Streptococcus pneumoniae (pneumococcus) is an important human pathogen that causes community-acquired pneumonia, pneumonia-derived sepsis (pneumonic sepsis) and associated multi-organ dysfunction, otitis media, and meningitis among young children and the elderly (Olson and Davis, 2020; Gadsby and Musher, 2022; Feldman and Anderson, 2016). Pneumococcal infections are linked to high rates of morbidity and mortality. The pneumococcal capsule is a major virulence factor accounting for more than 100 serotype diversity (Zafar et al., 2017; Paton and Trappetti, 2019; Geno et al., 2015). Vaccination remains the best practice to prevent pneumococcal diseases. However, full coverage of ≥100 serotypes with currently licensed 23-valent capsular polysaccharide vaccine (Pneumovax 23) and conjugated vaccines (PCV7, 10, 15, 20), is not feasible (Feldman and Anderson, 2014; Jaiswal et al., 2014; Mehr and Wood, 2012). The highly recombinogenic nature of pneumococcus has led to the emergence of vaccine escape strains through “capsule switching” to non-vaccine serotypes (Mehr and Wood, 2012; Manna et al., 2022; Feikin et al., 2013). Additionally, the increasing prevalence of antibiotic-resistant pneumococcal strains has impeded the momentum for eradicating pneumococcal diseases (Nikolaou et al., 2020; Li et al., 2023; Johnson et al., 2024), leading to increased healthcare costs.

Another major challenge in combatting pneumococcal diseases is its ability to induce natural competence. Pneumococcus develops a competent state by which it actively takes up exogenous DNA into its cells for genetic transformation (Griffith, 1928; Salvadori et al., 2019; Claverys et al., 2009), as well as regulation of virulence processes (Lin et al., 2016; Zhu et al., 2015; Lau et al., 2001). For decades, the molecular mechanisms of pneumococcal competence development have been intensively investigated using the capsule-deficient strains derived from the serotype 2 strain D39 (Griffith, 1928), including CP1200, CP1250, and R6, as well as its hypercompetent lineages R800 and Rx, under specific in vitro culture conditions (e.g., using the competence-optimized C + Y medium) (Stevens et al., 2011; Gagne et al., 2013; Mirouze et al., 2013; Guiral et al., 2006; Bergé et al., 2017). In contrast, D39 could only enter the competent state in the Todd Hewitt supplemented with yeast extract (THY) broth when exogenously provided with the quorum sensing peptide pheromone called competence stimulating peptide (CSP) (Lin et al., 2020). The competent state develops naturally in response to the threshold accumulation of CSP encoded by the comC gene (Havarstein et al., 1995; Pestova et al., 1996; Håvarstein et al., 1996). The pre-CSP is processed into a mature peptide and secreted through the ComAB transporters. When extracellular CSP reaches a critical concentration, it binds to ComD and activate the two-component histidine kinase-response regulator signal transduction system (TCSTS) ComDE (Martin et al., 2013). The binding of CSP to ComD results in the phosphorylation of ComE. Phosphorylated ComE acts as a transcription factor and binds to com-boxes on the promoter of some com genes to activate their transcription, including comAB, comCDE, and comX1comX2 operons, creating a positive feedback loop which amplifies the competence signal.

ComDE triggers a short transient burst of the competent state that augments the expression of three distinct but temporally overlapping phases of “early,” “late,” and “delayed” competence (com) genes during competence development (Rimini et al., 2000; Peterson et al., 2004). The expression of early com genes comABCDE is pivotal in initiating the transition of pneumococcal cells into the competent state and results in the expression of the alternative σ factor ComX, encoded by the two identical copies of the early com genes comX1 and comX2 (Luo et al., 2003; Luo and Morrison, 2003; Piotrowski et al., 2009; Lee and Morrison, 1999). ComX activates the transcription of late com genes, facilitating processes such as DNA uptake and recombination (Peterson et al., 2004; Luo and Morrison, 2003; Weyder et al., 2018), and enhances virulence mechanisms, partially via fratricide (Havarstein et al., 2006; Steinmoen et al., 2002; Eldholm et al., 2010) mediated release of pneumolysin (Zhu et al., 2015; Brown et al., 2014; Alhamdi et al., 2015; Price and Camilli, 2009; Zafar et al., 2017). These include genes encoding components of the DNA binding type IV pilus (comG operon), the DNA receptor (comEA), the DNA translocase (comFA) (Laurenceau et al., 2015), and DNA processing and recombination proteins (ssbB, dprA, and recA) (Mortier-Barrière et al., 2007; Attaiech et al., 2011), which collectively facilitate DNA transfer across the cellular membrane and its integration into the genome. Although the regulation of delayed com genes is less well understood, they are thought to play a significant role in the fitness of S. pneumoniae during infection and associated with stress response mechanisms (Rimini et al., 2000; Peterson et al., 2004; Dagkessamanskaia et al., 2004).

