Many encapsulated bacteria use capsules to cause invasive diseases. However, it remains largely unknown how the capsules enhance bacterial virulence under in vivo infection conditions. Here we show that the capsules primarily target the liver to enhance bacterial survival at the onset of blood-borne infections. In a mouse sepsis model, the capsules enabled human pathogens Streptococcus pneumoniae and Escherichia coli to circumvent the recognition of liver-resident macrophage Kupffer cells (KCs) in a capsular serotype-dependent manner. In contrast to effective capture of acapsular bacteria by KCs, the encapsulated bacteria are partially (low-virulence types) or completely (high-virulence types) “untouchable” for KCs. We finally identified the asialoglycoprotein receptor (ASGR) as the first known capsule receptor on KCs to recognize the low-virulence serotype-7F and -14 pneumococcal capsules. Our data identify the molecular interplay between the capsules and KCs as a master controller of the fate and virulence of encapsulated bacteria, and suggest that the interplay is targetable for therapeutic control of septic infections.
Introduction
Sepsis is a common and often lethal syndrome with acute organ dysfunction (Rudd et al., 2020). The sepsis pathogenesis is a complex process, from overgrowth of invading pathogens in the blood circulation to dysregulated immune responses and collateral organ injuries (Munford and Suffredini, 2014). Encapsulated bacteria are the most common sepsis pathogens in humans and domestic animals (Wen and Zhang, 2015). Capsules are the outermost structures of many bacteria and certain fungi and are known to be essential for the in vivo survival and virulence of many bacterial pathogens. The importance of the capsule in bacterial virulence was elegantly used by Avery and his colleagues to demonstrate DNA as the genetic material (Avery et al., 1944). The significance of the capsules is also manifested by successful prevention of several human diseases caused by encapsulated bacteria using capsular polysaccharide (CPS)–based vaccines. Bacterial capsules have also been therapeutically targeted to treat bacterial infections with anti-capsule antibodies (Chang et al., 2002), phage-derived capsule depolymerases (Pertics et al., 2021), and capsule biosynthesis inhibiting small molecules (Arshad et al., 2016).
Virtually all known capsules are composed of repeating polysaccharide units, which are typically produced by single centralized cps loci (Moradali and Rehm, 2020; Whitfield et al., 2020). The structures of CPSs differ greatly between bacterial species. Many bacteria also produce intraspecies structural variants of CPSs by variations in the cps genes, which is the basis for conventional serotyping of bacterial strains (Whitfield et al., 2020). This is exemplified by identification of 100 and 83 capsular serotypes in Streptococcus pneumoniae and Escherichia coli, respectively (Ganaie et al., 2020; Wen and Zhang, 2015). Structural variations in capsules are recognized as a mechanism for encapsulated bacteria to evade serotype-specific antibodies, which underlies the pneumococcal serotype replacement after vaccination with pneumococcal conjugate vaccines (Lo et al., 2019). Certain capsule types are more associated with the virulence traits of encapsulated bacteria than others in septic infections. This is manifested by the dominance of the low-numbered pneumococcal serotypes/serogroups in fatal pneumonia in pre-antibiotic era (e.g., serotypes 1–8; White, 1938). Similarly, K1 is the most frequently found capsule type of E. coli in neonatal meningitis patients (Robbins et al., 1974). However, it remains unclear how the capsule variants precisely impact bacterial virulence during septic infections.
Capsules have been designated as “slippery,” or antiphagocytic, for decades because of the resistance of encapsulated bacteria to phagocytic killing of host defense (Comstock and Kasper, 2006; Wen and Zhang, 2015). Certain physical properties of the capsules (e.g., hyperviscosity and negative charge) hinder the recognition and binding of phagocytes to encapsulated cells (Brown and Gresham, 2012; Nahm and Katz, 2012; Taylor and Roberts, 2002). Capsules interfere with the complement-mediated opsonophagocytosis by masking the cell wall peptidoglycan and membrane-associated proteins (de Cordoba et al., 1983; Pangburn and Muller-Eberhard, 1978; Wilkinson et al., 1979). Certain bacteria evade host adaptive immunity by producing capsules structurally similar to host glycans, a phenomenon known as molecular mimicry (Finne, 1982; Troy, 1992; Wessels, 2006). A recent study shows that the capsule promotes intracellular survival and translocation of S. pneumoniae by enhancing bacterial resistance to oxidative killing (Brissac et al., 2021). However, it remains unknown what immune organs and cells are primarily targeted by bacterial capsules for immune evasion during septic infections.
In this study, we characterized the fates of blood-borne S. pneumoniae and E. coli, two common encapsulated sepsis pathogens, in a mouse sepsis model. The data uncovered the liver and its resident macrophage Kupffer cells (KCs) as the major organ and cells targeted by encapsulated bacteria to circumvent the most potent antibacterial machinery of the mammalian hosts against blood-borne microbes. Using isogenic bacteria producing different types of capsules, we further stratified the disease potential of selected capsule types of S. pneumoniae and E. coli into low- and high-virulence categories (LV and HV, respectively). The virulence phenotypes of individual capsule types were found to be predominantly determined by the natures of receptor–ligand interactions between the CPSs and KCs during the early phase of septic infections.
Results
Capsular serotypes define pneumococcal virulence
Previous studies have documented significant differences in virulence among pneumococcal strains/serotypes (Briles et al., 1992; Sjostrom et al., 2006; White, 1938). To understand the molecular basis of this heterogeneity, we first compared the virulence levels of 144 encapsulated strains representing 34 serotypes in a murine septic infection model (Fig. 1 A). This revealed remarkable differences in virulence as characterized by mortality rate in the 7-d observation period. We empirically used 50% mortality as a cutoff value to divide these strains into the categories of HV (48 strains) and LV (96 strains; Fig. 1 B and Table S1). The HV and LV mortality phenotypes were also reflected by bacteremia level (Fig. 1, B and C) and survival time (Fig. 1 D). The mice infected with the HV strains typically progressed to more severe bacteremia at 12 and 24 h, as high as 109 CFU/ml blood. In contrast, infection with the LV strains displayed relatively low levels of bacteremia. Accordingly, the mice infected with the HV strains showed much shorter survival time than LV counterparts. Further investigation with selected HV and LV strains, as well as the acapsular mutant R6, revealed remarkably large margins in virulence among pneumococcal strains. The HV strain TIGR4 (type 4) and TH870 (type 6A) were still fully virulent at an extremely low inoculum of 100 CFU (Fig. S1 A); the LV strains ST556 (type 19F) and TH2919 (type 6B) became fully virulent when infection dose was increased to 108 CFU (Fig. S1 B). However, R6 remained avirulent even at the highest infection dose (Fig. S1 B). This result showed a wide range of disease potential for encapsulated pneumococci.
Further analysis revealed a clear correlation between virulence and capsular serotype (Fig. 1 E). While all tested strains of eight serotypes fell into the HV category (1, 2, 3, 4, 5, 8, 22F, and 38), those of 23 other serotypes uniformly showed the LV phenotype. Only the strains of serotypes 6A, 6B, and 6C exhibited both LV and HV phenotypes. To ascertain this relationship, we constructed isogenic strains expressing different capsular serotypes by genetic replacement of the cps gene cluster (Fig. 2 A). Although the cps deletion mutant of the HV strain TH870 (serotype 6A) became avirulent in the murine sepsis model, complementation with the heterologous cps gene clusters of serotype 1, 2, 3, 4, 5, 6A, or 8 conferred the HV virulence phenotype to the mutant with 100% mortality and persistent bacteremia (Fig. 2 B, red lines). In sharp contrast, the counterpart carrying the cps locus of serotype 6B, 7F, 9V, 14, 18C, 19F, 19A, or 23F completely abolished the original HV phenotype of TH870. All of the mice infected with these strains survived with barely detectable bacteremia (Fig. 2 B, green lines). The role of capsule serotypes in shaping pneumococcal virulence was also demonstrated in capsule-switched derivatives of strains TH2919 (serotype 6B; Fig. 2 C), TH197 (serotype 6A; Fig. S1 C), and TH2912 (serotype 14; Fig. S1 D). One exception of this causal relationship between capsule serotype and virulence phenotype was observed with a derivative of serotype 3 strain TH2891. Expressing the serotype 6A cps genes (an HV capsule type) in TH2891 resulted in an unexpected LV phenotype (Fig. S1 E). This outlying effect may be caused by functional mismatch between the cps genes and the genetic background of the recipient strains, as reported previously (Wen et al., 2016).
Finally, we characterized the virulence of isogenic type 6A and 6B strains that differed only in a single nucleotide at position 584 of wciP, encoding the rhamnosyl transferase for biosynthesis of the serogroup 6 capsules (Mavroidi et al., 2004). The type 6A and 6B capsules consist of identical tetrasaccharide repeating units with only a different linkage between the rhamnose and ribitol-5-phosphate residues (α1–3 in 6A, α1–4 in 6B; Han et al., 2012; Fig. 2 D). A G584 (serine195 in 6A)-to-A584 (asparagine195 in 6B) polymorphism in wciP defines the rhamnosyl-ribitol-5-P linkage and thereby the structural difference between the two capsules (Mavroidi et al., 2004). While a G-to-A mutation at position 584 of wciP in TH8706A resulted in an LV phenotype, an opposite (A-to-G) mutation in the corresponding site of TH29196B converted the otherwise LV strain to HV (Fig. 2 E).
