Machine learning-driven identification of serotype-independent pneumococcal vaccine candidates using samples from human infection challenge studies

Machine learning-driven identification of serotype-independent pneumococcal vaccine candidates using samples from human infection challenge studies | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Machine learning-driven identification of serotype-independent pneumococcal vaccine candidates using samples from human infection challenge studies Katerina S. Cheliotis, Patricia Gonzalez-Dias, Esther L. German, and 18 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7566878/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Identifying conserved, immunogenic proteins that confer protection against Streptococcus pneumoniae colonisation could enable development of serotype-independent vaccines. We analysed baseline samples from 86 healthy adults experimentally challenged with pneumococcal serotypes 6B or 15B to investigate whether immune responses to 75 universally expressed pneumococcal proteins associated with protection against colonisation. We measured serum IgG using a novel Luminex assay and cytokine responses from peripheral blood mononuclear cells following antigen stimulation. No individual IgG or cytokine marker correlated significantly with protection in univariate analysis. However, machine learning identified IgG responses to PdB, SP1069, and SP0899 as predictive of protection. MCP-1 responses to SP1069 and SP0899, and IL-17 production in response to SP0648-3 also correlated with protection. Elevated baseline IFN-γ, RANTES, and anti-protein IgG correlated with lower colonisation density. We highlight SP1069 and SP0899 as potential serotype-independent vaccine candidates and demonstrate the utility of machine learning to identify immune correlates of protection. Health sciences/Diseases Biological sciences/Immunology Biological sciences/Microbiology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Pneumonia is the leading infectious cause of death in infants worldwide, with more than 800,000 annual deaths among children under 5 years of age globally 1 – 3 . The most common aetiological agent of pneumonia is the bacterium Streptococcus pneumoniae. Nasopharyngeal colonisation with pneumococcus, as well as being a prerequisite for invasive disease, is a source of transmission in the community 4 . The pneumococcal conjugate vaccine (PCV) generates population herd protection providing substantial cost-benefit of vaccination 5 . Vaccine-induced humoral responses to the pneumococcal capsule polysaccharide protect against pneumococcal carriage acquisition and disease caused by vaccine specific serotypes 6 – 10 . However, current vaccines – pneumococcal polysaccharide vaccine (PPV23) and PCVs – offer protection against a limited number of the > 100 pneumococcal serotypes 11 leading to serotype replacement of non-vaccine types 12 , 13 . In addition, despite being included in PCVs, serotype 3 continues to circulate and cause a high disease burden particularly in older adults 14 – 16 . PCVs are also complex and costly to produce, which can present a barrier to vaccine access in resource-limited countries, particularly those graduating from GAVI support 17 . Pneumococcal protein antigens that are universally expressed across various serotypes may induce both humoral and cellular immune responses, offering broad protection. These antigens could potentially be introduced as stand-alone vaccines or be included as conjugate proteins in current polysaccharide-based vaccines 18 . Currently the regulatory approval pathway for next-generation pneumococcal vaccines in infants relies on the non-inferiority of IgG levels to capsular polysaccharide using a pre-defined threshold of 0.35ug/ml. In adults demonstration of non-inferior or superior opsonophagocytic antibody responses to each serotype compared to existing licensed vaccines is required 19 . Therefore, the regulatory pathway for capsular polysaccharide independent vaccines is not defined. The identification of serotype-independent cellular and humoral correlates of protection against pneumococcal infection could assist with this framework and help to accelerate development of serotype-independent vaccines. Systems-biology and artificial intelligence can be used to significantly enhance our understanding of immunity and aid in the identification of correlates of protection. Embedding the understanding of individual components and their interactions, beyond canonical antibody binding and neutralisation, may accelerate the design of more successful and long-lasting vaccines 20 . In this study, we leveraged a controlled human infection model (CHIM) to investigate immune responses to pneumococcal proteins that are conserved across serotypes, aiming to identify correlates of serotype-independent protection and to understand their impact on colonisation dynamics in the nasopharynx. To enable high-throughput assessment of protein-specific IgG responses, we established a Luminex assay targeting 75 pneumococcal protein antigens. Given the well-established role of IL-17 in mediating protection against S. pneumoniae in murine models, we also evaluated IL-17-driven cellular responses in protein-stimulated peripheral blood mononuclear cells (PBMCs). In parallel, we employed a multiplex cytokine assay to profile broader cytokine responses, with MCP-1 (CCL2), RANTES (CCL5), and IFN-γ selected for more detailed analysis. Methods Study Design The overall study design is shown in Figure 1. Briefly, at baseline (day -4 and day -5), participants were assessed before intranasal inoculation (day 0) with S. pneumoniae (Spn). Carriage acquisition was monitored until day 14 for Spn15B and day 29 for Spn6B. Immune responses were evaluated using Luminex assays, which measured IgG antibody responses to 75 pneumococcal antigens in 72 volunteers, and cytokine profiling, which detected 21 immune markers (including IL-17, IFNγ, RANTES, and MCP-1) in 63 volunteers. Carriage acquisition results are presented as bar graphs, depicting the number of individuals testing positive or negative for Spn6B and Spn15B at different time points. These results indicate the proportion of participants who acquired pneumococcal carriage following inoculation. Participant cohort inoculated with serotype 6B pneumococcus Blood samples were taken from healthy non-smoking participants between the ages of 18 and 50 who were intranasally inoculated with 80,000 CFU in 0.1ml solution of S. pneumoniae serotype 6B (Spn6B) strain BHN418 per naris (full sequence available GenBank: ASHP00000000). Samples were taken prior to inoculation (baseline). Cohorts were recruited as part EudraCT (2014-004634-26), a study to evaluate the effect of live attenuated influenza vaccine (LAIV) on experimental human pneumococcal colonisation (EHPC) 21,22 21,22 . All participants gave written informed consent, with approval from the North West NHS Research Ethics Committee (14/NW/1460). Nasal wash samples were taken on days 2, 6, 9, 14 and 29 post-challenge with Spn6B. Colonisation was determined by classical microbiology and the presence of Spn6B in nasal wash samples taken at pre-defined time-points between inoculation and 29 days post-inoculation. No participants had previously received any pneumococcal vaccine. Participant cohort inoculated with serotype 15B pneumococcus Healthy, non-smoking participants aged 18-50 years were intranasally inoculated with 80,000 CFU S. pneumoniae serotype 15B (Spn15B) per naris (clinical isolate 15B P1262. European Nucleotide Archive accession number: ERS2632437). Spn15B is included in the PPV23 formulation but not in the 13-valent PCV formulation. All participants gave informed written consent, ethical approval was obtained from the National Health Service Research Ethics Committee, Liverpool East (15/NW/0931). Serum samples were taken at baseline (pre-inoculation) (Trial identifier: ISRCTN 68323432; 20815). Colonisation was determined by classical microbiology and the presence of Spn15B in nasal wash samples taken at pre-defined time-points between inoculation and 14 days post-inoculation. Nasal wash samples were taken on days 2, 7 and 14 post-challenge with Spn15B. No participants had previously received any pneumococcal vaccine. Luminex assay The proteins used in the assay are given in Supplementary Table 1. The sequences encoding the alpha-helical N-terminal domain of PspA1, PspC6, PspC9, were cloned using DNA from strain BHN418 (serotype 6B pneumococcus). The sequence encoding the N-terminal domain of PspA4 was cloned from strain 255/00 (serotype 14, GenBank accession number EF649969). Recombinant proteins were produced at Butantan Institute laboratories as described previously 23 including pneumolysin toxoid B (PdB), a pneumolysin mutant with a tryptophan-to-phenylalanine substitution at position 433 and 0.1% haemolytic activity of wild-type pneumolysin 24 25 . The remaining 70 proteins in the library were produced at Boston Children’s Hospital as described 26 . All 75 pneumococcal protein antigens were conjugated to magnetic microspheres at a concentration of 5µg per million microspheres using the xMAP Antibody Coupling Kit as per the manufacturer’s instructions. The assay was validated for the detection of anti-protein IgG as described previously 27,28 . Full methods are described in detail in Supplementary Text 1. A stock of pooled sera taken 29 days post-challenge and diluted 1:50 in assay buffer was used as an internal positive control across all plates and assay buffer was used as a negative control to detect background median fluorescence intensity (MFI) values. Luminex data cleaning and statistical analysis Raw Luminex data were cleaned prior to analysis through the following workflow. The coefficient of variation (CV) of replicates was calculated for all analytes. If any analyte had greater than 25% CV and both replicates had greater than 50 microspheres per well, the analyte was removed from analysis. If any paired samples had less than 50 microspheres of a given analyte in each well, the analyte was removed from analysis. The determinant of 50 microspheres is based on recommendations by the Luminex Corporation that a recovery of less than 50 microspheres renders a result unreliable as well as previous work showing that lower bead counts elicit higher CV values 29 . All statistical analysis was carried out in RStudio (version 1.0.153). Statistical significance was determined via a Mann-Whitney or Wilcoxon test for unpaired and paired groups, respectively, followed by Benjamini Hochberg correction for multiple comparisons. For the machine learning analysis, any proteins with greater than 25% of values missing and any participants with more than 50% of values missing were removed from the analysis. Stimulation of human peripheral blood mononuclear cells (PBMCs) and cytokine/chemokine detection Baseline PBMCs from healthy participants inoculated with Spn6B were thawed with 50 µg/ml DNAse I (MilliporeSigma) in prewarmed RPMI containing 10% FBS and washed once in media including DNAse I. 1x10 6 PBMCs/ml from an initial cohort of 12 carriage-negative and 11 carriage-positive participants were stimulated with 8µg/ml of each protein from the complete library and 5µg/ml of heat-inactivated Spn6B for 7 days at 37°C under 5% CO 2 . The plates were then centrifuged to pellet the cells, after which the supernatants were collected to measure IL-17A concentration using human ELISA kits (Invitrogen). In parallel, supernatants from carriage-negative and carriage-positive participants from this initial cohort were combined and assessed for other cytokine and chemokine concentrations using a 30-plex magnetic human Luminex cytokine kit (Thermo Fisher Scientific) and acquired on an LX200. A log 2-fold change between carriage-negative and positive was calculated and a Wilcoxon paired test was used to identify cytokines/chemokines associated with a specific carriage status. Twenty-one antigens from a 70-protein antigen library were selected for further analysis. The antigens were selected based on (i) their ability to induce high IL-17A responses in carriage-negative individuals (ii) negative correlation between high IL-17A levels and low pneumococcal density in participants who were experimentally colonised (iii) their ability to induce high expression of cytokines/chemokines at baseline prior to challenge (iv) available published data in which the antigens have shown protection against challenge after immunisation in pre-clinical models of infection. PBMCs from an additional 22 carriage-negative and 21 carriage-positive participants were stimulated with the selected 21 antigens and cytokine/chemokine concentrations were measured to validate the results obtained with the initial cohort. Human ELISA kits were used to measure concentrations of IL-17A (Invitrogen), MCP-1 (CCL-2) (BD OptEIA TM ), IFN-γ (BD OptEIA TM ), and RANTES (CCL-5) (Biolegend) in the collected supernatants following the manufacturers recommendations. Non-stimulated (control) and protein-stimulated supernatants were diluted as follows: 1:2 for IL-17, 1:10 for RANTES, 1:200 for IFN-γ, and 1:3200 for MCP-1. ELISA data cleaning and statistical analysis As detailed above, any proteins with greater than 25% of values missing and any participants with more than 50% of values missing were removed from the analysis. Statistical significance was determined via a Wilcoxon test for unpaired groups, followed by Benjamini Hochberg correction for multiple comparisons. Area under the curve (AUC) calculation Area under the curve (AUC) was used to calculate overall colonising density of S. pneumoniae in the nasopharynx of experimentally colonised participants. AUC was calculated by trapezoidal rule from log 10 transformed density results at 2, 6, 16, 22, 27 and 36 days or after challenge on day 2, 7 or 14 with interpolation of missing data assuming a linear change in density between known values. The AUC of pneumococcal colonising density at each timepoint was derived and summarised using the above descriptive statistics and analysed using a generalised linear model with a single factor of treatment. AUC correlation analysis A Spearman test was performed to evaluate the correlation between the IgG and cytokine/chemokine levels elicited in response to each protein and the AUC values obtained between days 2 to 14 for each carriage-positive participant using the cor.test function in the software R. The analysis was performed using antibody response and cellular response from 34 and 27 participants, respectively. Correlations with p -value ≤ 0.05 were considered significant. AUC threshold definition Retrospectively, we calculated the AUC of nasopharyngeal Spn6B density between days 2 to 14 post-challenge in 128 experimentally colonised participants across our CHIM studies to define “Low”, “Mid” or “High” colonising density. For participants experimentally colonised with Spn15B, threshold AUC values were set using colonising density between days 2 to 14 from 9 participants. Threshold values differed for the two serotypes. The data distribution was used to split the groups for both Spn6B and Spn15B inoculated cohorts. The participants from different AUC groups were localised in different quartiles (Supplementary Figure 1). Machine learning analysis Missing values in numerical variables were imputed using the "mice" R package 30 . Numeric features were standardized using z-score normalization. The Random Forest algorithm, an ensemble machine learning method that constructs multiple decision trees using random feature subsets and aggregates their outputs to enhance predictive accuracy and minimize overfitting (28), implemented in the R package "randomForest". The model was trained with mtry = 1 and ntree = 500, while other hyperparameters were set to their default values. This algorithm was employed to identify baseline protein-specific antibodies and cytokines/chemokines produced