The spontaneously developed transient competent state observed in vitro for S. pneumoniae lasts for 30–40 min, after which it is rapidly shut off (Rimini et al., 2000; Peterson et al., 2004; Peterson et al., 2000). In a previously published DNA microarray study using S. pneumoniae TIGR4 strain, Peterson et al. (2004) reported that the provision of CSP induces the expression of early com genes rapidly, reaching peak expression between 7.5 and 10 min. A delay of approximately 5 min was observed for the late com genes to reach their maximum expression. Additionally, the mRNA levels for delayed com genes exhibited a continual increase and reached their peak by the first 15–17 min post-CSP treatment. However, in our recent in vivo imaging system (IVIS)-based live imaging study during pneumonic sepsis in CD-1 mice using a D39 reporter strain harboring a firefly luciferase transcriptionally fused to the late competence ssbB gene (D39-ssbB-luc) (Lin and Lau, 2019), D39 was able to enter the competent state naturally without the exogenously provided CSP1 (Lin et al., 2020). The competent state during the acute lung infection was prolonged and persistent and occurred only after 20–24 h post-infection (hpi). This was rapidly followed by the breach of the alveolar-capillary barrier, systemic bacteremia and sepsis. The naturally developed competent state was maintained throughout the septic stage until moribund mice were euthanized (Lin et al., 2020). In our most recent study using a more sensitive RNA-Seq approach during pneumonic sepsis, we found that the expression of the early and late com genes was elevated as early as 12-hpi, and revealed a temporal period of adaptation to the host lung environment due to metal scarcity before the development of natural competence (Oh et al., 2024).

Therefore, in this study, we aim to unravel the differences in competence-responsive genes in vitro versus during pneumonia-derived sepsis. We comprehensively analyzed the existing pneumonia-derived sepsis RNA-Seq data (Oh et al., 2024) along with in vivo gene expression data from a repeated mouse pneumonic sepsis experiment to compare the expression of selective “early,” “late,” and “delayed” com genes against the expression of these genes in vitro during CSP induced competence development. Our analysis clearly illustrates the similarities and differences between the expression of com genes in vitro versus in vivo during pneumonia-derived sepsis.

The competent state in S. pneumoniae is a transient cellular phase that allows the uptake of environmental DNA for genetic recombination and fitness. This state involves the induction of three distinct phases of early, late, and delayed com genes (Lin et al., 2016; Johnston et al., 2014). In this study, we compared the expression profiles of six early, late, and delayed com genes during the in vitro CSP1-induced competent state versus the naturally developed competent state during pneumonic sepsis. Consistent with previous findings in the serotype 4 TIGR4 strain (Griffith, 1928), our in vitro qPCR-based studies in the serotype 2 strain D39 revealed that the expression of early com genes was induced the earliest at approximately 10 min post-CSP1 treatment. Following the activation of early com genes, the late com genes were subsequently expressed at approximately 20 min after induction with CSP1. The expression of the delayed genes was only modestly increased and peaked at 30 min, especially for dnaK and hrcA. During acute lung infection, the expression of both early and late com genes began before 12-hpi and reached their maximum at 24-hpi. However, the expression of early and late com genes was maintained until the mice became moribund and had to be euthanized. Importantly, the peaked expression of late competence genes, such as cibA and cbpD, which encode the allolytic autolysins responsible for pneumolysin release, was followed by the breach of the alveolar-capillary barrier and systemic invasion. A significant divergence in the gene expression profile between in vitro versus pneumonic sepsis was observed in the expression of delayed genes, which were downregulated throughout lung infection from Time 0-hpi to >40-hpi.