Because invasive pneumococcal diseases are naturally developed from pneumonia (Musher and Thorner, 2014), we investigated the impact of capsule types on pneumococcal virulence phenotypes in respiratory infection model. In line with their virulence phenotypes in septic infection, the selected HV (3, 6A, and 8) and LV (14, 19F, and 23F) serotypes displayed the same patterns of virulence traits in mice infected by intratracheal (i.t.) inoculation. Whereas airway infection of serotypes 3, 6A, and 8 led to 100% mortality (Fig. 2 F), with gradually worsening bacteremia as early as 12 h after infection (Fig. 2 G), all of the mice infected with serotypes 14, 19F, and 23F survived with no or marginal bacteremia (Fig. 2, F and G). Together with the clinical surveys (Brueggemann et al., 2004; Sjostrom et al., 2006; Weinberger et al., 2010; White, 1938), these data demonstrate a decisive role of capsule structures in defining pneumococcal virulence.
HV capsules enable pneumococci to bypass hepatic capture
To determine how the capsules enhance bacterial survival during septic infection, we compared the bacterial dynamics in the blood and major organs in mice after i.v. infection with isogenic derivatives of encapsulated TH870 (serotype 6A; Fig. 3 A). In keeping with the importance of the capsule in pneumococcal virulence, infection with TH870 (WT) led to 100% mortality, but all of the mice infected with acapsular mutant (∆cps) survived (Fig. S2 A). Accordingly, the ∆cps mutant became undetectable in the blood circulation within the first 30 min, whereas the parental bacteria were abundantly present (Fig. 3 B). To characterize the fates of the ∆cps mutants disappearing from the circulation, we determined bacterial distribution in five major organs at various time points after i.v. inoculation. The vast majority of the ∆cps mutants (77.3%) were found to be already trapped in the liver as early as 5 min, whereas only a small fraction of the encapsulated counterparts was detected in the liver (21.9%; Fig. 3 C). At that time, the liver trapped 4.7-fold more ∆cps mutants than the WT strain (Fig. 3 D). The spleen housed the second highest proportions of bacteria for both TH870 (14.4%) and its acapsular mutant (15.3%); heart, kidney, and lung contained only residual levels of bacteria (Fig. 3 C). These data reveal that the capsule is extremely important for bacterial survival in the bloodstream by blocking hepatic capture at the very beginning of septic infection.
To define how the capsule type shapes pneumococcal virulence, we characterized the bacteremia kinetics of isogenic LV (serotypes 14, 19F, and 23F) and HV (serotypes 3, 6A, and 8) derivatives of TH870. The virulence phenotypes of isogenic capsule serotype switched strains obtained by i.v. inoculation fully recapitulated those by the i.p. infection route (Fig. 2). All of the mice infected with the HV counterparts succumbed to challenge, whereas those with the LV strains survived (Fig. S2 B). Accordingly, the HV counterparts were temporally reduced in the first 12 h, but rebounded later. In contrast, the LV serotypes were virtually undetectable in the bloodstream in the first 12 h and remained undetectable ever since (Fig. S2 C). Similar to the rapid clearance of the acapsular mutant (Fig. 3B), the three LV serotypes became barely detectable in the bloodstream in the first 10–20 min after i.v. infection (Fig. 3 E). By comparison, the HV counterparts were abundantly present at 30 min. 5 min after i.v. inoculation, the vast majority of the LV pneumococci were found in the liver (type 14, 98.7%; 19F, 81.5%; and 23F, 74.8%); whereas the majority of the HV bacteria were still circulating in the bloodstream with only small fractions in the liver (type 3, 11.2%; 6A, 21.9%; and 8, 16.4%; Fig. 3, F and G). This serotype-dependent sequestration of pneumococci in the liver still operated at 10 and 30 min (Fig. S2 D). Besides the liver, the spleen, but not the heart, kidneys, or lungs, contained small fractions of bacteria for serotypes 3 (10.5%), 6A (12.4%), 8 (11.6%), 19F (13.1%), and 23F (17.2%; Fig. 3 F). This result indicates that the capsule serotype–dependent virulence phenotypes in septic infection are decided by differential trapping of blood-borne bacteria in the liver at a very early phase.
To determine the fate of the LV pneumococci captured in the liver, we compared the total viable bacteria of six representative serotypes at various time points after i.v. inoculation. The results showed gradual but effective elimination of the LV serotypes in the first 30 min. Despite rapid deposition of the LV bacteria in the liver, the total bacteria of serotypes 14, 19F, and 23F recovered from the blood and five organs were similar to the inocula for all isogenic strains at 5 min (Fig. 3 H), indicating that the bacteria captured in the liver were not immediately killed. However, the CFU in the livers and other organs of mice infected by the LV bacteria were significantly diminished at later time points (Fig. 3, I and J). In particular, at 30 min, only minor fractions of the originally inoculated serotype-14 (4.7%), -19F (12.8%), and -23F (12.3%) bacteria were detected (Fig. 3 J). In sharp contrast, high proportions of the inocula were still present in the mice infected by serotypes 3 (91.2%), 6A (66.8%), and 8 (62.9%) at 30 min (Fig. 3 J). These data demonstrate that the liver effectively captures and kills the LV encapsulated pneumococci but not HV counterparts during the early phase of septic infection.
Because the spleen was the organ with the second-highest proportions of pneumococci (Fig. 3, C and F), we evaluated the contribution of the organ to the early clearance of blood-borne bacteria using splenectomized (SPX) mice. Surgical removal of the spleen did not impair the clearance of isogenic serotype-8 (HV) and -14 (LV) pneumococci by the liver in the first 30 min of infection (Fig. 3 K). In fact, the liver of the SPX mice contained virtually all viable bacteria of TH87014 at 30 min, which appeared to be a compensatory response of the liver in the absence of the spleen. The asplenic mice showed a relatively higher level of serotype-8 pneumococci in the circulation and liver than sham-operated mice (Fig. 3 L). This observation indicates that the spleen is able to trap a low level of HV pneumococci that are poorly caught by the liver, which agrees with previous findings that the spleen is important for combating S. pneumoniae and other encapsulated bacteria (Deniset et al., 2017; Theilacker et al., 2016). Together, these data demonstrate that the liver but not spleen is the dominant organ responsible for rapid elimination of LV serotypes at the onset of septic infection.
We finally assessed the role of the liver in the clearance of blood-borne bacteria derived from pneumococcal pneumonia by i.t. instillation of 107 CFU isogenic HV (type 8) or LV (type 14) pneumococci. Both the strains caused severe bacterial burden in the lungs at 12 h (Fig. 3 M). While the mice infected with serotype-8 pneumococci showed bacteria in the blood, liver, and spleen, bacteria were detected in the liver and spleen, but not in blood, of mice i.t. infected with serotype-14 bacteria. The disseminating serotype-8 bacteria were predominantly found in the blood circulation, with relatively small fractions in the liver and spleen. In sharp contrast, the vast majority of disseminating serotype-14 pneumococci were trapped in the liver, making the blood sterile (Fig. 3 N). This finding indicates that the liver captures pneumococci disseminating from the lung to the blood circulation in the same serotype-dependent manner as shown in the septic infection model.
LV serotypes evade hepatic capture at high infection doses
It is intriguing that a number of LV capsule types identified in mouse models are included in the current pneumococcal vaccines, since they are associated with invasive infections of humans, such as 6B, 9V, 14, 18C, 19A, 19F, and 23F (Feldman and Anderson, 2020). We reasoned that these LV serotypes are superior to acapsular strains in the escape of hepatic capture at relatively high levels of bacteremia. We thus compared the clearance kinetics between isogenic serotype 14 and acapsular derivatives of TH870 in mice infected i.v. with 108 CFU. This trial showed that the encapsulated LV pneumococci are much more capable of surviving in the blood circulation and evading hepatic capture than the acapsular mutant. While the Δcps strain was almost fully cleared from the circulation in the first 30 min, TH87014 was abundantly present in the bloodstream (144-fold higher than Δcps; Fig. 4 A). In keeping with the relatively poorer hepatic capture of TH87014 from the circulation (Fig. 4 B), 20% of its inoculum was still detectable in the host at 30 min, compared with only 5% of the Δcps counterpart (Fig. 4 C). This observation is in line with 50% lethality of asialoglycoprotein receptor (ASGR) deficient (ASGR1−/−) mice when this dose of TH87014 was used (see Fig. 8). This result indicates that the encapsulated LV capsules are able to evade hepatic capture when invading bacteria reach certain levels in the circulation.
To assess the functional range of the capsule-mediated immune evasion, we compared the clearance rates of isogenic serotype-3 and -14 strains from the circulation of mice at various infection doses. The LV pneumococci were fully eradicated from the circulation within 30 min at infection doses as high as 107 CFU (Fig. 4 D). In contrast, serotype-3 pneumococci were stably circulated at all the infection doses (Fig. 4 E). In particular, the HV bacteria were still detectable in the bloodstream even with an inoculum of 103 CFU. The striking difference between the LV and HV capsule types in immune evasion is also reflected by their 50% clearance times (CT50, the time for clearing 50% inoculum from the bloodstream; Fig. 4 F). The intrinsic differences among capsule types in escaping hepatic capture allowed us to rank 15 selected serotypes based on CT50 levels of isogenic strains. This analysis places serotypes 3 and 14 at the extreme ends of this immune evasion spectrum, with CT50 values of 0.4 and 24.3 min, respectively (Fig. 4 G). Together, these results demonstrate that the serotype-specific potency in evading hepatic capture defines the disease potential of pneumococcal capsule variants in septic infection.