in response to PBMC stimulation with protein antigens that best predicted carriage status post-challenge. Unlike univariate analysis, which assesses each feature independently, machine learning techniques such as Random Forest can uncover complex, multivariate relationships among features, improving predictive accuracy. Feature importance was assessed using the mean decrease in Gini index, a metric that quantifies how much a given feature contributes to reducing impurity (i.e., improving class separation) within the decision trees. A higher mean decrease in Gini index indicates a stronger contribution to classification accuracy. To identify the most relevant features and optimize model performance, we applied an iterative feature selection process based on feature importance. Initially, we trained the Random Forest model using all 75 available features and evaluated their importance based on the mean decrease in Gini index. We then progressively removed the least informative features, retraining the model at each iteration and assessing its predictive performance. This process was repeated several times, testing progressively smaller feature sets (e.g., 50, 40, etc.), until the optimal configuration was identified, the smallest number of features yielding the best model performance, minimizing error, and reducing the risk of overfitting. Model evaluation was conducted using 10-fold cross-validation. Results Serum IgG levels to 75 different protein antigens at baseline did not associate with protection against experimental pneumococcal carriage acquisition of Spn6B or Spn15B A Luminex-based multiplex assay using a library of 75 highly conserved pneumococcal proteins (Supplementary Table 1), was successfully optimised and validated 28 . No associations between baseline levels of serum IgG against any of the protein antigens and protection against carriage acquisition remained statistically significant after applying the Benjamini-Hochberg correction for multiple testing (Figure 2). Screening of antigens using human PBMCs and cellular responses A library of pneumococcal protein antigens was used to stimulate PBMCs collected from participants prior to experimental challenge with Spn6B. Due to the limited number of PBMCs obtained from each participant, we were not able to screen all protein antigens in all participants. IL-17A production has previously been described as a marker of protection against pneumococcal colonisation in mice 31,32 . IL-17A secretion following ex-vivo protein stimulation was observed for all proteins at 7 days of culture. To identify other cellular markers that could associate with a protective or susceptible profile in the CHIM, we pooled supernatants from PBMC cultures following individual protein stimulation from carriage-negative and carriage-positive participants and compared levels produced of 30 cytokines and chemokines using 30-plex Luminex in these two groups (Supplementary Figure 2). MCP-1 (CCL-2) was selected as a marker of protection based on its well-documented role in monocyte recruitment and the clearance of pneumococcal carriage 21 . RANTES (CCL5) and IFN-γ were selected due to their elevated levels in carriage-positive individuals and their known involvement in T cell-mediated immune responses 33 . We used the data obtained from the initial IL-17A screening of susceptible and protected profiles (Supplementary Figure 3), together with available data on immunisation with those antigens and protection against carriage in mice (24), to down-select a limited number of antigens (Supplementary Table 2) for PBMC stimulation assays. PBMCs from a further 22 carriage-negative and 21 Spn6B carriage-positive participants were stimulated with the down-selected 21 antigens and IL-17A, RANTES, MCP-1 and IFN-γ responses were measured. Baseline cytokine/chemokine levels in response to protein antigens are not associate with protection against experimental pneumococcal carriage acquisition with Spn6B After correction for multiple comparisons, there was no significant difference between baseline protein-specific cytokine/chemokine responses in participants susceptible to experimental colonisation with Spn6B and those who were protected against colonisation (Figure 3). Responses with p-values ≤ 0.05 for the carriage-positive and carriage-negative groups comparison prior to Benjamini-Hochberg correction are shown (Figure 3). Random Forest algorithm as an alternative method to rank antigens associated with pneumococcal carriage protection We used the Random Forest method 34 to assess the predictive power of baseline protein-specific antibody (IgG) and cytokine/chemokine levels in predicting susceptibility/protection to carriage acquisition. We trained the model using various features at baseline and selected the combination of features that generated the model with the least error. Antibody response patterns to the proteins PdB, SP0899, and SP1069 together emerged as the most important features, with the model achieving an out-of-bag (OOB) error rate of 36.05%. In terms of cytokine/chemokine responses, IL-17A production in response to SP0648-3 and MCP-1 production in response to SP1069 and SP0899 together were identified as the most relevant features, resulting in an OOB error of 38.98%. Notably, SP0899 and SP1069 were a feature of both humoral and cellular responses that were predictive of protection against colonisation (Figure 4). Higher baseline IgG against SP1069 and SP0899 was observed in the cohort of participants protected against experimental carriage acquisition (Figure 4). We also observed that those participants protected against colonisation had higher levels of MCP-1 produced in response to stimulation with SP1069 ( p = 0.035) and SP0899 ( p = 0.049) than those susceptible to colonisation at baseline (Figure 4). Correlation of antibody and cytokine/chemokine levels against the proteins selected by machine learning analysis To determine whether there was a relationship between antibody- and cell-mediated immune responses, we correlated responses identified as predictive of colonisation status by machine learning. Whilst SP0648-3 was found to be correlated in carriage positive participants across datasets ( Rho = 0.614; p = 0.009), SP1069 ( Rho = 0.378; p = 0.203) and SP0899 ( Rho = -0.242; p = 0.425) were not. None of the proteins were correlated in carriage negative participants across datasets (SP0648-3 ( Rho = -0.076; p = 0.781), SP1069 ( Rho = -0.422; p = 0.133), SP0899 ( Rho = -0.098; p = 0.781)). Correlation between protein-responses and AUC We generated area under the curve data for colonising density of the participants colonised with pneumococcus (Figure 5). No significant correlations were found following Benjamini Hochberg correction. Correlations showing p -value ≤ 0.05 prior to correction are highlighted in Figure 5 and Supplementary Tables 3 and 4. Participants were stratified by colonisation density and duration, expressed as AUC (Supplementary Figure 4). With respect to cell-mediated immunity, the levels of RANTES, MCP-1, and IFN-γ produced in response to protein stimuli (SP1154-2, SP0648-3, SP0079, and SP1683) differed across density groups (Supplementary Figure 4 B). Although differences were not statistically significant, we observed a trend in high levels of antibodies and cytokines/chemokines specific to protein antigens that were associated with lower colonisation density (Figure 6). We observed that IgG specific to 32 proteins (SP0321, SP0564, SP0346, SP1545, SP1032, SP0617, SP0620, SP0678, SP1479, SP1492, SP0079, SP0402, SP0629, SP0757, SP2145, SP0662-1, SP1683, SP0878, SP0582, SP1376, SP0724, SP1500, SP2083, SP0098, SP0127, SP0604, SP0648-1, SP2207, SP0742, SP0314, SP0648-2, PspA1) was elevated in participants who were either protected against experimental carriage acquisition, or in those who were experimentally colonised at a relatively low bacterial density (Figure 6A). We also observed that cytokine and chemokine responses were elevated in participants who were either protected against experimental carriage acquisition or in the low AUC density group, followed by the mid AUC density and high AUC density groups – IFN-γ and RANTES responses were particularly elevated (Figure 6B). Cytokine and chemokine levels in carriage-negative participants were generally not higher than those observed in the low AUC group. From the selected proteins, only IFN-γ and RANTES responses to SP1154-2 (IFN-γ p = 0.034; RANTES p = 0.024) and SP0079 (IFN-γ p = 0.012; RANTES p = 0.005) and the RANTES response to SP0648-1 (p = 0.027) have shown a p ≤ 0.05 before Benjamini-Hochberg correction (Figure 6B). As previously stated, SP1069 and SP0899 were selected by machine learning for carriage status classification. The IFN-γ response to SP1069 followed the same pattern, levels were highest in the carriage-negative group or in participants colonised at low bacterial density, and lowest in those with high-density colonisation. There was a very marginal difference in the levels of MCP-1 produced in response to SP0899 across groups. Discussion Machine learning and systems biology are increasingly employed in vaccine development to identify and design vaccine candidates as well as to identify correlates of successful immunisation, predict individual vaccine responses, and correlates of protection when a disease/infection endpoint is possible 35,36 . As the number of measurable biological parameters expands, traditional univariate analyses are inappropriate, increasing the risk of type II errors (failing to detect true associations). In contrast, machine learning enables a more comprehensive and rigorous analysis of complex, high-dimensional datasets, offering deeper insights into the multifactorial nature of immune responses. In this study, we leveraged machine learning to identify immune features predictive of protection against nasal colonisation. Our findings highlight the Random Forest algorithm’s utility in identifying proteins linked to protection against experimental pneumococcal carriage, demonstrating its potential to accelerate vaccine development. The ultimate aim of pneumococcal vaccination may not be sterilising immunity and complete elimination of nasopharyngeal colonisation, a natural immunising event 33,36,37,52 , but rather reducing colonisation density to prevent disease progression or transmission to others in the population, particularly during viral co-infections 53 . Suppressing bacterial density through protein-based vaccines could minimise transmission and reinforce immunity while maintaining mucosal exposure necessary for lung immunity 54 . Consistent with prior studies 36–38 , systemic anti-protein IgG at baseline did not correlate with protection against experimental colonisation by serotype 6B or 15B pneumococcus. Similarly, baseline cell-mediated responses in blood were not predictive of protection in our human cohort, despite evidence of Th17-mediated immunity against pneumococcal colonisation in murine models 26 . Using machine learning, we identified a group of three antigens — SP1069, SP0899, and PdB — eliciting IgG responses, and a group three antigens — SP1069, SP0899, and SP0648-3 — which stimulated protein-specific cellular responses, that were predictive of protection. Although we cannot yet determine whether these protein combinations could form the basis of a combination vaccine, this possibility warrants further investigation. Notably, SP1069 and SP0899 elicited both humoral and cellular responses. Although the exact functions of these proteins are not yet fully understood, immunisation of mice with a combination of cholera toxin alongside SP1069, SP0899 and SP0648-3 as individual antigens has led to a significant reduction in pneumococcal colonisation compared with the control groups 26 . Co-administration of multiple antigens with a cholera toxin adjuvant has been shown to enhance the protective efficacy in murine models 26 . Protein-based vaccines incorporating several antigens may thus offer broad coverage, while conjugating protein antigens to capsular polysaccharides in new PCV formulations could provide additional serotype-independent protection 39,40 as well as the advantage of cellular immunity induction. In humans, protection against pneumococcal disease requires both humoral and cellular immunity 31,41–46 . Similarly, murine models indicate the necessity of both immunity types for nasopharyngeal colonisation defence 26,41–43,47,48 . Using our human challenge model, we previously reported that colonisation-induced MCP-1 (CCL2) levels are associated with monocyte recruitment to the nasal mucosa and subsequent clearance of colonisation 21 . These results also support previous findings from mouse models showing that CCL2 signalling and monocyte recruitment are key mediators of pneumococcal carriage clearance 49 . Elevated MCP-1 levels in protected participants against experimental acquisition of Spn6B and its involvement in monocyte and neutrophil recruitment underline its significance in pneumococcal clearance 21,50,51 . In our cohort, baseline anti-protein IgG and cytokine responses correlated with lower colonisation density, suggesting roles for both humoral and cellular immunity in controlling carriage and preventing disease. Consistent with prior findings 55–58 , IFN-γ and RANTES secretion correlated with reduced colonisation density, likely enhancing innate immune cell recruitment and activity 59–61 . These results emphasise the importance of designing protein vaccines that elicit both antibody and cellular responses for effective protection. While promising, this work has limitations. Out-of-bag error rates (~36–39%) indicate only moderate predictive power, suggesting that further optimisation, such as increasing cohort size or incorporating mucosal immune parameters, may be needed to improve model accuracy. Immune responses to protein antigens in systemic circulation may differ from those in the nasal mucosa, and the study’s small sample size, comprised mainly of healthy young adults, may not reflect high-risk demographics. Validating these findings in samples from older adults participants will be important 37,62 . Immune correlates may vary across populations with differing pneumococcal burden. Our studies in Malawi 10,63 will enable validation across diverse populations and environmental contexts. Further investigation into antigen-specific nasal T-cell responses and sex-stratified immune differences could provide additional insights 27 . As immunology and vaccinology increasingly rely on complex, high-dimensional datasets, predictive computational models offer substantial advantages over traditional analyses by uncovering intricate relationships between immune responses and protection. The integration of advanced computational approaches with human challenge models, which uniquely allow sampling immediately before and after defined pathogen exposure, promises to transform how we define protective immunity and accelerate rational vaccine design. Despite the remarkable success of vaccines, which are estimated to have saved more than 150 million lives over the past 50 years 64 , major challenges persist. Unequal access to vaccines and the biological complexity of difficult-to-target pathogens threaten to stall progress in reducing the global burden of infectious diseases. A key scientific challenge is the identification of causal correlates of protection for high-frequency colonising pathogens such as Streptococcus pneumoniae , Staphylococcus aureus , and Escherichia coli , where repeated antigen exposures at mucosal surfaces shape layered humoral and cellular immune responses. Addressing these gaps will require integrating advanced computational approaches, experimental human models, and global public health strategies to translate immunological insights into next-generation vaccines. Declarations Data availability: Data are available upon reasonable request by email directed to the corresponding