The delayed competence genes in S. pneumoniae encode proteins commonly associated with stress responses, including chaperones, proteases, and heat shock proteins (Rimini et al., 2000; Peterson et al., 2004; Dagkessamanskaia et al., 2004). Our in vitro competence induction studies revealed that the expression of delayed genes was modestly elevated, aligning with previous findings in Streptococcus gordonii, a streptococcus closely related taxonomically to S. pneumoniae. In S. gordonii, stress-associated genes were not upregulated within 40 min post-CSP1 treatment, and some delayed genes showed synchronized expression with late genes with a strong signal detected 15 min after CSP1 induction (Vickerman et al., 2007). This observation aligns with our findings on hrcA and htrA expression in S. pneumoniae, which peaked at 20 min post CSP1 induction. One intriguing aspect of our findings is the downregulation of the many delayed genes (clpL, hrcA, dnaK, groL) throughout the entire course of lung infection, which is different than the expression profiles of these genes during competence development in vitro, which were either constitutive or peaked at 20 to 30 min post-exposure to CSP1, and without significant downregulation thereafter (Figure 4). The expression levels of gntR and htrA in the infected lungs were not appreciably changed. The GntR-like transcriptional regulator (SPD_1524) has been shown to sense environmental and nutritional changes by regulating sugar transport and metabolism (Li et al., 2018) and glutamine/glutamate metabolism (Hendriksen et al., 2008), and is induced upon bacitracin and LL-37 exposures (Majchrzykiewicz et al., 2010). An increase in gntR expression at the initial stage could indeed be pivotal for virulence, particularly as bacteria metabolize sugar and amino acid such as glutamine and glutamate as crucial nutrients essential for their survival and pathogenicity (Paixão et al., 2015; Härtel et al., 2011). It is possible that by 12-hpi and beyond, D39 bacteria had adapted to the stressful lung environment to allow the expression of other genes important for pneumococcal pathogenesis. Lastly, the HtrA protease is a pneumococcal CSP1-degrading serine protease regulated by the CiaRH two-component regulatory system responsible for terminating the competent state (Attaiech et al., 2015; Cassone et al., 2012). The htrA gene was most highly expressed during the early stage of infection (12-hpi) followed by a gradual decline during the remaining course of pneumonic sepsis and was unable to shut down the competent state that seemed to be important for pneumococcal pathogenesis that lasted until the infected mice entered the moribund stage (Figure 1).

The observed patterns of competence gene expression in vivo likely reflect complex spatial and temporal dynamics within the bacterial population during infection. It is plausible that competence induction occurs in waves, propagating through different regions of the pneumococcal population asynchronously. This would result in a heterogeneous population where some subpopulations are undergoing competence while others are in the refractory period. The competence regulatory mechanisms elucidated through in vitro studies are likely still operational in the in vivo setting. However, the complex environment encountered during infection may lead to variations in competence induction and shut-off that are not captured by population-level transcriptomics.

Another point that required further clarification is that in both in vitro and in vivo studies, S. pneumoniae was cultured in THY broth. However, the preparation of inoculum for mouse infections was subjected to an additional step involving blood agar plating. This modification was implemented due to the tendency of S. pneumoniae to form long chains, potentially leading to an underestimation of the actual number of bacterial inoculum. To mitigate this inaccuracy, we incorporated a blood agar plating step in our inoculum preparation, as described in Lin et al. (2020) which resulted in formation of mostly diplococci typically seen in clinical settings. The pneumococcus cells were subjected to a 3 rounds of rigorous washing process using sterile saline solution. This procedure effectively removed any residual media components, and the pneumococcus cells were subsequently collected for infection. This approach allows for more accurate quantification and standardization of bacterial cells across experiments, and enhances the reproducibility and reliability of our in vivo infection model. Therefore, it is unlikely that differences in the nutritional aspects of THY versus blood agar caused the differences in competence development between in vivo and in vitro conditions.