Phagocytic function of KC is differentially impaired by pneumococcal capsules
To identify the immune cells in the liver whose function was compromised by the capsules, we assessed the contribution of major professional phagocytes to hepatic clearance of acapsular ∆cps by selective depletion (Fig. 5 A; and Fig. S2, E and F). Although neutrophils are reported to be important for eliminating blood-borne microbes (Derby and Rogers, 1961; Munford and Suffredini, 2014; Rogers, 1956), depleting neutrophils with the 1A8 antibody did not affect the clearance rate (Fig. 5, B and C), hepatic capture (Fig. 5 D), and bacterial killing (Fig. 5 E) of ∆cps in the first 30 min. Treatment with the Gr1 antibody (depleting monocytes and neutrophils) also had little impact on the ∆cps clearance (Fig. 5, B–E). In contrast, exhausting liver-resident macrophage KCs led to a dramatic reduction in the clearance of acapsular pneumococci. Treating the Clec4f-DTR mice (Scott et al., 2016) with diphtheria toxin (DT) increased bacteremia levels by 42.9-, 72.8-, 599.2-, and 215.4-fold at 5, 10, 20, and 30 min (Fig. 5 B). CT50 was elongated from 0.53 min in control mice to 3.1 min in DT-treated mice (Fig. 5 C). The KC-depleted mice were severely impaired in shuttling acapsular pneumococci from the bloodstream to the liver (Fig. 5 D). While the livers of control mice carried 70.6% of the inoculum 10 min after infection, this value was reduced to only 9.9% in DT-treated mice. Depletion of KCs also led to significant impairment of bacterial killing: >30% of the inoculum was detected in the host at 30 min compared with <10% in the control (Fig. 5 E). The importance of KCs in the capture of acapsular bacteria was also demonstrated by intravital microscopy (IVM) imaging of liver sinusoids. Consistent with rapid capture of acapsular bacteria in the liver (Fig. 3 E), the ∆cps mutants were abundantly docked on KCs immediately after i.v. inoculation, while the parental bacteria mostly moved through the sinusoid vasculatures (Fig. 5 F and Video 1). The ∆cps bacteria were no longer detained in the liver sinusoids of the DT-treated mice owing to the lack of KCs (Fig. 5 G and Video 2). These results strongly suggest that the liver-resident macrophage KCs are responsible for effective capture and killing of the unencapsulated bacteria.
We further investigated the impact of different capsular types on the interplay of pneumococci and liver KCs. As visualized by IVM, the isogenic LV strains (types 14, 19F, and 23F) were rapidly and abundantly docked on KCs after i.v. inoculation, while the HV counterparts (types 3, 6A, and 8) mostly moved through the sinusoid vasculatures (Fig. 6 A; Fig. S2 G; and Video 3). As exemplified with serotype 14 pneumococci, each KC could bind as many as 20 LV bacterial cells compared with ≤1 HV bacterium 20 min after inoculation (Fig. 6 A, left panel). Not surprisingly, the DT-treated Clec4f-DTR mice were severely impaired in shuttling the LV pneumococci of serotypes 14, 19F, and 23F from the bloodstream to the liver (Fig. 6, B–D; Fig. S2 H; and Video 4). Selective detection of intramacrophage bacteria by pHrodo labeling revealed that 29.5% of the type 14 pneumococci were engulfed by KCs as early as 20 min (Fig. 6 E), indicating the bactericidal role of KCs by phagocytic killing. Functionally, the absence of KCs made the host highly susceptible to septic infection with LV type 14 pneumococci, as reflected by sustained bacteremia (Fig. S2 I) and 100% mortality of the KC-depleted mice (Fig. 6 F), akin to the HV-bacterial infection in normal mice (Fig. S2, B and C). These results thus demonstrate a pivotal role of KC-mediated early clearance for the host defense against septic infection of the LV capsule types.
Finally, we verified binding interactions between primary KCs and pneumococci. The freshly isolated KCs from mouse liver showed significant adherence to isogenic LV pneumococci of serotypes 14, 19F, and 23F but were poorly adherent to the HV counterparts of serotypes 3, 6A, and 8, even with opsonization by normal mouse serum (Fig. 6 G). In a similar manner, the primary human KCs caught 20–30% of the LV pneumococci in the ex vivo system, whereas binding to HV pneumococci was significantly lower (Fig. S2 J). Together, these experiments demonstrate that KCs are primarily responsible for the capture of LV pneumococci in the liver during septic infection; in contrast, the HV capsule variants inherently bypass the surveillance by KCs.
LV capsules are the ligands for KC recognition
Whereas KCs are known to express an array of pattern recognition receptors (PRRs) for clearance of blood-borne pathogens (Jenne and Kubes, 2013), no PRRs on KCs are known to recognize bacterial capsules. Rapid clearance of encapsulated LV pneumococci by KCs suggested that the CPSs of the LV pneumococci are potential ligands for host receptors on KCs. We tested this possibility by pretreatment with free CPSs of three representative LV serotypes (types 14, 19F, and 23F) before bacterial inoculation (Fig. 7 A). Intravenous injection of type 14 CPS (CPS14) immediately before inoculation of TH87014 expressing a homologous serotype 14 capsule resulted in dose-dependent reduction of bacterial clearance (Fig. 7 B, left panel). For instance, CT50 was elongated to 4.3 min in the mice treated with 400 μg of CPS14 from 0.35 min in the control (Fig. 7 B, right panel). The delayed clearance accompanied a dose-dependent decline in hepatic capture (Fig. 7 C). Ultimately, pretreatment with higher doses of CPS14 (e.g., 400 μg) made 56.1% of the inoculum viable at 30 min compared with only 2% in the control (Fig. 7 D). Significant inhibition against clearance of homologous serotypes was also achieved with CPS19F and CPS23F (Fig. 7 E; and Fig. S3, A and B). The blocking effect of the LV CPSs also accompanied significant reduction in KC-mediated bacterial capture in the liver sinusoids (Fig. 7 F and Video 5). However, similar treatment with CPSs of HV serotypes 3, 6A, and 8 did not yield any impact on the clearance of the homologous pneumococci (Fig. 7 E; and Fig. S3, C and D). The poor recognition of HV CPSs by KCs was further manifested by the observation that pretreatment of mice with different amounts of purified CPS3 did not affect the clearance of a sublethal dose of serotype-3 pneumococci (105 CFU; Fig. S3, E and F). This information strongly suggests that KCs capture the LV but not HV pneumococci through molecular recognition of CPS structures.
We next assessed the ligand specificity in KC’s recognition of the LV capsules by testing the impact of free CPSs on the clearance of heterologous serotypes. The clearance of TH87014 was blocked by pretreatment with CPS14 only, but not with any CPS of seven heterologous LV serotypes (6B, 7F, 9V, 18C, 19A, 19F, and 23F; Fig. 7 G and Fig. S3 G). The clearance of isogenic serotype 19F (Fig. 7 H and Fig. S3 H) and 23F (Fig. 7 I and Fig. S3 I) pneumococci was similarly attenuated in a serotype-specific manner by CPS19F and CPS23F, respectively, but not by CPSs of heterologous serotypes. The serotype-specific blocking indicates that the LV capsules are recognized by specific receptors on KCs. The null blocking of the HV CPSs further supports the concept that the HV capsules are poorly recognized by KCs.
The role of the complement system in the clearance of LV serotypes was investigated using mice lacking complement protein C3 or complement receptor of Ig superfamily (CRIg). Activated C3 promotes opsonophagocytic killing of pneumococci (Hyams et al., 2011), and CRIg is the major complement receptor on KCs, with significant contribution to bacterial capture (Helmy et al., 2006). As shown in Fig. 7 J, the clearance of serotype-19F and -23F pneumococci was dramatically impaired in the C3-deficient mice, but serotype-14 bacteria were still effectively eliminated from the blood of the C3−/− mice, indicating that C3 is necessary for KC recognition of serotype-19F and -23F but not -14 pneumococci. In keeping with this finding, CRIg−/− mice showed significantly slower clearance of serotype-19F and -23F pneumococci, particularly at 10 min (Fig. 7 K). In contrast, serotype-14 bacteria were removed from the bloodstream at a similar pace in both WT and CRIg−/− mice. These results show that the complement system is involved in the KC capture of certain but not all LV pneumococcal serotypes.
ASGR on KCs recognizes serotype-7F and -14 pneumococcal capsules
Because hepatic capture of serotype-14 pneumococci is independent of the complement system (Fig. 7 J), we sought to identify the CPS14 receptors on KCs by affinity pulldown of proteins from murine nonparenchymal cells (NPCs) with CPS14-coated beads (Fig. 8 A). The CPS8-conjugated beads were used as negative controls because this serotype is poorly captured by KCs (Fig. 6 A). Mass spectrometry analysis identified 81 membrane-associated proteins that were significantly enriched by the CPS14-conjugated beads (Fig. 8 B and Table S2).
Expression of the top hits in Chinese hamster ovary (CHO) cells resulted in significantly enhanced adhesion to TH87014 only with the murine ASGR1 construct. The ASGR1-expressing cells captured 45.5% of TH87014, but transfection with the other candidates (e.g., ASGR2, CD36, C1qbp, and endoglin) did not yield significant impact (Fig. 8 C). ASGR1 and ASGR2 are two integral subunits of ASGR, a C-type lectin that is uniquely expressed in the liver (Hoober, 2020). ASGR mediates endocytosis and lysosomal degradation of desialylated glycoproteins and platelets by recognizing terminal galactose or N-acetylgalactosamine (GalNAc) residues (Deppermann et al., 2020; Stockert, 1995) but has not been associated with the recognition of any pathogens. Consistent with an accessory role of the ASGR2 subunit in ASGR (Shia and Lodish, 1989), additional expression of ASGR2 in the ASGR1 transfectants further boosted adherence of TH87014, although expression of ASGR2 alone did not result in obvious pneumococcal binding (Fig. 8 C). Further analysis revealed specific binding of murine ASGR (mASGR) to pneumococci of serotype 7F but not any of the other 13 types tested thus far (Fig. 8 D).