authors at [email protected] and [email protected] Author Contributions Conceived the study - DMF, SJ, CS, BU, RM Data acquisition - KSC, CS, RT, JR, ELG, EM, YL, SP, ENM Generated the ELISA data - CS Performed the machine learning analysis - PGD Performed data analysis - KSC, PGD, CS Scientific discussion – AJP, SBG, HIN, SHP, ANAG, AMC, SPJ Project definition and discussion - KSC, PGD, CS, DMF, BU Provided protein library - RM, YJL, EM, RT PGD, KSC and CS jointly interpreted the results and wrote the manuscript with contributions from all co-authors. All authors read and approved the final manuscript. Declaration of Interests RM is a consultant to GSK and Merck, is a named inventor and patent holder on vaccine technologies, a member of the board of directors at Corner Therapeutics and the scientific advisory boards of Amplitude Therapeutics, Limmatech and Vitrivax. Y-JL is a consultant to GSK and is a named inventor and patent holder on vaccine technologies. AJP is Chair of the UK Department of Health and Social Care’s Joint Committee on Vaccines and Immunisation and was previously chair and is now a member of WHOs technical advisory group on Salmonella vaccines. AJP is a contributor to intellectual property on a COVID19 vaccine licensed by Oxford University Innovation to AstraZeneca. Oxford University has received funding for research on Salmonella vaccines the Serum Institute of India. Acknowledgments The authors thank all participants in the EHPC studies, as well as the Clinical Research Network North West Coast and the NIHR Comprehensive Local Research Network North West Coast for their continued support. We also extend our gratitude to the study clinical team members for their assistance in sample collection. This work by funded by grants from the Bill & Melinda Gates Foundation (OPP1117728) (to DMF), Medical Research Council (MR/M011569/1) (to SBG and DMF), Robert Austrian Award (to SPJ) and the Bacterial Vaccines (BactiVac) Network funded by the GCRF Networks in Vaccines Research and Development which was co-funded by the MRC and BBSRC (to DMF and CS). KSC was supported by the MRC Doctoral Training Programme (Award 1964515). References GBD 2019 Antimicrobial Resistance Collaborators. Global mortality associated with 33 bacterial pathogens in 2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet 400 , 2221–2248 (2022). Wahl, B. et al. Burden of Streptococcus pneumoniae and Haemophilus influenzae type b disease in children in the era of conjugate vaccines: global, regional, and national estimates for 2000-15. Lancet Glob. Health 6 , e744–e757 (2018). McAllister, D. A. et al. Global, regional, and national estimates of pneumonia morbidity and mortality in children younger than 5 years between 2000 and 2015: a systematic analysis. Lancet Glob. Health 7 , e47–e57 (2019). Weiser, J. N., Ferreira, D. M. & Paton, J. C. Streptococcus pneumoniae: transmission, colonization and invasion. Nat. Rev. Microbiol. 16 , 355–367 (2018). Ladhani, S. N., Andrews, N. & Ramsay, M. E. Summary of evidence to reduce the two-dose infant priming schedule to a single dose of the 13-valent pneumococcal conjugate vaccine in the national immunisation programme in the UK. Lancet Infect. Dis. 21 , e93–e102 (2021). Black, S. et al. Efficacy, safety and immunogenicity of heptavalent pneumococcal conjugate vaccine in children. Northern California Kaiser Permanente Vaccine Study Center Group. Pediatr. Infect. Dis. J. 19 , 187–195 (2000). Ota, M. O. C. et al. Pneumococcal antibody concentrations of subjects in communities fully or partially vaccinated with a seven-valent pneumococcal conjugate vaccine. PLoS ONE 7 , e42997 (2012). Collins, A. M. et al. First human challenge testing of a pneumococcal vaccine. Double-blind randomized controlled trial. Am. J. Respir. Crit. Care Med. 192 , 853–858 (2015). Nagel, J. et al. The association between antibody levels before and after 7-valent pneumococcal conjugate vaccine immunization and subsequent pneumococcal infection in chronic arthritis patients. Arthritis Res. Ther. 17 , 124 (2015). Bar-Zeev, N. et al. Impact and effectiveness of 13-valent pneumococcal conjugate vaccine on population incidence of vaccine and non-vaccine serotype invasive pneumococcal disease in Blantyre, Malawi, 2006-18: prospective observational time-series and case-control studies. Lancet Glob. Health 9 , e989–e998 (2021). Singh, A. & Dutta, A. K. Pneumococcal Vaccines - How Many Serotypes are Enough? Indian J. Pediatr. 85 , 47–52 (2018). Weinberger, D. M., Malley, R. & Lipsitch, M. Serotype replacement in disease after pneumococcal vaccination. Lancet 378 , 1962–1973 (2011). Olaya-Abril, A., Jiménez-Munguía, I., Gómez-Gascón, L., Obando, I. & Rodríguez-Ortega, M. J. Identification of potential new protein vaccine candidates through pan-surfomic analysis of pneumococcal clinical isolates from adults. PLoS ONE 8 , e70365 (2013). Tiley, K. S. et al. Nasopharyngeal Carriage of Pneumococcus in Children in England up to 10 Years After 13-Valent Pneumococcal Conjugate Vaccine Introduction: Persistence of Serotypes 3 and 19A and Emergence of 7C. J. Infect. Dis. 227 , 610–621 (2023). Pick, H. et al. Pneumococcal serotype trends, surveillance and risk factors in UK adult pneumonia, 2013-18. Thorax 75 , 38–49 (2020). Ladhani, S. N. et al. Rapid increase in non-vaccine serotypes causing invasive pneumococcal disease in England and Wales, 2000-17: a prospective national observational cohort study. Lancet Infect. Dis. 18 , 441–451 (2018). Chen, C. et al. Effect and cost-effectiveness of pneumococcal conjugate vaccination: a global modelling analysis. Lancet Glob. Health 7 , e58–e67 (2019). Lagousi, T., Basdeki, P., Routsias, J. & Spoulou, V. Novel Protein-Based Pneumococcal Vaccines: Assessing the Use of Distinct Protein Fragments Instead of Full-Length Proteins as Vaccine Antigens. Vaccines (Basel) 7 , (2019). Grant, L. R. et al. Persistence of IgG antibody following routine infant immunization with the 7-valent pneumococcal conjugate vaccine. Pediatr. Infect. Dis. J. 34 , e138-42 (2015). Britto, C. & Alter, G. The next frontier in vaccine design: blending immune correlates of protection into rational vaccine design. Curr. Opin. Immunol. 78 , 102234 (2022). Jochems, S. P. et al. Inflammation induced by influenza virus impairs human innate immune control of pneumococcus. Nat. Immunol. 19 , 1299–1308 (2018). Rylance, J. et al. Two randomized trials of the effect of live attenuated influenza vaccine on pneumococcal colonization. Am. J. Respir. Crit. Care Med. 199 , 1160–1163 (2019). Darrieux, M. et al. Fusion proteins containing family 1 and family 2 PspA fragments elicit protection against Streptococcus pneumoniae that correlates with antibody-mediated enhancement of complement deposition. Infect. Immun. 75 , 5930–5938 (2007). Paton, J. C. et al. Purification and immunogenicity of genetically obtained pneumolysin toxoids and their conjugation to Streptococcus pneumoniae type 19F polysaccharide. Infect. Immun. 59 , 2297–2304 (1991). Ferreira, D. M. et al. DNA vaccines expressing pneumococcal surface protein A (PspA) elicit protection levels comparable to recombinant protein. J. Med. Microbiol. 55 , 375–378 (2006). Lu, Y.-J. et al. Screening for Th17-Dependent Pneumococcal Vaccine Antigens: Comparison of Murine and Human Cellular Immune Responses. Infect. Immun. 86 , (2018). Cheliotis, K. S. et al. Influence of sex, season and environmental air quality on experimental human pneumococcal carriage acquisition: a retrospective cohort analysis. ERJ Open Research 8 , (2022). Identifying Novel Pneumococcal Pneumonia Vaccine Candidates - LSTM Online Archive. https://archive.lstmed.ac.uk/21171/. Bjornstal, O. et al. The impact of decreased bead count to determine concentrations of amyloid beta1-42, total-tau, and phosphorylated-tau181 in human cerebrospinal fluid using xMAP technology. J. Pharm. Sci. 100 , 4655–4663 (2011). Unable to find information for 800635. Lu, Y.-J. et al. Interleukin-17A mediates acquired immunity to pneumococcal colonization. PLoS Pathog. 4 , e1000159 (2008). Keech, C. A. et al. A Phase 1 Randomized, Placebo-controlled, Observer-blinded Trial to Evaluate the Safety and Immunogenicity of Inactivated Streptococcus pneumoniae Whole-cell Vaccine in Adults. Pediatr. Infect. Dis. J. 39 , 345–351 (2020). Mitsi, E. et al. Nasal Pneumococcal Density Is Associated with Microaspiration and Heightened Human Alveolar Macrophage Responsiveness to Bacterial Pathogens. Am. J. Respir. Crit. Care Med. 201 , 335–347 (2020). Breiman, L. Random Forests. Springer Science and Business Media LLC 45 , 5–32 (2001). Oberg, A. L., Kennedy, R. B., Li, P., Ovsyannikova, I. G. & Poland, G. A. Systems biology approaches to new vaccine development. Curr. Opin. Immunol. 23 , 436–443 (2011). Ferreira, D. M. et al. Controlled human infection and rechallenge with Streptococcus pneumoniae reveals the protective efficacy of carriage in healthy adults. Am. J. Respir. Crit. Care Med. 187 , 855–864 (2013). Adler, H. et al. Experimental human pneumococcal colonization in older adults is feasible and safe, not immunogenic. Am. J. Respir. Crit. Care Med. 203 , 604–613 (2021). Araujo, A. P. et al. Serum levels of anti-PspA and anti-PspC IgG decrease with age and do not correlate with susceptibility to experimental human pneumococcal colonization. PLoS ONE 16 , e0247056 (2021). Chichili, G. R. et al. Phase 1/2 study of a novel 24-valent pneumococcal vaccine in healthy adults aged 18 to 64 years and in older adults aged 65 to 85 years. Vaccine 40 , 4190–4198 (2022). Kaplonek, P. et al. A semisynthetic glycoconjugate provides expanded cross-serotype protection against Streptococcus pneumoniae. Vaccine 40 , 1038–1046 (2022). McCool, T. L. & Weiser, J. N. Limited role of antibody in clearance of Streptococcus pneumoniae in a murine model of colonization. Infect. Immun. 72 , 5807–5813 (2004). Malley, R. et al. CD4+ T cells mediate antibody-independent acquired immunity to pneumococcal colonization. Proc Natl Acad Sci USA 102 , 4848–4853 (2005). van Rossum, A. M. C., Lysenko, E. S. & Weiser, J. N. Host and bacterial factors contributing to the clearance of colonization by Streptococcus pneumoniae in a murine model. Infect. Immun. 73 , 7718–7726 (2005). Cohen, J. M. et al. Protective contributions against invasive Streptococcus pneumoniae pneumonia of antibody and Th17-cell responses to nasopharyngeal colonisation. PLoS ONE 6 , e25558 (2011). Wilson, R. et al. Protection against Streptococcus pneumoniae lung infection after nasopharyngeal colonization requires both humoral and cellular immune responses. Mucosal Immunol. 8 , 627–639 (2015). Wilson, R. et al. Naturally acquired human immunity to pneumococcus is dependent on antibody to protein antigens. PLoS Pathog. 13 , e1006137 (2017). Trzciński, K. et al. Effect of serotype on pneumococcal competition in a mouse colonization model. MBio 6 , e00902-15 (2015). Moffitt, K. L., Malley, R. & Lu, Y.-J. Identification of protective pneumococcal T(H)17 antigens from the soluble fraction of a killed whole cell vaccine. PLoS ONE 7 , e43445 (2012). Zhang, Z., Clarke, T. B. & Weiser, J. N. Cellular effectors mediating Th17-dependent clearance of pneumococcal colonization in mice. J. Clin. Invest. 119 , 1899–1909 (2009). Nikolaou, E. et al. Experimental Human Challenge Defines Distinct Pneumococcal Kinetic Profiles and Mucosal Responses between Colonized and Non-Colonized Adults. MBio 12 , (2021). Carr, M. W., Roth, S. J., Luther, E., Rose, S. S. & Springer, T. A. Monocyte chemoattractant protein 1 acts as a T-lymphocyte chemoattractant. Proc Natl Acad Sci USA 91 , 3652–3656 (1994). German, E. L. et al. Protective effect of PCV vaccine against experimental pneumococcal challenge in adults is primarily mediated by controlling colonisation density. Vaccine 37 , 3953–3956 (2019). Carr, O. J. J. et al. Nasopharyngeal pneumococcal colonization density is associated with severe pneumonia in young children in the lao people’s democratic republic. J. Infect. Dis. 225 , 1266–1273 (2022). Khan, M. N., Xu, Q. & Pichichero, M. E. Protection against Streptococcus pneumoniae Invasive Pathogenesis by a Protein-Based Vaccine Is Achieved by Suppression of Nasopharyngeal Bacterial Density during Influenza A Virus Coinfection. Infect. Immun. 85 , (2017). Trzciński, K. et al. Protection against nasopharyngeal colonization by Streptococcus pneumoniae is mediated by antigen-specific CD4+ T cells. Infect. Immun. 76 , 2678–2684 (2008). Hoe, E. et al. Reduced IL-17A Secretion Is Associated with High Levels of Pneumococcal Nasopharyngeal Carriage in Fijian Children. PLoS ONE 10 , e0129199 (2015). Vuononvirta, J., Peltola, V., Ilonen, J., Mertsola, J. & He, Q. The Gene Polymorphism of IL-17 G-152A is Associated with Increased Colonization of Streptococcus pneumoniae in Young Finnish Children. Pediatr. Infect. Dis. J. 34 , 928–932 (2015). Lu, Y.-J. et al. GMP-grade pneumococcal whole-cell vaccine injected subcutaneously protects mice from nasopharyngeal colonization and fatal aspiration-sepsis. Vaccine 28 , 7468–7475 (2010). Soudja, S. M. et al. Memory-T-cell-derived interferon-γ instructs potent innate cell activation for protective immunity. Immunity 40 , 974–988 (2014). Mancuso, R. I. et al. Impaired expression of CXCL5 and matrix metalloproteinases in the lungs of mice with high susceptibility to Streptococcus pneumoniae infection. Immun. Inflamm. Dis. 6 , 128–142 (2018). Shenoy, A. T. et al. Lung CD4+ resident memory T cells remodel epithelial responses to accelerate neutrophil recruitment during pneumonia. Mucosal Immunol. 13 , 334–343 (2020). Urban, B. C. et al. Inflammation of the nasal mucosa is associated with susceptibility to experimental pneumococcal challenge in older adults. Mucosal Immunol. 17 , 973–989 (2024). Morton, B. et al. A feasibility study of controlled human infection with Streptococcus pneumoniae in Malawi. EBioMedicine 72 , 103579 (2021). Shattock, A. J. et al. Contribution of vaccination to improved survival and health: modelling 50 years of the Expanded Programme on Immunization. Lancet 403 , 2307–2316 (2024). Additional Declarations Competing interest reported. RM is a consultant to GSK and Merck, is a named inventor and patent holder on vaccine technologies, a member of the board of directors at Corner Therapeutics and the scientific advisory boards of Amplitude Therapeutics, Limmatech and Vitrivax. Y-JL is a consultant to GSK and is a named inventor and patent holder on vaccine technologies. AJP is Chair of the UK Department of Health and Social Care’s Joint Committee on Vaccines and Immunisation and was previously chair and is now a member of WHOs technical advisory group on Salmonella vaccines. AJP is a contributor to intellectual property on a COVID19 vaccine licensed by Oxford University Innovation to AstraZeneca. Oxford University has received funding for research on Salmonella vaccines the Serum Institute of India. Supplementary Files SupplementaryMaterialProteinsPaper20Aug2025FV.docx GraphicalAbstract.