Competence development in S. pneumoniae is a complex process highly regulated by multiple factors, including specific host tissue micro-environmental conditions, nutrient availability, and accumulation of CSP. Through normal daily breathing, we inhale millions of particles and microbes. Most of these harmful substances and microbes are trapped by the mucous layer coating upper airways and cleared by mucociliary transport. Microbes that successfully reach the alveolar spaces are deposited on the pulmonary surfactant layer, which initiates complex strategies to counter each other. Similarly, during the initial stages of lung infection, pneumococcus senses changes in host body temperature, pH, ionic strength, and viscosity within the surfactant layer, forcing it to adapt to the host environment to establish infection. In addition, pneumococcal cells would need to overcome various first-line host innate immune defences in the alveolar lining layer such as antimicrobial peptides (MacNair et al., 2024), surfactant proteins and phospholipids (Kuang et al., 2020; Kuang et al., 2011; Tan et al., 2014; Awasthi, 2010; Tan et al., 2015), cell wall-degrading lysozyme (Davis et al., 2008), lactoferrin (Bitsaktsis et al., 2012), secretory phospholipase A2 (Movert et al., 2013), secretory leukocyte protease inhibitor (Abe et al., 1997), as well as phagocytic alveolar macrophages and neutrophils (Koppe et al., 2012) and the complement system (Syed et al., 2020). Pneumococci likely need to overcome these first lines of host defences before establishing successful infection and competence development. Our recent RNA-Seq analysis revealed that when compared against Time 0-hpi (baseline), 20, 42, and 24% of the top 50 most upregulated pneumococcal genes in infected mouse lungs at 12-hpi, 24-hpi and >40-hpi were com genes (Oh et al., 2024), indicating crucial roles played by the competence regulon in initial adaptation to the lung environment (0–12-hpi), followed by lung infection and breaching of the air-blood barrier (12–24-hpi), and maintenance of the septicemic stage (24–40-hpi) (Oh et al., 2024). Nevertheless, early competence genes were the first to be expressed when the pneumococcus cells entered a competent state. As a key regulators of competence, the early genes comCDE play indispensable roles in regulating competence development (Martin et al., 2013), consequently, these genes exhibit drastically higher expression levels in vitro. Our study confirms that these genes exhibit significantly elevated expression levels 10 min after CSP1 treatment in vitro (Figure 2A). Specifically, comC shows peak expression at 10 min with a 9.06-fold increase, while comD and comE display even more dramatic upregulation with 122-fold and 38.8-fold increases, respectively. These results underscore the rapid and robust transcriptional response of the ComCDE system to CSP1 induction. Unexpectedly, the expression for comD and comE were barely changed at 12-hpi but upregulated at 24-hpi by 7.5-fold and continuously maintained much above the baseline level even at the moribund state, as reflected by persistent activation of the competent state based on our RNA-Seq data.

All the in vitro experiments in this study were performed in THY broth using the S. pneumoniae D39 strain. D39 does not enter the competent state naturally in THY broth without the provision of CSP1. Our lab has previously confirmed this using the reporter strain D39-ssbB-luc and tested its bioluminescence activity, finding that it failed to enter the competent state naturally in the THY medium. After the addition of CSP1, we were able to monitor bioluminescence activity 30 min (indicating expression of ssbB late gene) after the provision of CSP1 (Lin et al., 2020). During acute pneumonia infection in mice, the competent state developed naturally without the exogenous provision CSP1. However, the conditions that are required for the competence development is unknown. During the lung infection, pneumococci compromise the alveolar-capillary barrier, allowing pneumococcal cells to enter the systemic circulation. This breach not only facilitates the spread of infection but also creates conditions favourable for competence development. The lung environment typically low in certain nutrients, especially the divalent cations. Specifically, the damaged alveolar-capillary barrier allows the influx of various substances, including divalent cations such as calcium, magnesium, and manganese, which are usually scarce in lung tissue. Pneumococcal cells can then uptake these nutrients, which have been shown to be important for the induction of competence (Oh et al., 2024). This nutrient availability likely contributes to the natural induction of competence in vivo, even in the absence of exogenous CSP1. As we have reported previously, the synthetic C + Y medium consists various divalent metal ions which support spontaneous induction of competence in S. pneumoniae CP1250 strains at competence permissive pH (Oh et al., 2024).

In addition, our previous research has also demonstrated that the exogenous provision of CSP1 has limited effect on the development of natural competence during pneumonia-derived sepsis. Surprisingly, exogenously provided CSP1 failed to modulate the onset kinetics of competence development in vivo. Interestingly, our previous findings revealed that competent D39 bacteria primarily spread the competence signal through direct cell-to-cell contact. This mechanism diverges from the classical quorum-sensing model, which relies heavily on the accumulation of extracellular CSP (Lin et al., 2020).