The serotype specificity of mASGR action was verified by dose-dependent inhibition with CPS7F and CPS14, but not CPS19F or CPS8 (Fig. 8 E and Fig. S4 A). Similar experiments with galactose and GalNAc, the known ASGR ligands (Stockert, 1995), showed that ASGR recognizes the terminal galactose in the repeating units of CPS7F and CPS14 (Fig. 8 E and Fig. S4 A). CPS14 was also shown to block mASGR-mediated adhesion of both serotype-14 and -7F pneumococci, but CPS7F inhibited only the adhesion of homologous serotype-7F strain. This result suggests that CPS14 is superior to CPS7F as a ligand for mASGR (Fig. 8 F). Consistent with sequence conservation between mouse and human ASGR (hASGR) proteins (Fig. S4 B), coexpression of hASGR1 and hASGR2 in CHO cells also led to serotype-specific adhesion of serotype-7F and -14 pneumococci (Fig. 8 G). Likewise, the hASGR-mediated recognition of CPS7F and CPS14 was blocked by the homologous CPSs, galactose or N-acetylgalactosamine (Fig. S4, C and D). The serotype-specific recognition of CPS7F and CPS14 by mASGR and hASGR was also confirmed with natural pneumococcal strains (Fig. S4, E and F). It should be noted that none of the seven HV CPSs (types 1, 2, 3, 4, 5, 6A, and 8) contains the terminal galactose in its repeating units, which agrees with the poor recognition of the HV capsule variants by KCs (Fig. 6 A). Moreover, ASGR did not show significant binding to several pneumococcal serotypes with terminal galactose in the capsules, such as 10A, 12F, 15C, 17F, and 33F. Analysis of the architectures of pneumococcal CPSs revealed that only the branched chain in CPS14 is identical to the endogenous ligand core structure Galβ1,4GlcNAc (Fig. S4 G; Li et al., 2017; Sorensen et al., 2009). The contribution of ASGR-mediated capsule recognition to KC capture of pneumococci was confirmed in vitro with KCs freshly isolated from mouse and human livers. Pretreatment of murine or human KCs with an anti-ASGR1 antibody resulted in significant reduction in cellular binding of TH87014 (Fig. 8 H). Together, these experiments identified ASGR as a novel PRR on KCs, recognizing serotype-7F and -14 pneumococcal capsules.
Finally, we determined the contribution of ASGR to KC capture of blood-borne pneumococci in ASGR1-deficient mice. Significant impairment in the clearance of serotype-14 pneumococci was observed in ASGR1−/− mice (Fig. 8 I). Compared with WT mice, ASGR1−/− mice showed significantly higher levels of bacteremia (5.3-, 16.7-, 32.0-, and 9.2-fold) at 5, 10, 20, and 30 min, respectively. In particular, pneumococci abundantly circulated in the bloodstream of ASGR1−/− mice at 20 and 30 min, although WT control showed undetectable bacteremia at those time points. ASGR1−/− did not exhibit an obvious defect in clearing the isogenic serotype-19F bacteria (Fig. 8 I). Moreover, ASGR1−/− mice showed 44.4% lethality upon i.v. challenge with 108 CFU of TH87014, a nonlethal dose for WT mice (Fig. 8 J). These results demonstrate that ASGR is involved in the recognition and capture of serotype-14 pneumococci by KCs in the liver of mice. It should be noted that the ASGR1−/− mice were relatively more capable of clearing serotype-14 pneumococci than the KC-depleted (Fig. 6 C) or CPS14-blocked (Fig. 7 B) mice. This discrepancy suggests that additional uncharacterized KC receptor(s) is also involved in the recognition of the serotype-14 capsule. Together, these data demonstrate that the C-type lectin receptor ASGR acts as a capsule recognition receptor on KCs to promote serotype-specific clearance of blood-borne pneumococci.
E. coli capsules compromise bacterial capture by KCs
Because capsule variation exists in many bacterial species, we further determined whether the capsule-mediated immune evasion against KC recognition and capture broadly operates in Gram-negative sepsis pathogens, using E. coli isolates from human sterile sites. Based on our pilot test of virulence phenotypes in the septic infection model (Table S1), we chose two LV (TH4509K2ab and TH14508KG2-1) and two HV (TH14515K1 sand TH14864K5) strains for further characterization. Each of the four strains represented a unique capsule (K) type. While the LV strains were not or transiently detected in the bloodstream with virtually full survival in mice, the mice infected by HV counterparts showed rapidly rising bacteremia and 100% mortality within 24 h after i.p. inoculation (Fig. S5, A and B). To define the importance of the capsules on E. coli virulence in septic infection, we first tested the virulence of the acapsular mutant of the HV strain TH14515K1. Intravenous inoculation of the mutant in mice did not lead to detectable bacteremia or mortality (Fig. 9 A, ∆cps), manifesting the importance of the K1 capsule in E. coli virulence.
We further assessed the relationship between capsule type and virulence phenotype by generating isogenic E. coli strains that differed only in the cps genes (Fig. S5 C). Placing the cps genes from the HV strain TH14515K1 or TH14864K5 converted the ∆cps mutant to the HV phenotype with violent bacteremia and 100% mortality (Fig. 9 A). In an opposite manner, transferring the cps genes from the LV strain TH4509K2ab or TH14508KG2-1 to the same recipient did not change the virulence phenotype of the mutant (Fig. 9 A). Along the same line, the isogenic K1 and K5 strains caused persistent bacteremia in the early phase of septic infection, but the K2ab and KG2-1 counterparts were rapidly cleared from the circulation (Fig. 9 B). CT50 values for the K1 (5.2 min) and K5 (6.3 min) strains were much longer than those for the K2ab (0.76 min) and KG2-1 (0.84 min) counterparts. These experiments demonstrate the type-specific contribution of capsules to E. coli virulence in septic infection.
Further investigation identified the vast majority of the “disappearing” LV strains in the liver after i.v. inoculation, but the HV strains were much less trapped in the liver (Fig. 9 C). Selective depletion of KCs but not neutrophils or inflammatory monocytes resulted in severely impaired clearance of the LV K2ab E. coli from the circulation (Fig. 9 D). Accordingly, the DT treatment resulted in significantly reduced LV E. coli cells in the liver (Fig. 9 E). Consistently, IVM revealed that the LV bacteria were much more abundantly captured by KCs than HV counterparts (Fig. 9 F and Video 6). This result shows that the virulence levels of encapsulated E. coli are predominantly shaped by capsule type–dependent escape of KC-mediated host defense in the liver. Although relatively small fractions of both LV and HV E. coli serotypes were also found in the spleen (Fig. 9 C), surgical removal of the spleen did not affect the early clearance of selected HV (type K1) or LV (type K2ab) bacteria. The overall patterns of bacterial clearance from the circulation were similar between SPX and SHM mice for K1 and K2ab bacteria in the first 30 min of septic infection (Fig. S5 D). It should be noted that asplenic mice showed modestly increased K1 bacteria in the blood (Fig. S5 E), suggesting that the spleen serves an active role in the clearance of HV E. coli cells that are poorly defended by the liver.
To understand the molecular mechanisms governing KC recognition of the LV E. coli strains, we tested the blocking effect of purified CPSs on bacterial clearance. Mice pretreated with three LV CPSs (K2ab, KG1-1, and KG2-1) significantly inhibited the clearance of the capsule type–homologous strains (Fig. 9 G and Video 7). CPS treatment extended the CT50 values of strains TH15511K2ab, TH14978KG1-1, and TH15991KG2-1 from 0.7–4.3 min to 19.9–30.0 min (Fig. S5 F). The delayed clearance accompanied attenuation of bacterial capture by the liver (Fig. 9 H). Similar treatment with CPSs from the HV strains did not yield significant impact on clearance (Fig. S5 G). We further assessed cross–capsule type blocking effect on clearance of the LV strain with the heterologous CPSs. The heterologous type KG1-1 and KG2-1 CPSs showed marginal inhibition against hepatic clearance of isogenic strain TH14515K2ab (Fig. 9, I and J; and Fig. S5 H). Finally, IVM imaging of the liver sinusoids revealed that treatment with the homologous but not heterologous CPSs led to significant reduction in the KC-bound LV E. coli (Fig. 9 K). The capsule type–specific blocking indicated that the LV E. coli bacteria were specifically recognized by CPS receptors on KCs.
The serotype-specific recognition of LV E. coli capsules was further confirmed by serotype-specific role of the complement system in the early clearance of LV E. coli. While the removal of E. coli KG2-1 from the bloodstream was significantly impaired in C3−/− mice, the C3 deficiency accelerated the clearance of serotype K2ab (Fig. S5 I). The latter finding can be explained by the C3 retardation mechanisms of bacterial clearance from the blood circulation (Broadley et al., 2016). Consistent with the contribution of C3 in the clearance, CRIg−/− mice showed significant deficiency in the clearance of E. coli KG2-1 (Fig. S5 J). Interestingly, the early clearance of E. coli K2ab was also delayed in CRIg−/− mice (Fig. S5 J), which was in line with the requirement of CRIg in the slow-track clearance of blood bacteria as reported previously (Broadley et al., 2016). Lastly, we tested potential recognition of LV E. coli capsules by ASGR using hASGR1-expressing CHO cells. No significant adhesion was detected among 20 E. coli invasive isolates tested thus far (representing 13 capsule types; data not shown). Together, these data highlight the importance of molecular interactions between the capsules and KCs in shaping the outcomes of septic infections by Gram-negative sepsis pathogens.