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7566878","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":527759683,"identity":"fe11531f-2145-4073-acd2-8b9a66122bcf","order_by":0,"name":"Katerina S. Cheliotis","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Katerina","middleName":"S.","lastName":"Cheliotis","suffix":""},{"id":527759684,"identity":"e69b2622-49af-4aba-8b98-965110fdb0b0","order_by":1,"name":"Patricia Gonzalez-Dias","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Patricia","middleName":"","lastName":"Gonzalez-Dias","suffix":""},{"id":527759685,"identity":"603a84c2-9ff5-4600-9847-2f3bbd8f7856","order_by":2,"name":"Esther L. German","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Esther","middleName":"L.","lastName":"German","suffix":""},{"id":527759686,"identity":"231dd9ae-9071-4e06-b00e-cc22ede388e4","order_by":3,"name":"André N.A. Gonçalves","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"André","middleName":"N.A.","lastName":"Gonçalves","suffix":""},{"id":527759687,"identity":"271d9d92-79bc-4f35-9637-a83304e50f6e","order_by":4,"name":"Elena Mitsi","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Elena","middleName":"","lastName":"Mitsi","suffix":""},{"id":527759688,"identity":"951eed9a-d09a-4ae6-8d74-86c8ef4d742a","order_by":5,"name":"Helder I. Nakaya","email":"","orcid":"","institution":"Universidade de São Paulo","correspondingAuthor":false,"prefix":"","firstName":"Helder","middleName":"I.","lastName":"Nakaya","suffix":""},{"id":527759689,"identity":"36592d8a-e4d9-4d76-94cf-021c9739ba1f","order_by":6,"name":"Elissavet Nikolaou","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Elissavet","middleName":"","lastName":"Nikolaou","suffix":""},{"id":527759690,"identity":"75d7ffae-2ce2-4681-9c99-5f62e2e8ba28","order_by":7,"name":"Shaun H. Pennington","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Shaun","middleName":"H.","lastName":"Pennington","suffix":""},{"id":527759691,"identity":"bc1d9d60-ef0f-4b14-a550-3ce37694c928","order_by":8,"name":"Sherin Pojar","email":"","orcid":"","institution":"University of Oxford","correspondingAuthor":false,"prefix":"","firstName":"Sherin","middleName":"","lastName":"Pojar","suffix":""},{"id":527759693,"identity":"dba94fb8-43f2-4e29-b6e6-be2135bbf64f","order_by":9,"name":"Jesús Reiné","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jesús","middleName":"","lastName":"Reiné","suffix":""},{"id":527759694,"identity":"efdd502c-78b4-4aee-8df1-c376e92fde73","order_by":10,"name":"Eliane N. Miyaji","email":"","orcid":"","institution":"Instituto Butantan","correspondingAuthor":false,"prefix":"","firstName":"Eliane","middleName":"N.","lastName":"Miyaji","suffix":""},{"id":527759696,"identity":"d41e8f37-6234-4c68-b2aa-9ce625a3577d","order_by":11,"name":"Rafaella Tostes","email":"","orcid":"","institution":"Instituto Butantan","correspondingAuthor":false,"prefix":"","firstName":"Rafaella","middleName":"","lastName":"Tostes","suffix":""},{"id":527759697,"identity":"6c3151e9-b63e-4247-b4e1-05b084c3b200","order_by":12,"name":"Andrea M. Collins","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Andrea","middleName":"M.","lastName":"Collins","suffix":""},{"id":527759699,"identity":"15b471a5-f25f-405b-a90a-048848281215","order_by":13,"name":"Stephen B. Gordon","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Stephen","middleName":"B.","lastName":"Gordon","suffix":""},{"id":527759700,"identity":"51e2e737-d79f-4ea8-8f3c-deea457750a7","order_by":14,"name":"Andrew J. Pollard","email":"","orcid":"","institution":"University of Oxford","correspondingAuthor":false,"prefix":"","firstName":"Andrew","middleName":"J.","lastName":"Pollard","suffix":""},{"id":527759701,"identity":"ae268fc0-4679-499b-955e-91299faf9e00","order_by":15,"name":"Ying-Jie Lu","email":"","orcid":"","institution":"Boston Children's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Ying-Jie","middleName":"","lastName":"Lu","suffix":""},{"id":527759702,"identity":"440b9691-6136-4b80-8ea4-96d9736c5935","order_by":16,"name":"Richard Malley","email":"","orcid":"","institution":"Boston Children's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Richard","middleName":"","lastName":"Malley","suffix":""},{"id":527759703,"identity":"82bda9f1-e020-494e-9f86-6585bf69b2ee","order_by":17,"name":"Simon P. Jochems","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Simon","middleName":"P.","lastName":"Jochems","suffix":""},{"id":527759704,"identity":"18ab5f50-2d13-4ff5-a20c-09523335585c","order_by":18,"name":"Britta Urban","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Britta","middleName":"","lastName":"Urban","suffix":""},{"id":527759705,"identity":"1e7d8b52-66c0-4c9a-8108-8db60e989cd6","order_by":19,"name":"Carla Solórzano","email":"","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":false,"prefix":"","firstName":"Carla","middleName":"","lastName":"Solórzano","suffix":""},{"id":527759706,"identity":"c069092d-9b13-4797-b147-f9bad61c70ae","order_by":20,"name":"Daniela M. Ferreira","email":"data:image/png;base64,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","orcid":"","institution":"Liverpool School of Tropical Medicine","correspondingAuthor":true,"prefix":"","firstName":"Daniela","middleName":"M.","lastName":"Ferreira","suffix":""}],"badges":[],"createdAt":"2025-09-08 18:23:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7566878/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7566878/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93461158,"identity":"0432a33c-0c66-4dd1-86bb-688e665e7c2e","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3462609,"visible":true,"origin":"","legend":"","description":"","filename":"ProteinsPaper20Aug2025FV.docx","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/d855e4aa76cf076ede403f39.docx"},{"id":93462196,"identity":"cbc33e7e-6c63-4f76-9a73-9e1167e471ba","added_by":"auto","created_at":"2025-10-14 06:29:11","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21154,"visible":true,"origin":"","legend":"","description":"","filename":"49e16ed37fb7429bb0e60176bf38e828.json","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/8c7272fec3eebb96bd7a15c3.json"},{"id":93461152,"identity":"0b46c33d-ac8a-42ed-a2d2-26bbb23edfba","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2966550,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterialProteinsPaper20Aug2025FV.docx","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/ac0d3683ea4484e39a3261b2.docx"},{"id":93462375,"identity":"02d991f0-2d2e-48cd-a139-b205db1d29f5","added_by":"auto","created_at":"2025-10-14 06:37:11","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":149602,"visible":true,"origin":"","legend":"","description":"","filename":"49e16ed37fb7429bb0e60176bf38e8281enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/7651999760e4b5661556ae10.xml"},{"id":93461161,"identity":"fd1a2119-3f94-486d-8b1a-89e63e50c7e4","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":879457,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/231282aab8907c5f44f62c51.png"},{"id":93461148,"identity":"7c2fd50a-e255-4539-a168-5bfa4cb36adf","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":91720,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/3befa5be72946c52105cbff1.png"},{"id":93461150,"identity":"c31532f8-61e7-480a-9793-aa184090f336","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":121723,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/6e90d00dd08f1e6d773f7324.png"},{"id":93461165,"identity":"0feff238-0bb6-4601-b0a3-0ceb50b9f5ca","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":78688,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/5a876cda762a5ada7b06b927.png"},{"id":93461164,"identity":"e467de2d-8b22-429f-8e31-f705974eba86","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":65965,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/fc2bb789dbf957459b895057.png"},{"id":93461156,"identity":"287d5e15-d5a2-465b-ba7b-db317b5ddfef","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":110905,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/614aa36d61a58f476fce4122.png"},{"id":93462198,"identity":"cba328cb-d64b-4df9-90f3-389e2962b36b","added_by":"auto","created_at":"2025-10-14 06:29:11","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53636,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/b51fae19ac3f85e0c95440da.png"},{"id":93461157,"identity":"c4482762-e9b9-4297-8a1c-b6e0b630afc6","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":75377,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/e438c5574fa117644b62fac0.png"},{"id":93462201,"identity":"bcc39245-6427-407e-9694-9dfb0c266089","added_by":"auto","created_at":"2025-10-14 06:29:12","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":132416,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/faeda3a116592bf941220e3e.png"},{"id":93461162,"identity":"5521c2d1-efb5-491d-941f-194d8d494c9c","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17022,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/8f9571436be6bce45b61f320.png"},{"id":93461170,"identity":"0b1837e4-28e6-4d1f-a047-197d43af49cd","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":46047,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/3756d42980b7df48ef327510.png"},{"id":93462199,"identity":"d3e2dccc-40f5-437b-9551-5032a023a32e","added_by":"auto","created_at":"2025-10-14 06:29:11","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":26019,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/c053ef8137f6b0548a39f72d.png"},{"id":93461155,"identity":"2cfecd91-b6cb-4ef1-80ab-9b62b07f736d","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":22049,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/7ef1c63f440a2358accda96c.png"},{"id":93461169,"identity":"8354e785-9b5a-4578-bc1b-0f4002c10551","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":34418,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/3c6429e03e8779377f112a89.png"},{"id":93461166,"identity":"69545838-27d7-4b0a-8f56-a08f28b7b173","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17591,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/9c47f1adb43f936184bbff89.png"},{"id":93461159,"identity":"52dd0b43-f7c2-4c17-b780-b348787d107d","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28052,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/93fb5fb081abf31b4f800d12.png"},{"id":93461167,"identity":"e2e98ec7-b917-40d7-9854-0582744d89bc","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"xml","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":143429,"visible":true,"origin":"","legend":"","description":"","filename":"49e16ed37fb7429bb0e60176bf38e8281structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/09c02876c6c242467dd7a3da.xml"},{"id":93461172,"identity":"62fd9531-6237-4210-b0ef-1ea27e55b7d4","added_by":"auto","created_at":"2025-10-14 06:21:12","extension":"html","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":166198,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/5acb1961612aac7605fb08d2.html"},{"id":93462193,"identity":"a714f520-6336-4000-8ce3-1d08eeabbbe8","added_by":"auto","created_at":"2025-10-14 06:29:10","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":181631,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudy design. \u003c/strong\u003eParticipant serum was collected at ‘Baseline’ (4 days for Spn6B and 5 days for Spn15B) prior to experimental inoculation with pneumococcus (Spn) at day 0 (D0), and carriage acquisition assessed until day 29 post-challenge (days 2, 6, 9, 14 and 29 post-challenge with Spn6B; days 2, 7 and 14 post-challenge with Spn15B). Antibody responses were evaluated using the Luminex assay. Cytokine/chemokine responses (IL-17A, IFN-γ, RANTES (CCL-5) and MCP-1 (CCL-2)) following the stimulation of PBMCs with pneumococcal protein antigens were evaluated by ELISA. The number of antigens and the number of participants per carriage status are shown. Bar plots show the number of participants that were colonised with pneumococcus (orange) and those that were protected against carriage acquisition (blue).\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/b0c4ce33ddd456483e8189e6.png"},{"id":93462195,"identity":"7b30574a-5b76-44df-8636-1f8a71997997","added_by":"auto","created_at":"2025-10-14 06:29:11","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":291031,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLevels of anti-protein IgG at baseline in participant cohorts subsequently challenged with Spn6B or 15B. \u003c/strong\u003e\u0026nbsp;Baseline levels of anti-protein serum IgG in healthy adults aged 18-50, measured as average net median fluorescence intensity (MFI). Bars indicate interquartile range. Spn15B carriage-negative n = 15, Spn15B carriage-positive n = 9; Spn6B carriage-negative n = 18, Spn6B carriage-positive n = 21. Responses with p-value ≤ 0.05 prior to Benjamini-Hochberg correction for multiple comparisons are shown; this significance was lost following correction.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/5854dca306adb64d7c996279.png"},{"id":93461145,"identity":"d3665feb-c096-4f15-80ef-1090f29b7bef","added_by":"auto","created_at":"2025-10-14 06:21:10","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":229386,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCytokine/chemokine responses to pneumococcal proteins\u003c/strong\u003e. Baseline levels of cytokines/chemokines produced by PBMCs stimulated with pneumococcal proteins in healthy adults aged 18-50. Bars indicate interquartile range. Spn6B carriage-negative (blue) n = 32, Spn6B carriage-positive (orange) n = 31. Responses with p-value ≤ 0.05 prior to Benjamini-Hochberg correction for multiple comparisons are shown; this significance was lost following correction.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/7430ee52bcbd14ac461dcf72.png"},{"id":93462194,"identity":"601af36b-911a-40e3-a06b-3ef49aa5e7db","added_by":"auto","created_at":"2025-10-14 06:29:11","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":315279,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHumoral and cellular responses to the machine learning best ranked proteins antigens. \u003c/strong\u003eBoxplots showing \u003cstrong\u003e(A) \u003c/strong\u003eLog2 transformed average net MFI for Spn carriage-positive (orange; n = 36) and Spn carriage-negative (blue; n = 50) participants and \u003cstrong\u003e(B) \u003c/strong\u003eLog2 transformed cytokine/chemokine concentration for Spn carriage-positive (orange; n = 31) and Spn carriage-negative (blue; n = 32) participants. Each dot represents data from an individual participant.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/6a2ff3a5a85b5bcb50a4ea3f.png"},{"id":93461147,"identity":"feea730a-b5f2-48cd-9cab-db4f37e75fb6","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":181309,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCorrelation of protein-specific IgG levels and cytokine/chemokine responses with colonising density (AUC). \u003c/strong\u003eCorrelation results obtained between \u003cstrong\u003e(A)\u003c/strong\u003e antibody (n= 34) and \u003cstrong\u003e(B)\u003c/strong\u003e cytokine/chemokine (n= 27) levels for each protein and the AUC values obtained at the interval between days 2 to 14 post-challenge for each Spn6B carriage-positive participant. Proteins with \u003cem\u003ep\u003c/em\u003e-value ≤0.05 (denoted by horizontal dashed line) were highlighted, blue dots represent protein responses negatively correlated with AUC, while red dots represent the positive correlations. Following Benjamini Hochberg correction for multiple comparisons, all significance was lost.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/e19784ccb3270340b5d16d71.png"},{"id":93461160,"identity":"02097759-dcc0-4670-86d5-0216af47da32","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":325642,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHumoral and cellular responses across AUC groups for selected proteins showing trend-specific patterns\u003c/strong\u003e. \u003cstrong\u003e(A) \u003c/strong\u003eBaseline levels of anti-protein serum IgG in healthy adults aged 18-50, measured as average net median fluorescence intensity (MFI). Bars indicate interquartile range. Spn15B carriage-negative n = 15, Spn15B carriage-positive n = 13, low AUC n = 8, mid n = 1 and high n = 4; Spn6B carriage-negative n = 18, Spn6B carriage-positive n = 21, low AUC group n = 4, mid n = 6 and high n = 11. \u003cstrong\u003e(B) \u003c/strong\u003eBaseline levels of selected cytokines/chemokines stimulated by pneumococcal proteins in PBMCs from healthy adults aged 18-50. Carriage-negative - grey; low AUC - yellow, mid AUC - orange; high AUC - red. Bars indicate interquartile range. Spn6B carriage-negative n = 18, Spn6B carriage-positive n = 27, low AUC group n = 4, mid n = 14 and high n = 9.