RNA-Seq is an extremely sensitive technique widely used to provide a comprehensive transcriptional profile, capable of detecting even low-abundance transcripts. Our previous RNA-Seq analysis revealed a significant upregulation of competence genes in S. pneumoniae as early as 12-hpi during pneumonic sepsis in mice. This finding suggests that the pneumococcus competence regulon plays a crucial role in pneumococci adaptation to the lung microenvironment during infection. However, when we monitored the progression of infection using the IVIS system, we did not detect a competent state until approximately 24-hpi (e.g., compare Figure 1A vs. Figure 1C). Interestingly, both early and late competence genes reached their peak expression at 24-hpi, marking the peak competent state where the late genes tend to express in relatively rapid succession following the expression of early genes. The expression of the ssbB gene (luciferase signals) was detectable only at 24-hpi, perhaps reaching an expression threshold where could be detected by the less sensitive IVIS-based live imaging. Interestingly, the IVIS data shows the strongest ssbB bioluminescence signal at 40-hpi, which shows a discrepancy compared to the peaked ssbB expression at 24-hpi by RNA-Seq. The stronger IVIS signal observed at 40-hpi can be attributed to a cascade of events within the host. As the alveolar-capillary barrier was breached, bacteria gained access to the nutrient-rich bloodstream, allowing them to spread systemically rather than remain confined to the lungs. This widespread dissemination throughout the host body resulted in a more intense, systemic bioluminescence signal output at 40-hpi. Furthermore, the nutrient-rich blood created conditions more favorable for pneumococcal competence development, leading to an increase in competence signal. Another plausible explanation is that, the IVIS imaging and RNA-Seq employ different technologies to assess gene expression, which can lead to apparent discrepancies in results. IVIS detects bioluminescence signals from reporter proteins expressed by S. pneumoniae, while RNA-Seq directly quantifies mRNA levels. This fundamental difference can result in temporal variations in peak expression detection. The observed delayed peak in the IVIS signal at 40-hpi likely reflects post-transcriptional regulation of SsbB. Even after mRNA levels have decreased, the reporter protein can continue to accumulate, resulting in a sustained or increased bioluminescence signal. Therefore, mRNA transcriptome levels do not always directly correlate with protein levels. Also, the half-life mRNA molecules is generally much shorter than proteins. This discrepancy is further supported by findings from Cheng et al. (2016) who reported that while mRNA concentrations declined, protein concentrations continued to increase.

It is worth noting that, the initial competence induction in the context of pneumococcal cells relies more on the adaptation to the lung environment rather than the number of pneumococcal cells present (Lin et al., 2020; Weiser et al., 2018). Lin et al. (2020) demonstrated that regardless of whether mice were inoculated with 5 × 10⁷, 5 × 10⁶ or 1 × 10⁶ CFU, competence induction was consistently detected at approximately 24-hpi in all cases. This observation suggests that the ability of bacteria to sense and respond to the host environment, rather than their initial numbers, is the key trigger for competence induction. The consistency in timing across different inoculum sizes underscores the importance of bacterial adaptation to the host milieu in regulating this crucial physiological process. In addition, the authors also revealed that the lung bacterial burden had minimal impact on competence induction kinetics. The authors further explored this by infecting mice with 5 × 10⁷ CFU of D39-ssbB-luc and performed the bacterial burden at 0, 12, 24, and >40-hpi. They found no significant difference in pneumococcal lung burden at 12 and 24-hpi compared to 0-hpi, with a significant increase to >10⁸ CFU only after 40-hpi. Notably, initial competence induction at approximately 24-hpi appeared to depend more on adaptation to the lung environment rather than bacterial cell numbers (Lin et al., 2020).

Our previous study demonstrated a high degree of correlation between RNA-Seq and qPCR validation results for 10 representative competence-specific genes, using the same RNA samples. This analysis yielded a Pearson correlation coefficient r of 0.8525 (R² = 0.7267) (Oh et al., 2024). In the current study, we expanded our analysis to 18 pneumococcal genes to further assess the reproducibility of the RNA-Seq data. Importantly, we utilized RNA extracted from a new independently performed pneumonic sepsis experiment. This expanded validation yielded a Pearson correlation coefficient r of 0.7775, further confirming the reliability of our RNA-Seq results. Minhas et al. (2020) employed the same approach to validate the reproducibility of their dual RNA-Seq data for both murine and pneumococcal genes. They used RNA isolated from two sources, the same isolated RNA for dual RNA-Seq and RNA from a repeated experiment. When using RNA from the same samples for dual RNA-Seq, they obtained a Pearson r of 0.8565 (R² = 0.7338), while using isolated RNA from a repeated experiment resulted in Pearson r of 0.8751 (R² = 0.7658) (Minhas et al., 2020). Overall, these results highlight the robustness and reproducibility of RNA-Seq data across different experimental conditions and biological systems. The high correlation between RNA-Seq and qPCR validation results, as well as the strong reproducibility demonstrated in both our studies and in Minhas et al. (2020) suggest that RNA-Seq provides reliable and consistent gene expression data. Importantly, this high level of reproducibility indicates that repeating RNA-Seq experiments may not be necessary in many cases, which is particularly advantageous given the considerable cost associated with RNA-Seq. This finding not only validates the reliability of RNA-Seq as a tool for gene expression analysis but also highlights its cost-effectiveness in experimental design.

In summary, a more comprehensive comparison of the similarities and differences of the gene expression patterns during in vitro and in vivo competence development may reveal new aspects of uncharacterized pneumococcal competence regulation. These new insights could contribute to the development of innovative strategies for controlling the burden of pneumococcal diseases by inhibiting the spread of virulence and antibiotic resistance genes mediated by the competence regulon.