Discussion
Capsules are an essential virulence factor in many encapsulated bacteria and fungi. A large body of literature has characterized capsules as antiphagocytic structures (Comstock and Kasper, 2006; Paton and Trappetti, 2019; Wen and Zhang, 2015), but it remains unknown which host tissue site is primarily targeted by capsules during systematic infection. This study uncovers the liver as the major organ where capsules maximally excise their virulence-enabling function. The liver has been shown to be a major organ to trap invading bacteria (Benacerraf et al., 1959; Brown et al., 1981a; Gregory et al., 1996; Mackaness, 1962; Martin et al., 1949; Rogers, 1956). We have verified that acapsular S. pneumoniae and E. coli are rapidly cleared from the bloodstream in the early phase of septic infections. However, encapsulated bacteria remain in the circulation in a capsule type– and infection dose–dependent manner. The capsule-mediated evasion of hepatic clearance allows encapsulated bacteria to survive and proliferate in the blood circulation and leads to uncontrolled bacteremia and septic death. These findings provide a conceptual explanation for the dominance of encapsulated bacteria in human septic infections.
KCs constitute ∼90% of total tissue macrophages in the body (Bilzer et al., 2006), with well-known function in capturing blood-borne bacteria (Broadley et al., 2016; Gola et al., 2021; Helmy et al., 2006; Kolaczkowska et al., 2015; Lee et al., 2010; Surewaard et al., 2016; Wong et al., 2013; Zeng et al., 2016) and fungi (Sun et al., 2019). This work reveals that the antibacterial function of KCs is differentially compromised by various types of capsules in S. pneumoniae and E. coli. The HV capsule types are extremely capable of circumventing KC capture, which is manifested by sustained bacteremia in mice and relatively poorer binding of primary human and mouse KCs to HV capsules. The uneven contribution of KCs to the clearance of pneumococcal serotypes also explains the previous finding that macrophage depletion with clodronate liposomes yields variable levels of impact on the clearance of different pneumococcal serotypes/strains (Gerlini et al., 2014).
This study has shown that KCs are able to recognize the LV capsules via capsule type-specific receptors. This is exemplified by our discovery of ASGR as the KC receptor for pneumococcal serotype-7F and -14 capsules. ASGR is exclusively expressed in the liver (Fagerberg et al., 2014; Monroe and Huber, 1994), particularly on hepatocytes and KCs (Azimifar et al., 2014; Consortium et al., 2018; MacParland et al., 2018). While it is known that KCs use ASGR to internalize galactosylated particles (Dong et al., 2008; Kolb-Bachofen et al., 1982) and desialylated platelets (Deppermann et al., 2020; Li et al., 2017; Sorensen et al., 2009), ASGR has not been previously associated with microbial recognition. Multiple lectin receptors have been shown to recognize capsules, such as SIGN-R1 (Kang et al., 2006; Kang et al., 2004; Koppel et al., 2005) and MRC-1 (McGreal et al., 2006; Subramanian et al., 2019; Zamze et al., 2002). However, none of these receptors have been linked to KC-specific recognition of encapsulated pathogens. To the best of our knowledge, ASGR is the first known KC-specific receptor that recognizes bacterial capsules.
Our data indicate that the major complement receptor CRIg on KCs is involved in KC recognition of LV capsules in a serotype-dependent manner. CRIg is specifically expressed by KCs (Helmy et al., 2006) and promotes KC capture of C3-opsonized bacteria (Broadley et al., 2016; Helmy et al., 2006; Zeng et al., 2016). We have revealed that CRIg-deficient mice are partially disabled in capturing isogenic pneumococci producing serotype-19F and -23F, but not -14, capsules. Accordingly, a similar serotype-dependent pattern in KC capture is manifested in C3−/− mice. Given the relatively poor deposition of C3 on bacterial capsules (Fine, 1975; Hyams et al., 2010), we postulate that C3 promotes CRIg binding to C3-opsonized encapsulated bacteria once C3 is activated on the bacterial surface. Previous studies have identified numerous soluble lectins in the plasma that recognize capsules, which activate C3 via the lectin complement activation pathway. In particular, ficolins of human and mouse recognize specific motifs on certain capsules of S. pneumoniae (Brady et al., 2014; Geno et al., 2018; Krarup et al., 2005) and many other bacteria (Aoyagi et al., 2005; Aoyagi et al., 2008; Hummelshoj et al., 2012; Krarup et al., 2005). It is possible that capsule-binding soluble lectins promote C3 deposition and activation on the bacterial surface and thereby enhance C3-mediated opsonophagocytosis of encapsulated bacteria via CRIg.
While some studies suggest that KCs directly kill microorganisms trapped in the liver (Benacerraf et al., 1959; Klein et al., 1994), others report that KCs lack the capacity of generating reactive oxygen radicals upon infection-associated signals (Lepay et al., 1985) and are unable to kill bacterial and protozoal pathogens taken up in the liver (Gregory et al., 2002; Gregory et al., 1996; Lepay et al., 1985). In the latter cases, neutrophils are described as the major helper to kill the microbes captured by KCs (Gregory et al., 2002; Gregory et al., 1996; Kolaczkowska et al., 2015). Our data strongly suggest that KCs are able to kill the bacteria attached to their surface without help of other professional phagocytes in the early phase of septic infections. The total bacteria in the liver were dramatically diminished in the first 30 min after i.v. inoculation in the absence of neutrophils and inflammatory monocytes, a time window far before massive recruitment of neutrophils upon blood-borne bacterial infection (Gregory et al., 1996). The bactericidal activity of KCs is also indicated by our observation that KCs take up LV pneumococci in the liver sinusoids.
Capsule types have been associated with disease potential of many encapsulated pathogens (Briles et al., 1992; Hyams et al., 2013; Robbins et al., 1974; White, 1938). However, genetic heterogeneities among bacterial strains and patients, as well as host immunological conditions, have precluded the establishment of a causal relationship. By using isogenic capsule variants, we have demonstrated a decisive role of capsule types/structures in shaping the survival and virulence potential of encapsulated bacteria in mice. At a given infection dose (106 CFU i.v. for S. pneumoniae, 107 CFU i.v. for E. coli), certain capsular types always confer the HV phenotype (e.g., S. pneumoniae serotypes 1–5 and E. coli serotypes K1 and K5), whereas the bacteria with other capsules constantly display the LV phenotype (S. pneumoniae serotypes 14 and 19F and E. coli serotypes K2ab and KG2-1). It should be noted that the LV capsules can overwhelm the KC-mediated defense and confer the HV phenotype at high infection doses, which may represent the situation when large numbers of encapsulated LV bacteria in local infections (e.g., pneumonia and peritonitis) continue to seed in the bloodstream and overrun the capacity of KC-mediated defense. The immune evasion conferred by the LV capsules appears to be accomplished by capsule coverage of multiple “vulnerable” pathogen-associated molecular patterns on the cell membrane/wall from be detected by the PRRs, such as lipoproteins, lipoteichoic acids, peptidoglycan, and LPS (Medzhitov, 2007). Molecular recognition of the LV capsules by KCs seems to represent a trade-off effect.
While the spleen is important for host defense against S. pneumoniae and other encapsulated bacteria (Bogart et al., 1972; Brown et al., 1981b; Leung et al., 1972; Shinefield et al., 1966; Whitaker, 1968), the precise mechanisms of this splenic immunity remain to be fully defined. This work demonstrates that the spleen is important for clearance of the HV serotypes that are poorly recognized by liver macrophages, but this famous immune organ is dispensable to natural defense against LV capsules because of effective clearance of LV serotypes in the liver. Our finding is consistent with the neutrophil-driven clearance of HV pneumococci (e.g., serotype-2 D39) in the spleen (Deniset et al., 2017). The striking difference between the liver and spleen in the capacity of clearing blood-borne bacteria appears to be determined by the receptors uniquely or differentially expressed by KCs and splenic cells. This notion is supported by serotype-specific capture of LV S. pneumoniae and E. coli serotypes by KCs. Although the immune cells responsible for the clearance of encapsulated bacteria in the spleen remain to be defined, splenic macrophages are likely involved in this function. Deniset et al. (2017) have identified red pulp macrophage as the major phagocyte to capture HV pneumococci in the spleen of mice, which agrees with the observation that Leishmania infection–induced proliferation of red pulp macrophages in mice protects septic infection of HV pneumococci (Kirby et al., 2009). Splenic marginal zone macrophages have been shown to recognize pneumococcal capsules by the SIGN-R1 lectin receptor (Kang et al., 2006; Kang et al., 2004; Lanoue et al., 2004). Interestingly, CD169+ metallophilic macrophages in the marginal zone of the spleen permit intracellular proliferation of HV pneumococci, which allows resurrection of bacteremia after the blood is sterilized (Ercoli et al., 2018). The importance of the spleen in the defense of HV serotypes warrants future investigations on the precise contributions of various splenic resident macrophages.