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/a883c37b3577dec8d453090d.png"},{"id":94044979,"identity":"596c28da-c811-4a9b-9aca-407720ef3042","added_by":"auto","created_at":"2025-10-21 21:01:24","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2303005,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/2b1adbee-7286-4e3e-927d-b8dd09eb17d7.pdf"},{"id":93461149,"identity":"4e9f0522-2a8a-4fcc-9ec2-30b64c52445e","added_by":"auto","created_at":"2025-10-14 06:21:11","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2966550,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterialProteinsPaper20Aug2025FV.docx","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/380ffa3d6d392f14b687fe36.docx"},{"id":93461143,"identity":"ba1c900e-3647-46a9-bb40-42c8c6beb1ab","added_by":"auto","created_at":"2025-10-14 06:21:10","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":135070,"visible":true,"origin":"","legend":"","description":"","filename":"GraphicalAbstract.docx","url":"https://assets-eu.researchsquare.com/files/rs-7566878/v1/f2b70eaa50dbd31ce4eea3b7.docx"}],"financialInterests":"Competing interest reported. RM is a consultant to GSK and Merck, is a named inventor and patent holder on vaccine technologies, a member of the board of directors at Corner Therapeutics and the scientific advisory boards of Amplitude Therapeutics, Limmatech and Vitrivax. Y-JL is a consultant to GSK and is a named inventor and patent holder on vaccine technologies. AJP is Chair of the UK Department of Health and Social Care’s Joint Committee on Vaccines and Immunisation and was previously chair and is now a member of WHOs technical advisory group on Salmonella vaccines. AJP is a contributor to intellectual property on a COVID19 vaccine licensed by Oxford University Innovation to AstraZeneca. Oxford University has received funding for research on Salmonella vaccines the Serum Institute of India.","formattedTitle":"Machine learning-driven identification of serotype-independent pneumococcal vaccine candidates using samples from human infection challenge studies","fulltext":[{"header":"Introduction","content":"\u003cp\u003ePneumonia is the leading infectious cause of death in infants worldwide, with more than 800,000 annual deaths among children under 5 years of age globally \u003csup\u003e\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. The most common aetiological agent of pneumonia is the bacterium \u003cem\u003eStreptococcus pneumoniae.\u003c/em\u003e Nasopharyngeal colonisation with pneumococcus, as well as being a prerequisite for invasive disease, is a source of transmission in the community \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThe pneumococcal conjugate vaccine (PCV) generates population herd protection providing substantial cost-benefit of vaccination \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. Vaccine-induced humoral responses to the pneumococcal capsule polysaccharide protect against pneumococcal carriage acquisition and disease caused by vaccine specific serotypes \u003csup\u003e\u003cspan additionalcitationids=\"CR7 CR8 CR9\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. However, current vaccines \u0026ndash; pneumococcal polysaccharide vaccine (PPV23) and PCVs \u0026ndash; offer protection against a limited number of the \u0026gt;\u0026thinsp;100 pneumococcal serotypes \u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e leading to serotype replacement of non-vaccine types \u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. In addition, despite being included in PCVs, serotype 3 continues to circulate and cause a high disease burden particularly in older adults \u003csup\u003e\u003cspan additionalcitationids=\"CR15\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. PCVs are also complex and costly to produce, which can present a barrier to vaccine access in resource-limited countries, particularly those graduating from GAVI support \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003ePneumococcal protein antigens that are universally expressed across various serotypes may induce both humoral and cellular immune responses, offering broad protection. These antigens could potentially be introduced as stand-alone vaccines or be included as conjugate proteins in current polysaccharide-based vaccines \u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eCurrently the regulatory approval pathway for next-generation pneumococcal vaccines in infants relies on the non-inferiority of IgG levels to capsular polysaccharide using a pre-defined threshold of 0.35ug/ml. In adults demonstration of non-inferior or superior opsonophagocytic antibody responses to each serotype compared to existing licensed vaccines is required\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. Therefore, the regulatory pathway for capsular polysaccharide independent vaccines is not defined. The identification of serotype-independent cellular and humoral correlates of protection against pneumococcal infection could assist with this framework and help to accelerate development of serotype-independent vaccines.\u003c/p\u003e\u003cp\u003eSystems-biology and artificial intelligence can be used to significantly enhance our understanding of immunity and aid in the identification of correlates of protection. Embedding the understanding of individual components and their interactions, beyond canonical antibody binding and neutralisation, may accelerate the design of more successful and long-lasting vaccines \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eIn this study, we leveraged a controlled human infection model (CHIM) to investigate immune responses to pneumococcal proteins that are conserved across serotypes, aiming to identify correlates of serotype-independent protection and to understand their impact on colonisation dynamics in the nasopharynx. To enable high-throughput assessment of protein-specific IgG responses, we established a Luminex assay targeting 75 pneumococcal protein antigens. Given the well-established role of IL-17 in mediating protection against \u003cem\u003eS. pneumoniae\u003c/em\u003e in murine models, we also evaluated IL-17-driven cellular responses in protein-stimulated peripheral blood mononuclear cells (PBMCs). In parallel, we employed a multiplex cytokine assay to profile broader cytokine responses, with MCP-1 (CCL2), RANTES (CCL5), and IFN-γ selected for more detailed analysis.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cem\u003eStudy Design\u003c/em\u003e\u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe overall study design is shown in Figure 1. Briefly, at baseline (day -4 and day -5), participants were assessed before intranasal inoculation (day 0) with \u003cem\u003eS. pneumoniae\u003c/em\u003e (Spn). Carriage acquisition was monitored until day 14 for Spn15B and day 29 for Spn6B.\u003c/p\u003e\n\u003cp\u003eImmune responses were evaluated using Luminex assays, which measured IgG antibody responses to 75 pneumococcal antigens in 72 volunteers, and cytokine profiling, which detected 21 immune markers (including IL-17, IFN\u0026gamma;, RANTES, and MCP-1) in 63 volunteers.\u003c/p\u003e\n\u003cp\u003eCarriage acquisition results are presented as bar graphs, depicting the number of individuals testing positive or negative for Spn6B and Spn15B at different time points. These results indicate the proportion of participants who acquired pneumococcal carriage following inoculation. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eParticipant cohort inoculated with serotype 6B pneumococcus\u003c/em\u003e \u003c/p\u003e\n\u003cp\u003eBlood samples were taken from healthy non-smoking participants between the ages of 18 and 50 who were intranasally inoculated with 80,000 CFU in 0.1ml solution of\u003cem\u003e S. pneumoniae\u003c/em\u003e serotype 6B (Spn6B) strain BHN418 per naris (full sequence available GenBank: ASHP00000000). Samples were taken prior to inoculation (baseline). Cohorts were recruited as part EudraCT (2014-004634-26), a study to evaluate the effect of live attenuated influenza vaccine (LAIV) on experimental human pneumococcal colonisation (EHPC) \u003csup\u003e21,22\u003c/sup\u003e\u003csup\u003e21,22\u003c/sup\u003e. All participants gave written informed consent, with approval from the North West NHS Research Ethics Committee (14/NW/1460). Nasal wash samples were taken on days 2, 6, 9, 14 and 29 post-challenge with Spn6B. Colonisation was determined by classical microbiology and the presence of Spn6B in nasal wash samples taken at pre-defined time-points between inoculation and 29 days post-inoculation. No participants had previously received any pneumococcal vaccine. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eParticipant cohort inoculated with serotype 15B pneumococcus\u003c/em\u003e \u003c/p\u003e\n\u003cp\u003eHealthy, non-smoking participants aged 18-50 years were intranasally inoculated with 80,000 CFU \u003cem\u003eS. pneumoniae \u003c/em\u003eserotype 15B (Spn15B) per naris (clinical isolate 15B P1262. European Nucleotide Archive accession number: ERS2632437). Spn15B is included in the PPV23 formulation but not in the 13-valent PCV formulation. All participants gave informed written consent, ethical approval was obtained from the National Health Service Research Ethics Committee, Liverpool East (15/NW/0931). Serum samples were taken at baseline (pre-inoculation) (Trial identifier: ISRCTN 68323432; 20815). Colonisation was determined by classical microbiology and the presence of Spn15B in nasal wash samples taken at pre-defined time-points between inoculation and 14 days post-inoculation. Nasal wash samples were taken on days 2, 7 and 14 post-challenge with Spn15B. No participants had previously received any pneumococcal vaccine. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eLuminex assay\u003c/em\u003e \u003c/p\u003e\n\u003cp\u003eThe proteins used in the assay are given in Supplementary Table 1. The sequences encoding the alpha-helical N-terminal domain of PspA1, PspC6, PspC9, were cloned using DNA from strain BHN418 (serotype 6B pneumococcus). The sequence encoding the N-terminal domain of PspA4 was cloned from strain 255/00 (serotype 14, GenBank accession number EF649969). Recombinant proteins were produced at Butantan Institute laboratories as described previously \u003csup\u003e23\u003c/sup\u003e including pneumolysin toxoid B (PdB), a pneumolysin mutant with a tryptophan-to-phenylalanine substitution at position 433 and 0.1% haemolytic activity of wild-type pneumolysin \u003csup\u003e24\u003c/sup\u003e\u003csup\u003e25\u003c/sup\u003e. The remaining 70 proteins in the library were produced at Boston Children\u0026rsquo;s Hospital as described \u003csup\u003e26\u003c/sup\u003e. \u003c/p\u003e\n\u003cp\u003eAll 75 pneumococcal protein antigens were conjugated to magnetic microspheres at a concentration of 5\u0026micro;g per million microspheres using the xMAP Antibody Coupling Kit as per the manufacturer\u0026rsquo;s instructions. The assay was validated for the detection of anti-protein IgG as described previously \u003csup\u003e27,28\u003c/sup\u003e. Full methods are described in detail in Supplementary Text 1. A stock of pooled sera taken 29 days post-challenge and diluted 1:50 in assay buffer was used as an internal positive control across all plates and assay buffer was used as a negative control to detect background median fluorescence intensity (MFI) values. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eLuminex data cleaning and statistical analysis\u003c/em\u003e \u003c/p\u003e\n\u003cp\u003eRaw Luminex data were cleaned prior to analysis through the following workflow. The coefficient of variation (CV) of replicates was calculated for all analytes. If any analyte had greater than 25% CV and both replicates had greater than 50 microspheres per well, the analyte was removed from analysis. If any paired samples had less than 50 microspheres of a given analyte in each well, the analyte was removed from analysis. The determinant of 50 microspheres is based on recommendations by the Luminex Corporation that a recovery of less than 50 microspheres renders a result unreliable as well as previous work showing that lower bead counts elicit higher CV values \u003csup\u003e29\u003c/sup\u003e. All statistical analysis was carried out in RStudio (version 1.0.153). Statistical significance was determined via a Mann-Whitney or Wilcoxon test for unpaired and paired groups, respectively, followed by Benjamini Hochberg correction for multiple comparisons. For the machine learning analysis, any proteins with greater than 25% of values missing and any participants with more than 50% of values missing were removed from the analysis. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eStimulation of human peripheral blood mononuclear cells (PBMCs) and cytokine/chemokine detection\u003c/em\u003e \u003c/p\u003e\n\u003cp\u003eBaseline PBMCs from healthy participants inoculated with Spn6B were thawed with 50 \u0026micro;g/ml DNAse I (MilliporeSigma) in prewarmed RPMI containing 10% FBS and washed once in media including DNAse I. 1x10\u003csup\u003e6\u003c/sup\u003e PBMCs/ml from an initial cohort of 12 carriage-negative and 11 carriage-positive participants were stimulated with 8\u0026micro;g/ml of each protein from the complete library and 5\u0026micro;g/ml of heat-inactivated Spn6B for 7 days at 37\u0026deg;C under 5% CO\u003csub\u003e2\u003c/sub\u003e. The plates were then centrifuged to pellet the cells, after which the supernatants were collected to measure IL-17A concentration using human ELISA kits (Invitrogen). In parallel, supernatants from carriage-negative and carriage-positive participants from this initial cohort were combined and assessed for other cytokine and chemokine concentrations using a 30-plex magnetic human Luminex cytokine kit (Thermo Fisher Scientific) and acquired on an LX200. A log 2-fold change between carriage-negative and positive was calculated and a Wilcoxon paired test was used to identify cytokines/chemokines associated with a specific carriage status. \u003c/p\u003e\n\u003cp\u003eTwenty-one antigens from a 70-protein antigen library were selected for further analysis. The antigens were selected based on (i) their ability to induce high IL-17A responses in carriage-negative individuals (ii) negative correlation between high IL-17A levels and low pneumococcal density in participants who were experimentally colonised (iii) their ability to induce high expression of cytokines/chemokines at baseline prior to challenge (iv) available published data in which the antigens have shown protection against challenge after immunisation in pre-clinical models of infection. \u003c/p\u003e\n\u003cp\u003ePBMCs from an additional 22 carriage-negative and 21 carriage-positive participants were stimulated with the selected 21 antigens and cytokine/chemokine concentrations were measured to validate the results obtained with the initial cohort. Human ELISA kits were used to measure concentrations of IL-17A (Invitrogen), MCP-1 (CCL-2) (BD OptEIA\u003csup\u003eTM\u003c/sup\u003e), IFN-\u0026gamma; (BD OptEIA\u003csup\u003eTM\u003c/sup\u003e), and RANTES (CCL-5) (Biolegend) in the collected supernatants following the manufacturers recommendations. Non-stimulated (control) and protein-stimulated supernatants were diluted as follows: 1:2 for IL-17, 1:10 for RANTES, 1:200 for IFN-\u0026gamma;, and 1:3200 for MCP-1. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eELISA data cleaning and statistical analysis\u003c/em\u003e \u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAs detailed above, any proteins with greater than 25% of values missing and any participants with more than 50% of values missing were removed from the analysis. Statistical significance was determined via a Wilcoxon test for unpaired groups, followed by Benjamini Hochberg correction for multiple comparisons. