional role is unclear. There are only very limited data on the role of Tregs during S. pneumoniae infection; mouse data suggest Tregs are protective, modulating the duration of S. pneumoniae nasopharyngeal colonisation and preventing the development of septicaemia during pneumonia, perhaps by preventing epithelial/endothelial barrier breakdown 6,19 . However, these data were obtained using immune naïve mice and also run counter to data showing Tregs usually increase susceptibility to infection with other pathogens by suppressing inflammatory responses 46,47 . The recruitment of Tregs to the site of S. pneumoniae infection may be important for resolution of the inflammatory responses. The intradermal model provides a method for defining this potential role of Tregs during S. pneumoniae infection using human tissue and could be important for our understanding of how pneumonic consolidation usually resolves without causing persistent lung damage. Confirmation of the functional identify of the different CD4 + T cell populations recruited to the blister fluid at the site of S. pneumoniae intradermal challenge would require additional assays such as intracellular cytokine staining (Th17 CD4 + T cells) or suppression of inflammatory responses (CD4 + CD25 high CD127 low as Tregs), but any additional analyses to flow cytometry were prevented by the low numbers of recovered cells. More detailed characterisation of the CD4 + T and CD8 + T cell sub-populations, their influence on the immunological and inflammatory response to S. pneumoniae, and the identification of which antigens they recognise are all important subjects for future investigation using the intradermal challenge model.
In summary, we describe a novel safe human challenge model using intradermal injection of killed S. pneumoniae that results in an acute inflammatory response similar to that seen in animal models of pneumonia, and which was then followed by an influx of lymphocytes by 48 h which included surprisingly high proportions of CD8 + T cells and probable memory Tregs cells. This intradermal model can be used to characterise in detail the human immune response to S. pneumoniae challenge, and can therefore to investigate why age and a range of diseases and/or treatments that modulate the immune response are associated with increased susceptibility to S. pneumoniae. Furthermore, the model could be readily adapted for investigating multiple other bacterial pathogens, thereby extending the range of infections for which relevant human data can be obtained.

Methods
Healthy volunteers. A total of 11 (7 male, 4 female) healthy volunteers aged 18 to 35 years old were recruited for this study, with lymphocyte data obtained from a subset of 6 subjects (5 male, 1 female). None of the subjects had received vaccination against S. pneumoniae. The experiments were approved by the UCL Research Ethics Committee (ref. 7577/001) with written informed consent obtained from all participants, and all methods were performed in accordance with the relevant guidelines and regulations.
Preparation of UV-killed S. pneumoniae. Capsular serotype 4 S. pneumoniae TIGR4 (a gift from Professor Jeffrey Weiser, University of Pennsylvania) was grown overnight on Columbia agar blood plate (SGL, 8022) at 37 °C in 5% CO 2 , then transferred into 15mls of autoclaved THY broth (Sigma Aldrich, T1438) supplemented with 0.5% yeast extract (Sigma Aldrich, Y1625) in a 50 ml falcon tube (Greiner, T2318-500EA) and cultured at 37 °C in 5% CO 2 with the cap loosely replaced. On reaching an optical density 600 of 0.4 (approximately 1 × 10 8 colony forming units [CFU] per ml) the S. pneumoniae cultures were centrifuged at 13,000×g for 10 min, the supernatant discarded, and bacteria re-suspended in sterile PBS (Sigma Aldrich, D8537) in a sterile petri dish then exposed to UV light (302 nm, ChemiDoc; Bio-Rad, UK) for 1 h. The UV-killed S. pneumoniae were collected into a sterile 50 ml falcon tube, washed with sterile normal saline and centrifugation at 13,000×g for 10 min. Aliquots of 1.5 × 10 8 UV-killed S. pneumoniae in 1 ml of sterile saline were frozen at −80 °C in autoclaved 10% glycerol (VWR 24,388.260) diluted in distilled water until required for injection. The S. pneumoniae in the UV stocks were confirmed to be dead by culture (probable limit of detection of < 1 in 1 million bacteria) by the University College London Hospitals Microbiology department. At pre-specified time points, a suction chamber with a 10 mm diameter hole connected to a negative pressure instrument (NP-4, Electronic Diversities Ltd., MD, USA) was secured over the site of injection, and negative pressure gradually applied until a fully formed blister was visible. The pressure was then gradually returned to the baseline, and the blister pierced with a 23G needle (Fisher Scientific, NN-2332R) and the fluid collected using a 200 μl pipette. The blister area of the forearm of the volunteers was cleaned with 0.5% cetrimide spray (Savlon) and a large protective dressing applied (Mepore).
Laser doppler imaging. At specified time points before blister induction doppler scans were taken of the intradermal injection site using a Laser Doppler Imager (Moor LDI-HIR, Moor Instruments Ltd, UK) and the colour-coded image analysed using moorLDI software (Version 5). Blood flow (referred to as arbitrary "perfusion units") was quantified by multiplying the pixel number by the mean blood flow signal after subtraction of pixels below a fixed threshold of 300 perfusion units.
Cell preparation and flow cytometry. After lysing red blood cells with ACK buffer (Lonza, 10-548E), leukocytes were isolated from bloodby an initial wash step in PBS with centrifugation at 400×g for 5 min, followed by resuspension in staining buffer (PBS supplemented with 0.5% fetal bovine serum [FBS, Thermo Fisher Scientific, 26140079] and 0.4% 2 mM EDTA [ThermoFisher Scientific, 15575020]). Leukocytes were isolated from blister fluid by centrifugation at 400×g for 5 min, removal of the supernatant for cytokine analysis, then re-suspendion in ACK lysis buffer for 1 min, followed by a PBS washing step, centrifugation at 400×g and resuspension in 100 μl of staining buffer. Cell counts were obtained using a haemocytometer and Trypan blue to exclude dead cells. For flow cytometry, 10 6  Enzyme-linked immunosorbent assay (ELISA). Cytokine concentrations were measured in the blister fluid using TNFα (DY210), IL-6 (DY206), IL-1β (DY201), IL-10 (DY217B), IL-17 (DY317), IL-22 (DY782), IL-8 (DY208) kits from R&D Systems as per the manufacturer's instructions and streptavidin-HRP and tetramethylbenzidine (Invitrogen, 002023) detection as previously described 38 . Absorbance was read on a plate reader (Versamax, Sunnyvale, CA, USA) at 450 nm minus 540 nm. Concentration of cytokines in the samples were calculated by comparison against the standard curve.