Although this work focuses on the early phase of blood-borne infections by encapsulated bacteria, the mechanisms governing the capsule interactions with KCs can have profound implications for the understanding of pneumococcal pneumonia and other bacterial diseases often leading to septic infections. Approximately 50% of patients with pneumococcal pneumonia develop bacteremia, and bacterial dissemination to the blood is associated with severe disease and prognosis (Musher and Thorner, 2014). Our data strongly suggest that bacterial serotype and the health conditions of the liver and spleen are among the decisive risk factors for transforming pneumococcal primary infection in the lungs into a septic infection. The abundant presence of pneumococci in the liver, but not in the blood, of mice infected i.t. with LV serotypes indicates that the LV pneumococci disseminated from the lungs are effectively cleared in the liver, which prevents bacterial proliferation in the blood and lethality. Hypersusceptibility of KC-depleted mice to a sublethal infection with LV pneumococci is consistent with the fact that the individuals with chronic liver diseases are associated with higher risk of developing sepsis with bacteremia (Ashare et al., 2009; Foreman et al., 2003). In sharp contrast, severe bacteremia, poor bacterial deposition in the liver, and 100% mortality in mice infected i.t. with HV pneumococci corroborate our finding in the septic infection model that HV capsules prevent hepatic clearance of invading bacteria. Consistent with the importance of the spleen in the defense against HV pneumococci (Deniset et al., 2017), the spleen carries more bacteria than the liver in mice infected with the HV pneumococci.
In summary, this work reveals that the liver serves as the frontline immune organ to defend blood-borne infections of acapsular and LV encapsulated bacteria via KC-mediated pathogen recognition mechanisms, but hepatic function is severely compromised by HV capsules, leaving the spleen as the major immune organ to control relatively low levels of septic infections by HV serotypes. In the context of the success in the prevention of invasive pneumococcal disease by CPS-based vaccines (Briles et al., 2019), these findings have provided molecular and cellular insights into potential mechanisms governing the immunoprotection efficacy of the pneumococcal vaccines in controlling infections of S. pneumoniae. It is tempting to hypothesize that vaccine-elicited antibodies against pneumococcal capsules, particularly HV serotypes, makes HV pneumococci more approachable by KCs and/or splenic immune cells.
Materials and methods
Bacterial strains and cultivation
All of the S. pneumoniae and E. coli strains used in this study are described in Table S3. Pneumococci were cultured in Todd-Hewitt broth with 0.5% yeast extract or tryptic soy agar plates with 3% defibrinated sheep blood at 37°C with 5% CO2 as described (Lu et al., 2006). E. coli strains were grown in Luria-Bertani (LB) broth or agar plates at 37°C under aerobic conditions. Antibiotics were added when necessary at the following concentrations: 400 μg/ml for kanamycin, 150 μg/ml for streptomycin, 4 μg/ml for chloramphenicol, 5 μg/ml for gentamycin, 200 μg/ml for apramycin, and 300 μg/ml for spectinomycin.
Bacterial mutagenesis
The materials and procedures used to generate S. pneumoniae and E. coli mutants are described in Table S3. Pneumococcal capsule-switch strains were constructed by natural transformation as described (Li et al., 2016). Briefly, each background strain was first converted to gain streptomycin resistance by transformation with the rpsL1 allele (Sung et al., 2001). The capsule biosynthesis genes in the cps locus were then replaced by the Janus Cassette (JC) using the amplicons of the up- and downstream sequences with strain-specific primers to generate the Δcps::JC background strains, which were used to generate unmarked cps deletion mutant with the fusion PCR segments of the up- and downstream arms by counterselection with streptomycin as described in Table S3. For capsule switching, the sequence of each target cps locus between the dexB and aliA genes along with the up- and downstream sequences (∼3 kb each) was amplified from genomic DNA of the donor strains using PrimeSTAR GXL DNA polymerase (Takara) and primers as described in Table S3 and used to transform the Δcps::JC mutants. To improve transformation efficiency, before amplification of the target cps loci, a chloramphenicol-resistant cat gene was inserted at the 3′ end of the aliA gene in the donor genomes by transformation with the fusion fragments as described in Table S3, except serotype 5 strain TH2940, which was constructed with kanamycin-resistant kan gene. The cat and kan markers were amplified from plasmid pEVP3 and the JC segment with primers Pr12140/Pr12141 and Pr12618/Pr12619, respectively, and then fused to the 5′ and 3′ flanking sequences of each target locus as described in Table S3. The capsule-switching transformants were selected on tryptic soy agar blood plates in the presence of streptomycin and chloramphenicol, except serotype 5 transformants that were selected with streptomycin and kanamycin. In some cases, capsule-switching strains were obtained by transformation with genomic DNA of the donor strains when repeated trials did not yield desirable transformants with the donor cps amplicons, which were backcrossed at least three times before being used for any experimentation as described previously (Trzcinski et al., 2003). The serotypes of all derivatives were genotypically confirmed by PCR and DNA sequencing with diagnostic primer sets as described previously (Pai et al., 2006).
E. coli capsule-switching strains were constructed in the TH14515 (K1) background in a manner similar to that described previously (Wang et al., 2018a) and illustrated in Fig. S5 C. In brief, the Cas9 encoding plasmid pCasKP-apr was introduced into recipient strain TH14515 by electroporation. The resulting Cas9-positive strain TH15325 was maintained at 30°C with apramycin and transformed with the spacer-containing plasmid pTH15353, which contained the up- and downstream homologous sequences of the cps locus of the recipient strain at 30°C in the presence of apramycin and spectinomycin, resulting in strain TH15399. TH15399 was used to generate the cps-null mutant TH15403 by curing both the plasmids cultivation at 37°C (pCasKP-apr) in the presence of 5% sucrose (pTH15353), or to produce the isogenic capsule type switching strain TH15401 by curing pTH15353 with sucrose. To improve transformation efficiency, a transitional strain was generated for each cps donor sequence by inserting the 5′ and 3′ arm sequences (1.5 kb each) of the donor and recipient cps loci in the cps locus of TH15401. Specifically, the 5′ and 3′ arm sequences of the recipient and donor cps loci, as well as the plasmid pTH15358, were amplified with the primers outlined in Table S3; the five amplicons linked with the NEB Assembly Mix and electroporated into TH15401. The pSGKP-spe derivatives were cured by incubating the bacteria at 30°C in the presence of apramycin and 5% sucrose. Donor cps sequences were amplified from the donor genomes with the primers outlined in Table S3 and the spacer-containing plasmid were outlined in Table S3. Capsule-switched recombinants were constructed by coelectroporation of corresponding donor cps sequences and spacer-containing plasmid into the corresponding transitional recipient strains. Capsule-switched recombinants were selected on LB plate with apramycin and spectinomycin, and these two plasmids were cured as mentioned before. The accuracy of all the genetic constructs was confirmed by PCR amplification.
E. coli genome sequencing
E. coli genomes were sequenced as described previously (Li et al., 2012). Genomic DNA was extracted using the HiPure Bacterial DNA Kit according to the manufacturer’s instructions (Magen) and used to generated genomic libraries by VAHTS Universal DNA Library Prep Kit for Illumina for DNA sequencing on the Illumina NovaSeq 6000 platform with 150-bp paired-end modules. The raw reads were filtered to remove low-quality reads and assembled by SPAdes genome assembler v3.14.1. The E. coli K-antigen serotype was determined through web Basic Local Alignment Search Tool (BLAST). The sequences of cps loci were aligned to the standard database Nucleotide collection (nr/nt) using Megablast with default parameters. The K-antigen group was defined according to specific genes in group 1 (wzi and wzy), group 2 or 3 (kpsF, kpsE and kpsM), and group 4 (wzy). Individual K-type was designated based on the identity of the cps sequence to previously reported serotypes. In certain cases, the cps sequences cannot be aligned to any defined cps locus in E. coli, and these new types were designated as KG1-1 (K-antigen in group 1, number 1, GenBank accession MZ339221) and KG2-1 (MZ339219) in this study, respectively.
Mouse infection
All infection experiments were conducted in sex-matched C57BL/6 or CD1 mice (6–8 weeks old) according to the animal protocols approved by the Institutional Animal Care and Use Committee in Tsinghua University. All of the knockout/in mice were maintained on the C57BL/6 background. Clec4F-DTR mice were generated as described previously (Scott et al., 2016). C3−/− and ASGR1−/− mice were purchased from The Jackson Laboratory and Cyagen, respectively. CRIg−/− mice (Helmy et al., 2006) were acquired from Genentech.
Septic infections were carried out by i.p. or i.v. injection essentially as described previously (Wang et al., 2018b). Lung infections were performed by i.t. inhalation after anesthesia. In brief, bacterial inoculum was prepared by resuspending frozen stocks in Ringer’s solution to desirable concentrations based on the predetermined CFU for each stock. Precise concentrations were determined by plating the inoculum immediately before infection. Desirable bacterial CFU were inoculated in a final volume of 200 μl for i.p., 100 μl for i.v., or 30 μl for i.t. infection. A higher infection dose of E. coli strains (i.p. 107 CFU) than S. pneumoniae (i.p. 104 CFU) was used for virulence assessment because the former were generally less virulent in the sepsis model.