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eArea under the curve (AUC) calculation\u003c/em\u003e\u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eArea under the curve (AUC) was used to calculate overall colonising density of \u003cem\u003eS. pneumoniae\u003c/em\u003e in the nasopharynx of experimentally colonised participants. AUC was calculated by trapezoidal rule from log 10 transformed density results at 2, 6, 16, 22, 27 and 36 days or after challenge on day 2, 7 or 14 with interpolation of missing data assuming a linear change in density between known values. The AUC of pneumococcal colonising density at each timepoint was derived and summarised using the above descriptive statistics and analysed using a generalised linear model with a single factor of treatment.\u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAUC correlation analysis \u003c/em\u003e \u003c/p\u003e\n\u003cp\u003eA Spearman test was performed to evaluate the correlation between the IgG and cytokine/chemokine levels elicited in response to each protein and the AUC values obtained between days 2 to 14 for each carriage-positive participant using the cor.test function in the software R. The analysis was performed using antibody response and cellular response from 34 and 27 participants, respectively. Correlations with \u003cem\u003ep\u003c/em\u003e-value \u0026le; 0.05 were considered significant. \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAUC threshold definition\u003c/em\u003e \u003c/p\u003e\n\u003cp\u003eRetrospectively, we calculated the AUC of nasopharyngeal Spn6B density between days 2 to 14 post-challenge in 128 experimentally colonised participants across our CHIM studies to define \u0026ldquo;Low\u0026rdquo;, \u0026ldquo;Mid\u0026rdquo; or \u0026ldquo;High\u0026rdquo; colonising density. For participants experimentally colonised with Spn15B, threshold AUC values were set using colonising density between days 2 to 14 from 9 participants. Threshold values differed for the two serotypes. The data distribution was used to split the groups for both Spn6B and Spn15B inoculated cohorts. The participants from different AUC groups were localised in different quartiles (Supplementary Figure 1). \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eMachine learning analysis \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eMissing values in numerical variables were imputed using the \u0026quot;mice\u0026quot; R package \u003csup\u003e30\u003c/sup\u003e. Numeric features were standardized using z-score normalization. The Random Forest algorithm, an ensemble machine learning method that constructs multiple decision trees using random feature subsets and aggregates their outputs to enhance predictive accuracy and minimize overfitting (28), implemented in the R package \u0026quot;randomForest\u0026quot;. The model was trained with mtry = 1 and ntree = 500, while other hyperparameters were set to their default values. This algorithm was employed to identify baseline protein-specific antibodies and cytokines/chemokines produced in response to PBMC stimulation with protein antigens that best predicted carriage status post-challenge. Unlike univariate analysis, which assesses each feature independently, machine learning techniques such as Random Forest can uncover complex, multivariate relationships among features, improving predictive accuracy. \u003c/p\u003e\n\u003cp\u003eFeature importance was assessed using the mean decrease in Gini index, a metric that quantifies how much a given feature contributes to reducing impurity (i.e., improving class separation) within the decision trees. A higher mean decrease in Gini index indicates a stronger contribution to classification accuracy. \u003c/p\u003e\n\u003cp\u003eTo identify the most relevant features and optimize model performance, we applied an iterative feature selection process based on feature importance. Initially, we trained the Random Forest model using all 75 available features and evaluated their importance based on the mean decrease in Gini index. We then progressively removed the least informative features, retraining the model at each iteration and assessing its predictive performance. This process was repeated several times, testing progressively smaller feature sets (e.g., 50, 40, etc.), until the optimal configuration was identified, the smallest number of features yielding the best model performance, minimizing error, and reducing the risk of overfitting. Model evaluation was conducted using 10-fold cross-validation. \u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cem\u003eSerum IgG levels to 75 different protein antigens at baseline did not associate with protection against experimental pneumococcal carriage acquisition of Spn6B or Spn15B\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eA Luminex-based multiplex assay using a library of 75 highly conserved pneumococcal proteins (Supplementary Table 1), was successfully optimised and validated \u003csup\u003e28\u003c/sup\u003e. No associations between baseline levels of serum IgG against any of the protein antigens and protection against carriage acquisition remained statistically significant after applying the Benjamini-Hochberg correction for multiple testing (Figure 2). \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eScreening of antigens using human PBMCs and cellular responses \u003c/em\u003e\u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eA library of pneumococcal protein antigens was used to stimulate PBMCs collected from participants prior to experimental challenge with Spn6B. Due to the limited number of PBMCs obtained from each participant, we were not able to screen all protein antigens in all participants. IL-17A production has previously been described as a marker of protection against pneumococcal colonisation in mice\u003csup\u003e31,32\u003c/sup\u003e. IL-17A secretion following ex-vivo protein stimulation was observed for all proteins at 7 days of culture. \u003c/p\u003e\n\u003cp\u003eTo identify other cellular markers that could associate with a protective or susceptible profile in the CHIM, we pooled supernatants from PBMC cultures following individual protein stimulation from carriage-negative and carriage-positive participants and compared levels produced of 30 cytokines and chemokines using 30-plex Luminex in these two groups (Supplementary Figure 2). MCP-1 (CCL-2) was selected as a marker of protection based on its well-documented role in monocyte recruitment and the clearance of pneumococcal carriage \u003csup\u003e21\u003c/sup\u003e. RANTES (CCL5) and IFN-\u0026gamma; were selected due to their elevated levels in carriage-positive individuals and their known involvement in T cell-mediated immune responses \u003csup\u003e33\u003c/sup\u003e. \u003c/p\u003e\n\u003cp\u003eWe used the data obtained from the initial IL-17A screening of susceptible and protected profiles (Supplementary Figure 3), together with available data on immunisation with those antigens and protection against carriage in mice (24), to down-select a limited number of antigens (Supplementary Table 2) for PBMC stimulation assays. PBMCs from a further 22 carriage-negative and 21 Spn6B carriage-positive participants were stimulated with the down-selected 21 antigens and IL-17A, RANTES, MCP-1 and IFN-\u0026gamma; responses were measured. \u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eBaseline cytokine/chemokine levels in response to protein antigens are not associate with protection against experimental pneumococcal carriage acquisition with Spn6B \u003c/em\u003e\u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAfter correction for multiple comparisons, there was no significant difference between baseline protein-specific cytokine/chemokine responses in participants susceptible to experimental colonisation with Spn6B and those who were protected against colonisation (Figure 3). Responses with p-values \u0026le; 0.05 for the carriage-positive and carriage-negative groups comparison prior to Benjamini-Hochberg correction are shown (Figure 3). \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eRandom Forest algorithm as an alternative method to rank antigens associated with pneumococcal carriage protection\u003c/em\u003e\u003cem\u003e\u003cbr\u003e\u003c/em\u003eWe used the Random Forest method \u003csup\u003e34\u003c/sup\u003e to assess the predictive power of baseline protein-specific antibody (IgG) and cytokine/chemokine levels in predicting susceptibility/protection to carriage acquisition. We trained the model using various features at baseline and selected the combination of features that generated the model with the least error. Antibody response patterns to the proteins PdB, SP0899, and SP1069 together emerged as the most important features, with the model achieving an out-of-bag (OOB) error rate of 36.05%. In terms of cytokine/chemokine responses, IL-17A production in response to SP0648-3 and MCP-1 production in response to SP1069 and SP0899 together were identified as the most relevant features, resulting in an OOB error of 38.98%. \u003c/p\u003e\n\u003cp\u003eNotably, SP0899 and SP1069 were a feature of both humoral and cellular responses that were predictive of protection against colonisation (Figure 4). Higher baseline IgG against SP1069 and SP0899 was observed in the cohort of participants protected against experimental carriage acquisition (Figure 4). We also observed that those participants protected against colonisation had higher levels of MCP-1 produced in response to stimulation with SP1069 (\u003cem\u003ep = \u003c/em\u003e0.035) and SP0899 (\u003cem\u003ep = \u003c/em\u003e0.049) than those susceptible to colonisation at baseline (Figure 4). \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCorrelation of antibody and cytokine/chemokine levels against the proteins selected by machine learning analysis\u003c/em\u003e\u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eTo determine whether there was a relationship between antibody- and cell-mediated immune responses, we correlated responses identified as predictive of colonisation status by machine learning. Whilst SP0648-3 was found to be correlated in carriage positive participants across datasets (\u003cem\u003eRho =\u003c/em\u003e 0.614; \u003cem\u003ep =\u003c/em\u003e 0.009), SP1069 (\u003cem\u003eRho =\u003c/em\u003e 0.378; \u003cem\u003ep =\u003c/em\u003e 0.203) and SP0899 (\u003cem\u003eRho =\u003c/em\u003e -0.242; \u003cem\u003ep =\u003c/em\u003e 0.425) were not. None of the proteins were correlated in carriage negative participants across datasets (SP0648-3 (\u003cem\u003eRho =\u003c/em\u003e -0.076; \u003cem\u003ep =\u003c/em\u003e 0.781), SP1069 (\u003cem\u003eRho = \u003c/em\u003e-0.422; \u003cem\u003ep =\u003c/em\u003e 0.133), SP0899 (\u003cem\u003eRho =\u003c/em\u003e -0.098; \u003cem\u003ep =\u003c/em\u003e 0.781)). \u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCorrelation between protein-responses and AUC\u003c/em\u003e\u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eWe generated area under the curve data for colonising density of the participants colonised with pneumococcus (Figure 5). No significant correlations were found following Benjamini Hochberg correction. Correlations showing \u003cem\u003ep\u003c/em\u003e-value \u0026le; 0.05 prior to correction are highlighted in Figure 5 and Supplementary Tables 3 and 4. Participants were stratified by colonisation density and duration, expressed as AUC (Supplementary Figure 4). With respect to cell-mediated immunity, the levels of RANTES, MCP-1, and IFN-\u0026gamma; produced in response to protein stimuli (SP1154-2, SP0648-3, SP0079, and SP1683) differed across density groups (Supplementary Figure 4 B). \u003c/p\u003e\n\u003cp\u003eAlthough differences were not statistically significant, we observed a trend in high levels of antibodies and cytokines/chemokines specific to protein antigens that were associated with lower colonisation density (Figure 6). We observed that IgG specific to 32 proteins (SP0321, SP0564, SP0346, SP1545, SP1032, SP0617, SP0620, SP0678, SP1479, SP1492, SP0079, SP0402, SP0629, SP0757, SP2145, SP0662-1, SP1683, SP0878, SP0582, SP1376, SP0724, SP1500, SP2083, SP0098, SP0127, SP0604, SP0648-1, SP2207, SP0742, SP0314, SP0648-2, PspA1) was elevated in participants who were either protected against experimental carriage acquisition, or in those who were experimentally colonised at a relatively low bacterial density (Figure 6A). We also observed that cytokine and chemokine responses were elevated in participants who were either protected against experimental carriage acquisition or in the low AUC density group, followed by the mid AUC density and high AUC density groups \u0026ndash; IFN-\u0026gamma; and RANTES responses were particularly elevated (Figure 6B). Cytokine and chemokine levels in carriage-negative participants were generally not higher than those observed in the low AUC group. From the selected proteins, only IFN-\u0026gamma; and RANTES responses to SP1154-2 (IFN-\u0026gamma; \u003cem\u003ep =\u003c/em\u003e 0.034; RANTES \u003cem\u003ep =\u003c/em\u003e 0.024) and SP0079 (IFN-\u0026gamma; \u003cem\u003ep =\u003c/em\u003e 0.012; RANTES \u003cem\u003ep =\u003c/em\u003e 0.005) and the RANTES response to SP0648-1 (p = 0.027) have shown a \u003cem\u003ep\u003c/em\u003e \u0026le; 0.05 before Benjamini-Hochberg correction (Figure 6B).\u003c/p\u003e\n\u003cp\u003eAs previously stated, SP1069 and SP0899 were selected by machine learning for carriage status classification. The IFN-\u0026gamma; response to SP1069 followed the same pattern, levels were highest in the carriage-negative group or in participants colonised at low bacterial density, and lowest in those with high-density colonisation. There was a very marginal difference in the levels of MCP-1 produced in response to SP0899 across groups. \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eMachine learning and systems biology are increasingly employed in vaccine development to identify and design vaccine candidates as well as to identify correlates of successful immunisation, predict individual vaccine responses, and correlates of protection when a disease/infection endpoint is possible \u003csup\u003e35,36\u003c/sup\u003e. As the number of measurable biological parameters expands, traditional univariate analyses are inappropriate, increasing the risk of type II errors (failing to detect true associations). In contrast, machine learning enables a more comprehensive and rigorous analysis of complex, high-dimensional datasets, offering deeper insights into the multifactorial nature of immune responses. In this study, we leveraged machine learning to identify immune features predictive of protection against nasal colonisation. Our findings highlight the Random Forest algorithm\u0026rsquo;s utility in identifying proteins linked to protection against experimental pneumococcal carriage, demonstrating its potential to accelerate vaccine development.\u003c/p\u003e\n\u003cp\u003eThe ultimate aim of pneumococcal vaccination may not be sterilising immunity and complete elimination of nasopharyngeal colonisation, a natural immunising event \u003csup\u003e33,36,37,52\u003c/sup\u003e, but rather reducing colonisation density to prevent disease progression or transmission to others in the population, particularly during viral co-infections \u003csup\u003e53\u003c/sup\u003e. Suppressing bacterial density through protein-based vaccines could minimise transmission and reinforce immunity while maintaining mucosal exposure necessary for lung immunity \u003csup\u003e54\u003c/sup\u003e.