Bacteria in the bloodstream were assessed by retroorbital bleeding and CFU plating. CT50 was calculated by nonlinear regression analysis of bacteremia kinetics for individual strains using the formula T = ln(1 − 50/plateau)/(−K), in which plateau and K were generated by one-phase association of the clearance ratio using GraphPad Prism. CT50 was designated as 30 min if less than half of the inoculum was cleared from the blood. Total viable bacteria burden in each animal was calculated as the sum of CFU values in the blood and major organs. Bacteria in the liver, spleen, lung, kidney, and heart were measured by CFU plating of tissue homogenates and presented as CFU per organ. Animal survival was recorded in a 7-d period or at a humane endpoint (body weight loss >20%). Competitive inhibition of bacterial clearance in the circulation of mice was similarly performed by i.v. injection purified CPSs 2 min before i.v. inoculation of encapsulated bacteria.
Splenectomy
SPX mice were prepared as described (Coil et al., 1978). The mice were anesthetized with 400 mg/kg of avertin (Sigma-Aldrich) and received 8 mg/kg of meloxicam (Sigma-Aldrich) before surgery. The peritoneum was opened on the left side to suture the spleen pedicle before removing the spleen and closing the peritoneum. SHM mice were treated in the same manner without removing the spleen. All animals were allowed to recover for ≥10 days before infection.
IVM
IVM imaging of mouse liver was carried out with inverted confocal laser scanning microscopy as described (Wang and Kubes, 2016). The hepatic microvasculature and KCs were visualized by i.v. injection of 2.5 μg AF594 anti-CD31 and AF647 anti-F4/80 antibodies, respectively, 30 min before i.v. inoculation with 5 × 107 CFU of FITC-labeled S. pneumoniae or E. coli. Images were acquired with Leica TCS SP8 confocal microscope using 10×/0.45 NA and 20×/0.80 NA HC PL APO objectives. The microscope was equipped with Acousto Optics without filters. Fluorescence signals were detected by photomultiplier tubes and hybrid photo detectors (600 × 600 pixels for time-lapse series and 1,024 × 1,024 pixels for photographs). Three laser excitation wavelengths (488, 585, and 635 nm) were employed by white light laser (1.5 mw, Laser kit WLL2, 470–670 nm). Real-time imaging was monitored for 2.5–5 min after infection. Five to 10 random fields of view at 10 min were selected to calculate the bacteria number per field of view.
Fluorescence labeling
S. pneumoniae and E. coli strains were labeled with FITC as described (Wright and Jong, 1986) with minor modifications. Briefly, 108 CFU of bacteria were suspended in 1 ml PBS containing 200 μg/ml FITC (Sigma-Aldrich) and incubated for 30 min at room temperature (RT) in the dark. The FITC-labeled bacteria were washed three times with 0.5 ml PBS to remove unbound FITC. pHrodo-labeled pneumococci were prepared according to the manufacturer’s instructions with the exception of using 0.001 mM pHrodo Red dye instead of 0.5–1.0 mM suggested by the manufacturer.
Immune cell depletion
Neutrophils and neutrophils/inflammatory monocytes were depleted by i.p. injection of 500 μg anti-Ly6G antibody (clone 1A8; BioXCell) and 500 μg anti-Ly6G/Ly6C antibody (Gr1; clone NIMP-R14; BioXCell), respectively, 24 h before infection as described previously (Xiong et al., 2015). The same dose of isotype-matched IgG was used as a negative control. Specific depletion of KCs was achieved by i.p. injection of 10 ng/g DT in the Clec4F-DTR mice 24 h before infection as described previously (Scott et al., 2016). The depletion efficiencies were verified by flow cytometry as shown in Fig. S2, E and F.
Isolation of liver NPCs
Liver NPCs were isolated by a collagenase-DNase digestion procedure as described (Li et al., 2014). In brief, mice were euthanized, and liver was perfused from the portal vein with 5 ml of digestion buffer (HBSS with 0.5 mg/ml collagenase IV, 20 μg/ml DNase I, and 0.5 mM CaCl2). The liver was excised, minced into small pieces (<2 mm), suspended in 10 ml digestion buffer, and incubated at 37°C for 30 min with constant rotation at 300 rpm. The homogenate was filtered through a 70-μm strainer and chilled on ice for 5 min, then washed by centrifugation and resuspension in ice-cold HBSS. Residual red blood cells were lysed by resuspending the pellet in 1 ml RBC lysis solution (BioLegend) and incubated on ice for 1 min before the reaction was terminated with HBSS. The cell suspension was then centrifuged at 50 g for 2 min at 4°C to collect the top aqueous phase containing NPCs. Human liver sections were similarly processed without perfusion. For in vitro bacteria-binding assay, the NPCs were resuspended in RPMI 1640 without FBS and seeded into 96-well cell culture plate at 5 × 104 cells/well. KCs were enriched by cellular adherence to the plastic surface after incubation at 37°C, 5% CO2 for 15 min. Nonadhering cells were removed by gently washing with RPMI 1640. This procedure typically yielded 1 × 104 cells/well for KCs that were used immediately for bacterial binding experiments. Otherwise, the NPCs were used for flow cytometry analysis without adherence. Human liver sections were obtained from freshly disposed tissues of liver surgery or transplantation patients with approval by the Human Subject Study Committee of the Strategic Support Force Medical Center (K2021-012-01).
Flow cytometry
Flow cytometry was carried out essentially as described (Scott et al., 2016). Mouse liver NPCs (1 × 106) or blood immune cells from 200 μl whole blood were blocked for 10 min in 50 μl FACS buffer (PBS with 3% FBS) with 1% anti-CD16/32 antibody, followed by staining for 20 min with APC-Cy7 anti-CD45 (1/200), APC anti-CD31 (1/200), BV605 anti-CD11b (1/500), FITC anti-F4/80 (1/200), PB anti-Ly6C (1/500), and PE anti-Ly6G (1/500). Before analysis, cells were stained with 5 μl 7-AAD for viability test. Viable cell populations were gated as KCs (CD45+CD31−CD11blowF4/80high), splenic red pulp macrophages (CD45+CD11blowF4/80high), inflammatory monocytes (CD45+CD31−CD11b+Ly6ChighLy6G−), and neutrophils (CD45+CD31−CD11b+Ly6C+Ly6Ghigh).
Purification of CPSs
Pneumococcal and E. coli CPSs were purified as described with minor modifications (Kobayashi et al., 2018; Lee et al., 2020). Pneumococci were cultured to OD620 0.5 in Todd-Hewitt broth with 0.5% yeast extract, and E. coli strains were grown overnight in nutrient-rich medium (2 g/liter KH2PO4, 9.7 g/liter K2HPO4, 1 g/liter (NH4)2SO4, 0.1 g/liter MgCl2, 0.5 g/liter sodium citrate, 10 g/liter glucose, and 2 g/liter yeast extract, pH 7.0–7.5). Bacterial cultures were chilled on ice and centrifuged at 10,000 g for 20 min at 4°C. The pellets were resuspended in 1/10 original culture volume in 50 mM citrate buffer (pH 4.5) with 0.1% Zwittergent 3-14 detergent (Millipore) and incubated at 42°C for 30 min. Cell debris was removed by centrifugation at 10,000 g for 5 min at 4°C. CPS-containing supernatants were filtered through a 0.2-μm Nalgene Rapid-Flow filter (Thermo Fisher Scientific). The solution was mixed with one-fourth volume of ethanol and stirred at 4°C for 2 h. Precipitated nucleic acids were removed by centrifugation at 10,000 g for 15 min at 4°C. CPSs in the supernatants were further precipitated by increasing the concentration of ethanol to 80% and incubation at 4°C for 2 h. The precipitants were collected by centrifugation at 10,000 g for 20 min at 4°C. The CPS pellets were air dried and resuspended in sterile H2O. For E. coli CPSs, the LPSs were detoxified by incubation in 85% ethanol with 0.1 M NaOH at 37°C and subsequent neutralization with 0.1 M acetic acid. The CPSs were pelleted by centrifugation at 10,000 g for 20 min at 4°C, air dried, and dissolved in sterile PBS. CPSs were concentrated by ultracentrifugation with 100-kD cutoff Ultra-centrifugal Filter Unit (Merck Millipore), and concentration was measured by the phenol-sulfuric acid method (Kobayashi et al., 2018). In brief, to the standard D-glucose (concentration from 0 to 0.1 mg/ml) or CPS sample, 6% phenol and sulfuric acid were added to the tubes sequentially; absorbance at 490 nm was measured after incubating at RT for 30 min. CPS concentration was calculated according to the standard curve.
Screening of CPS14-binding proteins
Pneumococcal CPS-binding proteins were isolated from membrane proteins of the liver NPCs by an affinity screening using CPS-coated latex beads as illustrated in Fig. 8 A. The membrane proteins were prepared from murine NPCs using the Mem-PER Plus Membrane Protein Extraction Kit (Thermo Fisher Scientific) with Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific) according to the manufacturer’s instructions. The resulting proteins were concentrated through a 3-kD cutoff Ultra-centrifugal Filter Unit (Merck Millipore), exchanged into PBS with 0.05% Tween 20 (PBST), and quantified with the BCA Assay Kit (Beyotime).
Serotype-8 and -14 CPSs of S. pneumoniae were separately coated onto carboxyl latex beads as described (Schlottmann et al., 2006). Briefly, 2.5 mg of the CPSs were dissolved in 2.5 ml H2O and mixed with 200 μl of 200 mg/ml 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium chloride (DMTMM; Sigma-Aldrich). The mixture was incubated for 1 h on a rotator at RT. The DMTMM-modified CPSs were buffer-exchanged by ultracentrifugation through a 10-kD cutoff Ultra-centrifugal Filter Unit (Merck Millipore) into 2.5 ml PBS. The resulting DMTMM-CPSs were mixed with 1 × 109 carboxyl latex beads (2 μm; Invitrogen) and incubated on a rotator overnight at RT. The beads were washed twice with 2.5 ml PBST and stored in 2.5 ml blocking buffer (PBS with 1% BSA and 0.05% NaN3) at 4°C in the dark.
To isolate CPS-binding proteins, 1 × 108 CPS-coated beads were mixed with 100 μg of the membrane protein preparation from the murine NPCs in 500 μl PBST containing 2 mM CaCl2 and MgCl2. The mixtures were rotated at RT for 1 h and washed three times with PBST by centrifugation and resuspension. The beads were resuspended in 50 μl PBST with 1 × SDS-PAGE loading buffer and heated at 100°C for 5 min to dissociate CPS-binding proteins from the beads. Samples were spun at 12,000 g for 5 min at 4°C, and the supernatants were used for quantitative protein identification by mass spectrometry.
Protein mass spectrometry
Murine proteins were quantitatively identified by mass spectrometry as described (Liu et al., 2019). Samples with 40 μg total proteins were separated on SDS-PAGE gels to excise protein bands for in-gel digestion and subsequent procedures for liquid chromatography–tandem mass spectrometry analysis using a Thermo-Dionex Ultimate 3000 HPLC system combined with the Thermo Orbitrap Fusion mass spectrometer. The spectra from each liquid chromatography–tandem mass spectrometry run were searched against the mouse Unreviewed TrEMBL FASTA database (release 2020_05) using Proteome Discovery searching algorithm (v1.4). Proteins with two or more unique peptide matches were selected to calculate protein abundance, which was calculated as the median of all peptide hits belonging to the same proteins. Protein abundance was compared between the CPS14- and CPS8-coated beads, and proteins with more than twofold preference to CPS14 beads were regarded as CPS14-binding candidates.
Expression of mouse and human proteins
Expression of murine and human proteins in CHO cells was carried into pCDH or pFUGW vectors as described (Zhang et al., 2020). The full-length cDNAs of the target mouse genes were amplified from the total RNA isolated from sorted KCs using TRIzol Reagent (Invitrogen) using a Maxima H Minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) and cloned with a His6 tag at the C terminus. The human ASGR1 and ASGR2 were synthesized according to references NM_001671 and NM_001181, digested with XbaI/EcoRI, and ligated to pCDH and pFUGW, respectively. The ligation mixtures were transformed into E. coli DH5α and selected on LB plates with 100 μg/ml ampicillin. The recombinant plasmids were confirmed by DNA sequencing and extracted using HiPure Plasmid EF Micro Kit (Magen) for subsequent transfection. The relevant primers and resulting plasmids are listed in Table S3.
Stably expressing cell lines were constructed by lentiviral transduction. First, recombinant plasmids (either pCDH or pFUGW backbone) were transfected into HEK293T cells using Lipofectamine 2000 (Invitrogen) together with lentiviral packaging vectors pMD2.G and psPAX2 (gifts from Didier Trono at École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland; Addgene). HEK293T cells were grown in DMEM with 10% FBS and 1× penicillin/streptomycin. The lentiviral particles were harvested at 48 h and filtered through a 0.45-μm syringe filter unit (Millipore) to remove cell debris. The pCDH-lentivirus was used to infect CHO cells with 8 μg/ml Polybrene (Sigma-Aldrich). The transfectants were selected with 5 μg/ml puromycin for 7 d. For coexpression, the secondary pFUGW-lentivirus was further transduced into CHO cell lines containing pCDH-lentivirus and selected under 300 μg/ml zeocin for 10 d. DNA and protein sequence analyses were carried out with the Lasergene 15.0.0 for Macintosh.
In vitro bacterial binding assay
Bacterial binding to host cells was assessed essentially as described (Zhang et al., 2000). Primary murine and human KCs were prepared in 96-well cell culture plates as described above. CHO transfectants were seeded in 48-well plates and grown to 90–100% confluence (∼1.5 × 105 cells/well). At the time of binding experiments, growth media were replaced with basic RPMI 1640 (KCs) or F-12K (CHO cells) without serum and antibiotics. Bacteria were added at a multiplicity of infection of 1:1 in a total volume of 50 or 200 μl per well for 96- or 48-well plates, respectively, followed by centrifugation at 500 g for 5 min to mimic the sheer force that bacteria typically experience in the blood circulation and to maximize the contact between bacteria and host cells. As exemplified in Fig. S4 H, the mild centrifugation facilitated receptor-specific bacterial adhesion to host cells but did not affect nonspecific bacteria adhesion. For primary KCs, 10% normal mouse or human serum was added at the same time with bacteria. The mixtures were incubated for 30 min at 37°C with 5% CO2. The unbound bacteria were enumerated by CFU plating of the supernatants after vigorous shaking 5–10 times to reduce nonspecific binding. The host cells were thoroughly washed to remove free bacteria and lysed with 50–200 μl ice-cold sterile H2O to enumerate cell-associated bacteria by CFU plating of the lysates. Bacterial binding was calculated by dividing the cell-associated CFU by the CFU of total bacteria in the assay system. Competitive blockage of bacterial binding was accomplished by adding 20–500 μg/ml of purified CPS, 0.1–100 mM of galactose, glucose, N-acetyl-D-glucosamine (GlcNAc) or GalNAc, and 5 μg/ml rabbit polyclonal anti-mouse or anti-human ASGR1 antibodies and the isotype control to the medium at the same time with bacteria.
Quantification and statistical analysis
Statistical analysis was performed using GraphPad Prism software v8.0.0. All data are presented as mean ± SEM unless otherwise indicated. Unpaired t test was used to determine the statistical significance between two groups. One- and two-way ANOVA multiple comparisons test was used to analyze data between multiple groups. Survival curves were analyzed by log-rank test. The numbers of samples and biological replicates and P values are indicated in figure legends.
Online supplemental material
Fig. S1 shows the capsular serotype-dependent variation in the pneumococcal virulence, relating to Figs. 1 and 2. Fig. S2 illustrates the capsule-related capture of pneumococci by liver KCs, flow cytometry analysis of selective depletion efficiency, and the essential role of KCs for long-term protection against LV pneumococci, relating to Figs. 3, 5, and 6. Fig. S3 shows the serotype-specific inhibition of pneumococcal clearance by purified CPSs, relating to Fig. 7. Fig. S4 shows the specific binding of serotype-7F and -14 pneumococci to both mouse and human ASGR, relating to Fig. 8. Fig. S5 demonstrates the capsular type–dependent evasion of hepatic clearance by invasive E. coli, relating to Fig. 9. Table S1 summarizes the virulence phenotypes of S. pneumoniae and E. coli clinical isolates. Table S2 shows the screening result of CPS14-binding proteins by mass spectrometry. Table S3 summarizes bacterial strains, plasmids, primers, and mutagenesis procedures in this study. Videos 1 and 2 reveal the potent KC capture of unencapsulated pneumococci, relating to Fig. 5. Videos 3 and 4 show the capsular type–dependent KC capture of LV but not HV pneumococci, relating to Fig. 6. Video 5 illustrates the inhibition of free type 14 CPSs to KC capture of serotype 14 pneumococci, relating to Fig. 7. Videos 6 and 7 show the capsular type–dependent KC capture of E. coli and the inhibition effect of LV type K2ab CPSs, relating to Fig. 9.
Acknowledgments
We thank Quanjiang Ji (ShanghaiTech University, Shanghai, China) for pCasKP-apr and pSGKP-spe plasmids; Wanli Liu (Tsinghua University, Beijing, China) for the CHO cells; Fen Qu (302 Hospital of PLA, Beijing, China), Yingchun Xu (Peking Union Medical College Hospital, Beijing, China), and So Hyun Kim (The Catholic University of Korea, Seoul, South Korea) for the bacterial isolates; and the Tsinghua research platforms for assistance in animal experimentation (Laboratory Animal Research Center), flow cytometry, IVM imaging (Center for Cell Biology), and protein mass spectrometry (Center for Proteomics).
This work was supported by grants from National Natural Science Foundation of China to J.-R. Zhang (31820103001 and 31530082), Tsinghua University Spring Breeze Fund to J.-R. Zhang (20201080767), and the Tsinghua-Peking Joint Center for Life Sciences Postdoctoral Foundation and the China Postdoctoral Science Foundation to H. An (2016M590085).
Author contributions: Conceptualization, H. An, J. Li, J.M. Leong, and J.-R. Zhang; experimentation, H. An, C. Qian, Y. Huang, J. Li, X. Tian, J. Feng, J. Hu, Y. Fang, F. Jiao, Y. Zeng, and X. Huang; methodology, X. Meng, Z. Zeng, Y. Chen, Y. Wu, J. Wang, X. Lin, X. Liu, Y. Chen, M. Guilliams, M.R. Oggioni, H. Deng, R. Zhang, H. Wang, and J. Wu; data analysis, H. An, C. Qian, Y. Huang, X. Tian, J. Li, H. Deng, J.M. Leong and J.-R. Zhang; composition, H. An, C. Qian, Y. Huang, J. Li, and J.-R. Zhang; funding, H. An, X. Liu, J.-W. Veening, and J.-R. Zhang.
References
Competing Interests
Disclosures: M.R. Oggioni has a PhD student jointly with GSK Vaccines working on the pathogenesis of pneumonia (BBSRC grant BB/S507052/1). No other disclosures were reported.
Author notes
H. An, C. Qian, Y. Huang, J. Li, and X. Tian contributed equally to this paper.