\u003c/p\u003e\n\n\u003cp\u003eConsistent with prior studies \u003csup\u003e36\u0026ndash;38\u003c/sup\u003e, systemic anti-protein IgG at baseline did not correlate with protection against experimental colonisation by serotype 6B or 15B pneumococcus. Similarly, baseline cell-mediated responses in blood were not predictive of protection in our human cohort, despite evidence of Th17-mediated immunity against pneumococcal colonisation in murine models \u003csup\u003e26\u003c/sup\u003e. Using machine learning, we identified a group of three antigens \u0026mdash; SP1069, SP0899, and PdB \u0026mdash; eliciting IgG responses, and a group three antigens \u0026mdash; SP1069, SP0899, and SP0648-3 \u0026mdash; which stimulated protein-specific cellular responses, that were predictive of protection. Although we cannot yet determine whether these protein combinations could form the basis of a combination vaccine, this possibility warrants further investigation. Notably, SP1069 and SP0899 elicited both humoral and cellular responses. Although the exact functions of these proteins are not yet fully understood, immunisation of mice with a combination of cholera toxin alongside SP1069, SP0899 and SP0648-3 as individual antigens has led to a significant reduction in pneumococcal colonisation compared with the control groups \u003csup\u003e26\u003c/sup\u003e. Co-administration of multiple antigens with a cholera toxin adjuvant has been shown to enhance the protective efficacy in murine models \u003csup\u003e26\u003c/sup\u003e. Protein-based vaccines incorporating several antigens may thus offer broad coverage, while conjugating protein antigens to capsular polysaccharides in new PCV formulations could provide additional serotype-independent protection \u003csup\u003e39,40\u003c/sup\u003e as well as the advantage of cellular immunity induction.\u003c/p\u003e\n\n\u003cp\u003eIn humans, protection against pneumococcal disease requires both humoral and cellular immunity \u003csup\u003e31,41\u0026ndash;46\u003c/sup\u003e. Similarly, murine models indicate the necessity of both immunity types for nasopharyngeal colonisation defence \u003csup\u003e26,41\u0026ndash;43,47,48\u003c/sup\u003e. Using our human challenge model, we previously reported that colonisation-induced MCP-1 (CCL2) levels are associated with monocyte recruitment to the nasal mucosa and subsequent clearance of colonisation \u003csup\u003e21\u003c/sup\u003e. These results also support previous findings from mouse models showing that CCL2 signalling and monocyte recruitment are key mediators of pneumococcal carriage clearance \u003csup\u003e49\u003c/sup\u003e. Elevated MCP-1 levels in protected participants against experimental acquisition of Spn6B and its involvement in monocyte and neutrophil recruitment underline its significance in pneumococcal clearance \u003csup\u003e21,50,51\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eIn our cohort, baseline anti-protein IgG and cytokine responses correlated with lower colonisation density, suggesting roles for both humoral and cellular immunity in controlling carriage and preventing disease. Consistent with prior findings \u003csup\u003e55\u0026ndash;58\u003c/sup\u003e, IFN-\u0026gamma; and RANTES secretion correlated with reduced colonisation density, likely enhancing innate immune cell recruitment and activity \u003csup\u003e59\u0026ndash;61\u003c/sup\u003e. These results emphasise the importance of designing protein vaccines that elicit both antibody and cellular responses for effective protection.\u003c/p\u003e\n\u003cp\u003eWhile promising, this work has limitations. Out-of-bag error rates (~36\u0026ndash;39%) indicate only moderate predictive power, suggesting that further optimisation, such as increasing cohort size or incorporating mucosal immune parameters, may be needed to improve model accuracy. Immune responses to protein antigens in systemic circulation may differ from those in the nasal mucosa, and the study\u0026rsquo;s small sample size, comprised mainly of healthy young adults, may not reflect high-risk demographics. Validating these findings in samples from older adults participants will be important \u003csup\u003e37,62\u003c/sup\u003e. Immune correlates may vary across populations with differing pneumococcal burden. Our studies in Malawi \u003csup\u003e10,63\u003c/sup\u003e will enable validation across diverse populations and environmental contexts. Further investigation into antigen-specific nasal T-cell responses and sex-stratified immune differences could provide additional insights \u003csup\u003e27\u003c/sup\u003e. \u003c/p\u003e\n\u003cp\u003eAs immunology and vaccinology increasingly rely on complex, high-dimensional datasets, predictive computational models offer substantial advantages over traditional analyses by uncovering intricate relationships between immune responses and protection. The integration of advanced computational approaches with human challenge models, which uniquely allow sampling immediately before and after defined pathogen exposure, promises to transform how we define protective immunity and accelerate rational vaccine design.\u003c/p\u003e\n\u003cp\u003eDespite the remarkable success of vaccines, which are estimated to have saved more than 150 million lives over the past 50 years\u003csup\u003e64\u003c/sup\u003e, major challenges persist. Unequal access to vaccines and the biological complexity of difficult-to-target pathogens threaten to stall progress in reducing the global burden of infectious diseases. A key scientific challenge is the identification of causal correlates of protection for high-frequency colonising pathogens such as \u003cem\u003eStreptococcus pneumoniae\u003c/em\u003e, \u003cem\u003eStaphylococcus aureus\u003c/em\u003e, and \u003cem\u003eEscherichia coli\u003c/em\u003e, where repeated antigen exposures at mucosal surfaces shape layered humoral and cellular immune responses. Addressing these gaps will require integrating advanced computational approaches, experimental human models, and global public health strategies to translate immunological insights into next-generation vaccines.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability:\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eData are available upon reasonable request by email directed to the corresponding authors at \u003cu\[email protected]\u003c/u\u003e and \u003cu\[email protected]\u003c/u\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eConceived the study - DMF, SJ, CS, BU, RM\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eData acquisition - KSC, CS, RT, JR, ELG, EM, YL, SP, ENM\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGenerated the ELISA data - CS\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePerformed the machine learning analysis - PGD\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePerformed data analysis - KSC, PGD, CS\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eScientific discussion \u0026ndash; AJP, SBG, HIN, SHP, ANAG, AMC, SPJ\u003c/p\u003e\n\u003cp\u003eProject definition and discussion - KSC, PGD, CS, DMF, BU\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eProvided protein library - RM, YJL, EM, RT\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePGD, KSC and CS jointly interpreted the results and wrote the manuscript with contributions from all co-authors. All authors read and approved the final manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of Interests\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eRM is a consultant to GSK and Merck, is a named inventor and patent holder on vaccine technologies, a member of the board of directors at Corner Therapeutics and the scientific advisory boards of Amplitude Therapeutics, Limmatech and Vitrivax. Y-JL is a consultant to GSK and is a named inventor and patent holder on vaccine technologies. AJP is Chair of the UK Department of Health and Social Care\u0026rsquo;s Joint Committee on Vaccines and Immunisation and was previously chair and is now a member of WHOs technical advisory group on \u003cem\u003eSalmonella\u003c/em\u003e vaccines. AJP is a contributor to intellectual property on a COVID19 vaccine licensed by Oxford University Innovation to AstraZeneca. Oxford University has received funding for research on Salmonella vaccines the Serum Institute of India.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe authors thank all participants in the EHPC studies, as well as the Clinical Research Network North West Coast and the NIHR Comprehensive Local Research Network North West Coast for their continued support. We also extend our gratitude to the study clinical team members for their assistance in sample collection. This work by funded by grants from the Bill \u0026amp; Melinda Gates Foundation (OPP1117728) (to DMF), Medical Research Council (MR/M011569/1) (to SBG and DMF), Robert Austrian Award (to SPJ) and the Bacterial Vaccines (BactiVac) Network funded by the GCRF Networks in Vaccines Research and Development which was co-funded by the MRC and BBSRC (to DMF and CS). KSC was supported by the MRC Doctoral Training Programme (Award 1964515).\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;\u0026ensp;GBD 2019 Antimicrobial Resistance Collaborators. Global mortality associated with 33 bacterial pathogens in 2019: a systematic analysis for the Global Burden of Disease Study 2019. \u003cem\u003eLancet\u003c/em\u003e \u003cstrong\u003e400\u003c/strong\u003e, 2221\u0026ndash;2248 (2022).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;\u0026ensp;Wahl, B. \u003cem\u003eet al.\u003c/em\u003e Burden of Streptococcus pneumoniae and Haemophilus influenzae type b disease in children in the era of conjugate vaccines: global, regional, and national estimates for 2000-15. \u003cem\u003eLancet Glob. Health\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, e744\u0026ndash;e757 (2018).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;\u0026ensp;McAllister, D. A. \u003cem\u003eet al.\u003c/em\u003e Global, regional, and national estimates of pneumonia morbidity and mortality in children younger than 5 years between 2000 and 2015: a systematic analysis. \u003cem\u003eLancet Glob. Health\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, e47\u0026ndash;e57 (2019).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;\u0026ensp;Weiser, J. N., Ferreira, D. M. \u0026amp; Paton, J. C. Streptococcus pneumoniae: transmission, colonization and invasion. \u003cem\u003eNat. Rev. Microbiol.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 355\u0026ndash;367 (2018).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;\u0026ensp;Ladhani, S. N., Andrews, N. \u0026amp; Ramsay, M. E. Summary of evidence to reduce the two-dose infant priming schedule to a single dose of the 13-valent pneumococcal conjugate vaccine in the national immunisation programme in the UK. \u003cem\u003eLancet Infect. Dis.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, e93\u0026ndash;e102 (2021).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;\u0026ensp;Black, S. \u003cem\u003eet al.\u003c/em\u003e Efficacy, safety and immunogenicity of heptavalent pneumococcal conjugate vaccine in children. Northern California Kaiser Permanente Vaccine Study Center Group. \u003cem\u003ePediatr. Infect. Dis. J.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 187\u0026ndash;195 (2000).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;\u0026ensp;Ota, M. O. C. \u003cem\u003eet al.\u003c/em\u003e Pneumococcal antibody concentrations of subjects in communities fully or partially vaccinated with a seven-valent pneumococcal conjugate vaccine. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, e42997 (2012).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;\u0026ensp;Collins, A. M. \u003cem\u003eet al.\u003c/em\u003e First human challenge testing of a pneumococcal vaccine. Double-blind randomized controlled trial. \u003cem\u003eAm. J. Respir. Crit. Care Med.\u003c/em\u003e \u003cstrong\u003e192\u003c/strong\u003e, 853\u0026ndash;858 (2015).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;\u0026ensp;Nagel, J. \u003cem\u003eet al.\u003c/em\u003e The association between antibody levels before and after 7-valent pneumococcal conjugate vaccine immunization and subsequent pneumococcal infection in chronic arthritis patients. \u003cem\u003eArthritis Res. Ther.\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 124 (2015).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Bar-Zeev, N. \u003cem\u003eet al.\u003c/em\u003e Impact and effectiveness of 13-valent pneumococcal conjugate vaccine on population incidence of vaccine and non-vaccine serotype invasive pneumococcal disease in Blantyre, Malawi, 2006-18: prospective observational time-series and case-control studies. \u003cem\u003eLancet Glob. Health\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, e989\u0026ndash;e998 (2021).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Singh, A. \u0026amp; Dutta, A. K. Pneumococcal Vaccines - How Many Serotypes are Enough? \u003cem\u003eIndian J. Pediatr.\u003c/em\u003e \u003cstrong\u003e85\u003c/strong\u003e, 47\u0026ndash;52 (2018).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Weinberger, D. M., Malley, R. \u0026amp; Lipsitch, M. Serotype replacement in disease after pneumococcal vaccination. \u003cem\u003eLancet\u003c/em\u003e \u003cstrong\u003e378\u003c/strong\u003e, 1962\u0026ndash;1973 (2011).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Olaya-Abril, A., Jim\u0026eacute;nez-Mungu\u0026iacute;a, I., G\u0026oacute;mez-Gasc\u0026oacute;n, L., Obando, I. \u0026amp; Rodr\u0026iacute;guez-Ortega, M. J. Identification of potential new protein vaccine candidates through pan-surfomic analysis of pneumococcal clinical isolates from adults. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, e70365 (2013).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Tiley, K. S. \u003cem\u003eet al.\u003c/em\u003e Nasopharyngeal Carriage of Pneumococcus in Children in England up to 10 Years After 13-Valent Pneumococcal Conjugate Vaccine Introduction: Persistence of Serotypes 3 and 19A and Emergence of 7C. \u003cem\u003eJ. Infect. Dis.\u003c/em\u003e \u003cstrong\u003e227\u003c/strong\u003e, 610\u0026ndash;621 (2023).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Pick, H. \u003cem\u003eet al.\u003c/em\u003e Pneumococcal serotype trends, surveillance and risk factors in UK adult pneumonia, 2013-18. \u003cem\u003eThorax\u003c/em\u003e \u003cstrong\u003e75\u003c/strong\u003e, 38\u0026ndash;49 (2020).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Ladhani, S. N. \u003cem\u003eet al.\u003c/em\u003e Rapid increase in non-vaccine serotypes causing invasive pneumococcal disease in England and Wales, 2000-17: a prospective national observational cohort study. \u003cem\u003eLancet Infect. Dis.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 441\u0026ndash;451 (2018).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Chen, C. \u003cem\u003eet al.\u003c/em\u003e Effect and cost-effectiveness of pneumococcal conjugate vaccination: a global modelling analysis. \u003cem\u003eLancet Glob. Health\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, e58\u0026ndash;e67 (2019).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Lagousi, T., Basdeki, P., Routsias, J. \u0026amp; Spoulou, V. Novel Protein-Based Pneumococcal Vaccines: Assessing the Use of Distinct Protein Fragments Instead of Full-Length Proteins as Vaccine Antigens. \u003cem\u003eVaccines (Basel)\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, (2019).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Grant, L. R. \u003cem\u003eet al.\u003c/em\u003e Persistence of IgG antibody following routine infant immunization with the 7-valent pneumococcal conjugate vaccine. \u003cem\u003ePediatr. Infect. Dis. J.\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, e138-42 (2015).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Britto, C. \u0026amp; Alter, G. The next frontier in vaccine design: blending immune correlates of protection into rational vaccine design. \u003cem\u003eCurr. Opin. Immunol.\u003c/em\u003e \u003cstrong\u003e78\u003c/strong\u003e, 102234 (2022).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Jochems, S. P. \u003cem\u003eet al.\u003c/em\u003e Inflammation induced by influenza virus impairs human innate immune control of pneumococcus. \u003cem\u003eNat. Immunol.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 1299\u0026ndash;1308 (2018).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Rylance, J. \u003cem\u003eet al.\u003c/em\u003e Two randomized trials of the effect of live attenuated influenza vaccine on pneumococcal colonization. \u003cem\u003eAm. J. Respir. Crit. Care Med.\u003c/em\u003e \u003cstrong\u003e199\u003c/strong\u003e, 1160\u0026ndash;1163 (2019).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Darrieux, M. \u003cem\u003eet al.\u003c/em\u003e Fusion proteins containing family 1 and family 2 PspA fragments elicit protection against Streptococcus pneumoniae that correlates with antibody-mediated enhancement of complement deposition. \u003cem\u003eInfect. Immun.\u003c/em\u003e \u003cstrong\u003e75\u003c/strong\u003e, 5930\u0026ndash;5938 (2007).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Paton, J. C. \u003cem\u003eet al.\u003c/em\u003e Purification and immunogenicity of genetically obtained pneumolysin toxoids and their conjugation to Streptococcus pneumoniae type 19F polysaccharide. \u003cem\u003eInfect. Immun.\u003c/em\u003e \u003cstrong\u003e59\u003c/strong\u003e, 2297\u0026ndash;2304 (1991).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Ferreira, D. M. \u003cem\u003eet al.\u003c/em\u003e DNA vaccines expressing pneumococcal surface protein A (PspA) elicit protection levels comparable to recombinant protein. \u003cem\u003eJ. Med. Microbiol.\u003c/em\u003e \u003cstrong\u003e55\u003c/strong\u003e, 375\u0026ndash;378 (2006).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Lu, Y.-J. \u003cem\u003eet al.\u003c/em\u003e Screening for Th17-Dependent Pneumococcal Vaccine Antigens: Comparison of Murine and Human Cellular Immune Responses. \u003cem\u003eInfect. Immun.\u003c/em\u003e \u003cstrong\u003e86\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Cheliotis, K. S. \u003cem\u003eet al.\u003c/em\u003e Influence of sex, season and environmental air quality on experimental human pneumococcal carriage acquisition: a retrospective cohort analysis. \u003cem\u003eERJ Open Research\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Identifying Novel Pneumococcal Pneumonia Vaccine Candidates - LSTM Online Archive. https://archive.lstmed.ac.uk/21171/.\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Bjornstal, O. \u003cem\u003eet al.\u003c/em\u003e The impact of decreased bead count to determine concentrations of amyloid beta1-42, total-tau, and phosphorylated-tau181 in human cerebrospinal fluid using xMAP technology. \u003cem\u003eJ. Pharm. Sci.\u003c/em\u003e \u003cstrong\u003e100\u003c/strong\u003e, 4655\u0026ndash;4663 (2011).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Unable to find information for 800635.\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Lu, Y.-J. \u003cem\u003eet al.\u003c/em\u003e Interleukin-17A mediates acquired immunity to pneumococcal colonization. \u003cem\u003ePLoS Pathog.\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, e1000159 (2008).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Keech, C. A. \u003cem\u003eet al.\u003c/em\u003e A Phase 1 Randomized, Placebo-controlled, Observer-blinded Trial to Evaluate the Safety and Immunogenicity of Inactivated Streptococcus pneumoniae Whole-cell Vaccine in Adults. \u003cem\u003ePediatr. Infect. Dis. J.\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, 345\u0026ndash;351 (2020).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Mitsi, E. \u003cem\u003eet al.\u003c/em\u003e Nasal Pneumococcal Density Is Associated with Microaspiration and Heightened Human Alveolar Macrophage Responsiveness to Bacterial Pathogens. \u003cem\u003eAm. J. Respir. Crit. Care Med.\u003c/em\u003e \u003cstrong\u003e201\u003c/strong\u003e, 335\u0026ndash;347 (2020).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Breiman, L. Random Forests. \u003cem\u003eSpringer Science and Business Media LLC\u003c/em\u003e \u003cstrong\u003e45\u003c/strong\u003e, 5\u0026ndash;32 (2001).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Oberg, A. L., Kennedy, R. B., Li, P., Ovsyannikova, I. G. \u0026amp; Poland, G. A. Systems biology approaches to new vaccine development. \u003cem\u003eCurr. Opin. Immunol.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 436\u0026ndash;443 (2011).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Ferreira, D. M. \u003cem\u003eet al.\u003c/em\u003e Controlled human infection and rechallenge with Streptococcus pneumoniae reveals the protective efficacy of carriage in healthy adults. \u003cem\u003eAm. J. Respir. Crit. Care Med.\u003c/em\u003e \u003cstrong\u003e187\u003c/strong\u003e, 855\u0026ndash;864 (2013).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Adler, H. \u003cem\u003eet al.\u003c/em\u003e Experimental human pneumococcal colonization in older adults is feasible and safe, not immunogenic. \u003cem\u003eAm. J. Respir. Crit. Care Med.\u003c/em\u003e \u003cstrong\u003e203\u003c/strong\u003e, 604\u0026ndash;613 (2021).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Araujo, A. P. \u003cem\u003eet al.\u003c/em\u003e Serum levels of anti-PspA and anti-PspC IgG decrease with age and do not correlate with susceptibility to experimental human pneumococcal colonization. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, e0247056 (2021).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Chichili, G. R. \u003cem\u003eet al.\u003c/em\u003e Phase 1/2 study of a novel 24-valent pneumococcal vaccine in healthy adults aged 18 to 64 years and in older adults aged 65 to 85 years. \u003cem\u003eVaccine\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 4190\u0026ndash;4198 (2022).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Kaplonek, P. \u003cem\u003eet al.\u003c/em\u003e A semisynthetic glycoconjugate provides expanded cross-serotype protection against Streptococcus pneumoniae. \u003cem\u003eVaccine\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 1038\u0026ndash;1046 (2022).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;McCool, T. L. \u0026amp; Weiser, J. N. Limited role of antibody in clearance of Streptococcus pneumoniae in a murine model of colonization. \u003cem\u003eInfect. Immun.\u003c/em\u003e \u003cstrong\u003e72\u003c/strong\u003e, 5807\u0026ndash;5813 (2004).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Malley, R. \u003cem\u003eet al.\u003c/em\u003e CD4+ T cells mediate antibody-independent acquired immunity to pneumococcal colonization. \u003cem\u003eProc Natl Acad Sci USA\u003c/em\u003e \u003cstrong\u003e102\u003c/strong\u003e, 4848\u0026ndash;4853 (2005).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;van Rossum, A. M. C., Lysenko, E. S. \u0026amp; Weiser, J. N. Host and bacterial factors contributing to the clearance of colonization by Streptococcus pneumoniae in a murine model. \u003cem\u003eInfect. Immun.\u003c/em\u003e \u003cstrong\u003e73\u003c/strong\u003e, 7718\u0026ndash;7726 (2005).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Cohen, J. M. \u003cem\u003eet al.\u003c/em\u003e Protective contributions against invasive Streptococcus pneumoniae pneumonia of antibody and Th17-cell responses to nasopharyngeal colonisation. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, e25558 (2011).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Wilson, R. \u003cem\u003eet al.\u003c/em\u003e Protection against Streptococcus pneumoniae lung infection after nasopharyngeal colonization requires both humoral and cellular immune responses. \u003cem\u003eMucosal Immunol.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 627\u0026ndash;639 (2015).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Wilson, R. \u003cem\u003eet al.\u003c/em\u003e Naturally acquired human immunity to pneumococcus is dependent on antibody to protein antigens. \u003cem\u003ePLoS Pathog.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, e1006137 (2017).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Trzciński, K. \u003cem\u003eet al.\u003c/em\u003e Effect of serotype on pneumococcal competition in a mouse colonization model. \u003cem\u003eMBio\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, e00902-15 (2015).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Moffitt, K. L., Malley, R. \u0026amp; Lu, Y.-J. Identification of protective pneumococcal T(H)17 antigens from the soluble fraction of a killed whole cell vaccine. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, e43445 (2012).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Zhang, Z., Clarke, T. B. \u0026amp; Weiser, J. N. Cellular effectors mediating Th17-dependent clearance of pneumococcal colonization in mice. \u003cem\u003eJ. Clin. Invest.\u003c/em\u003e \u003cstrong\u003e119\u003c/strong\u003e, 1899\u0026ndash;1909 (2009).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Nikolaou, E. \u003cem\u003eet al.\u003c/em\u003e Experimental Human Challenge Defines Distinct Pneumococcal Kinetic Profiles and Mucosal Responses between Colonized and Non-Colonized Adults. \u003cem\u003eMBio\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Carr, M. W., Roth, S. J., Luther, E., Rose, S. S. \u0026amp; Springer, T. A. Monocyte chemoattractant protein 1 acts as a T-lymphocyte chemoattractant. \u003cem\u003eProc Natl Acad Sci USA\u003c/em\u003e \u003cstrong\u003e91\u003c/strong\u003e, 3652\u0026ndash;3656 (1994).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;German, E. L. \u003cem\u003eet al.\u003c/em\u003e Protective effect of PCV vaccine against experimental pneumococcal challenge in adults is primarily mediated by controlling colonisation density. \u003cem\u003eVaccine\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 3953\u0026ndash;3956 (2019).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Carr, O. J. J. \u003cem\u003eet al.\u003c/em\u003e Nasopharyngeal pneumococcal colonization density is associated with severe pneumonia in young children in the lao people\u0026rsquo;s democratic republic. \u003cem\u003eJ. Infect. Dis.\u003c/em\u003e \u003cstrong\u003e225\u003c/strong\u003e, 1266\u0026ndash;1273 (2022).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Khan, M. N., Xu, Q. \u0026amp; Pichichero, M. E. Protection against Streptococcus pneumoniae Invasive Pathogenesis by a Protein-Based Vaccine Is Achieved by Suppression of Nasopharyngeal Bacterial Density during Influenza A Virus Coinfection. \u003cem\u003eInfect. Immun.\u003c/em\u003e \u003cstrong\u003e85\u003c/strong\u003e, (2017).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Trzciński, K. \u003cem\u003eet al.\u003c/em\u003e Protection against nasopharyngeal colonization by Streptococcus pneumoniae is mediated by antigen-specific CD4+ T cells. \u003cem\u003eInfect. Immun.\u003c/em\u003e \u003cstrong\u003e76\u003c/strong\u003e, 2678\u0026ndash;2684 (2008).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Hoe, E. \u003cem\u003eet al.\u003c/em\u003e Reduced IL-17A Secretion Is Associated with High Levels of Pneumococcal Nasopharyngeal Carriage in Fijian Children. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e0129199 (2015).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Vuononvirta, J., Peltola, V., Ilonen, J., Mertsola, J. \u0026amp; He, Q. The Gene Polymorphism of IL-17 G-152A is Associated with Increased Colonization of Streptococcus pneumoniae in Young Finnish Children. \u003cem\u003ePediatr. Infect. Dis. J.\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 928\u0026ndash;932 (2015).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Lu, Y.-J. \u003cem\u003eet al.\u003c/em\u003e GMP-grade pneumococcal whole-cell vaccine injected subcutaneously protects mice from nasopharyngeal colonization and fatal aspiration-sepsis. \u003cem\u003eVaccine\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 7468\u0026ndash;7475 (2010).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Soudja, S. M. \u003cem\u003eet al.\u003c/em\u003e Memory-T-cell-derived interferon-\u0026gamma; instructs potent innate cell activation for protective immunity. \u003cem\u003eImmunity\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 974\u0026ndash;988 (2014).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Mancuso, R. I. \u003cem\u003eet al.\u003c/em\u003e Impaired expression of CXCL5 and matrix metalloproteinases in the lungs of mice with high susceptibility to Streptococcus pneumoniae infection. \u003cem\u003eImmun. Inflamm. Dis.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 128\u0026ndash;142 (2018).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Shenoy, A. T. \u003cem\u003eet al.\u003c/em\u003e Lung CD4+ resident memory T cells remodel epithelial responses to accelerate neutrophil recruitment during pneumonia. \u003cem\u003eMucosal Immunol.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 334\u0026ndash;343 (2020).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Urban, B. C. \u003cem\u003eet al.\u003c/em\u003e Inflammation of the nasal mucosa is associated with susceptibility to experimental pneumococcal challenge in older adults. \u003cem\u003eMucosal Immunol.\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 973\u0026ndash;989 (2024).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Morton, B. \u003cem\u003eet al.\u003c/em\u003e A feasibility study of controlled human infection with Streptococcus pneumoniae in Malawi. \u003cem\u003eEBioMedicine\u003c/em\u003e \u003cstrong\u003e72\u003c/strong\u003e, 103579 (2021).\u003c/li\u003e\n\u003cli\u003e\u0026emsp;\u0026ensp;Shattock, A. J. \u003cem\u003eet al.\u003c/em\u003e Contribution of vaccination to improved survival and health: modelling 50 years of the Expanded Programme on Immunization. \u003cem\u003eLancet\u003c/em\u003e \u003cstrong\u003e403\u003c/strong\u003e, 2307\u0026ndash;2316 (2024). \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7566878/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7566878/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIdentifying conserved, immunogenic proteins that confer protection against \u003cem\u003eStreptococcus pneumoniae\u003c/em\u003ecolonisation could enable development of serotype-independent vaccines.\u003c/p\u003e\n\u003cp\u003eWe analysed baseline samples from 86 healthy adults experimentally challenged with pneumococcal serotypes 6B or 15B to investigate whether immune responses to 75 universally expressed pneumococcal proteins associated with protection against colonisation. We measured serum IgG using a novel Luminex assay and cytokine responses from peripheral blood mononuclear cells following antigen stimulation.\u003c/p\u003e\n\u003cp\u003eNo individual IgG or cytokine marker correlated significantly with protection in univariate analysis. However, machine learning identified IgG responses to PdB, SP1069, and SP0899 as predictive of protection. MCP-1 responses to SP1069 and SP0899, and IL-17 production in response to SP0648-3 also correlated with protection. Elevated baseline IFN-γ, RANTES, and anti-protein IgG correlated with lower colonisation density.\u003c/p\u003e\n\u003cp\u003eWe highlight SP1069 and SP0899 as potential serotype-independent vaccine candidates and demonstrate the utility of machine learning to identify immune correlates of protection.\u003c/p\u003e","manuscriptTitle":"Machine learning-driven identification of serotype-independent pneumococcal vaccine candidates using samples from human infection challenge studies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-14 06:21:05","doi":"10.21203/rs.3.rs-7566878/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c8c6ea32-240e-4b07-97b4-eb749e8dcf16","owner":[],"postedDate":"October 14th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":56101777,"name":"Health sciences/Diseases"},{"id":56101778,"name":"Biological sciences/Immunology"},{"id":56101779,"name":"Biological sciences/Microbiology"}],"tags":[],"updatedAt":"2025-10-21T20:53:16+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-14 06:21:05","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7566878","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7566878","identity":"rs-7566878","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}