A type 6A clinical isolate of S. pneumoniae was tested in a previously described infant mouse model with intranasal (IN) challenge at age 4 days [14][15]. An advantage of this model is that infant mice are susceptible to a low colonizing dose [16]. To examine the effect of influenza A virus (IAV) co-infection, the mouse-adapted strain x31 was inoculated at age 8 days (schematically represented in Fig 1A). The density of colonization was determined in nasal lavages obtained at age 12 days. All pups were heavily colonized and mice co-infected with IAV showed a significantly higher burden of organisms compared to controls (Fig 1B).

Otitis media was assessed in lavages of the middle ear cavities obtained through the tympanic membrane on day of age 12 using microsurgery. In pilot experiments, we confirmed that the presence of pneumococci in middle ear lavages correlated with an acute inflammatory response as indicated by an increase in numbers of neutrophils (CD45⁺, CD11b⁺, Ly6G⁺ cells) detected by flow cytometry (data not shown). Pneumococci were detected in 87% of middle ear samples among the IAV co-infected group compared to 21% of the PBS control group (Fig 1C).

To detect systemic infection after intranasal challenge, the observation period was extended and beginning at age 13 days mice developed signs of sepsis. Pneumococcal bacteremia was confirmed by blood culture in moribund mice or in mice surviving until the conclusion of the observation period at 18 days of age. Sepsis was not detected prior to age 12 days. All mice eventually developed invasive infection and there was no statistical difference in the time course of sepsis (or bacteremia) between IAV co-infected and PBS control animals (p = 0.85) (Fig 1D).

To examine pup-to-pup transmission, one 4-day old pup was selected from the litter (index mouse) and colonized with S.pneumoniae. On day 8, both the index and all other pups (contact mice) were inoculated with IAV or PBS. The transmission experiment followed the previously described protocol except it was terminated at age 12 days [14]. A shorter observation period was used to minimize transmission among contact mice, since once acquiring the organism a contact mouse could become a source of transmission. It was not possible to abbreviate this observation period further since in pilot experiments at time points prior to age 12 days transmission was unusual. Under the conditions tested, 70% of contact mice became colonized by age 12 days. In contrast, for the control group where all pups were inoculated with PBS instead of IAV at age 8 days, no transmission events were observed (Fig 1E).

Previously, we reported that transmission was proportional to the level of pneumococcal shedding from the index pup [14]. To evaluate the impact of IAV on pneumococcal shedding, pups were treated as index mice (colonized at day of age 4 and infected with IAV or PBS at age 8 days) and the number of shed bacteria measured daily from day 8 to 12. IAV co-infection increased the mean shedding of pneumococci on days 10–12 by ~10–50 fold (Fig 1F).

Thus, following intranasal challenge with this single isolate, infant mice became colonized, developed respiratory tract (otitis media) and invasive infection (bacteremia/sepsis), and transmitted the organism among the litter. Influenza A significantly increased the burden of colonization and risk of otitis media. Additionally, influenza A co-infection was permissive for transmission–an effect that correlated with enhanced shedding.

To assess bacterial population bottlenecks, three mutants, each tagged with a different antibiotic resistance marker inserted into the gene encoding the pneumococcal IgA1 protease were constructed. This gene was selected because mice do not express a protease susceptible form of immunoglobulin and it was previously shown that iga mutants show no defect in murine infection [17]. There was no difference in the in vitro growth characteristics of the three strains alone or in combination (S1A Fig and S2B Fig). When inoculated equally in combination their overall ability to colonize the infant mouse nasopharynx was equivalent (S2A Fig) and indistinguishable from the parent strain. The size of population bottlenecks was estimated by examining the distribution of isogenic mutants passing through each stage in pathogenesis with or without IAV co-infection.

At age 4 days, the pups were inoculated with an equal mixture of the three strains. Although there was some variation in the colonization density of each strain for individual pups, at age 12 days no single strain consistently prevailed over the others and all pups were colonized with all three strains with (Fig 2A and 2B2, upper panels) or without IAV co-infection (S2B Fig). In pilot experiments, we also established that all pups were colonized at similar levels with these strains at earlier time points. 13 of 15 middle ear samples with pneumococci following IAV co-infection also contained all three mutants (representative data from the right middle ear in Fig 2A, lower panel). There were too few mice with otitis media in the absence of IAV to assess the bottleneck size in this experimental condition. Our findings indicated that with this low inoculum and at this time point there is not a tight population bottleneck during colonization of infant mice (with or without IAV co-infection) or in the transition of bacteria to a normally sterile site (middle ear cavity) in the upper respiratory tract in the setting of IAV (summarized in Table 1).

While all pups had three strains in nasal lavages when sacrificed, following IN challenge fewer than three mutants were detected in the majority of blood cultures with (Fig 3A, upper panel) or without (Fig 3A, lower panel) IAV co-infection. The most probable size of the bottleneck was evaluated by a mathematical model based on a Poisson distribution where the predicted and observed number of strains/mouse was compared for population bottlenecks of 1,2,3 or ≥4 cells (see Materials and Methods). This calculation suggested at some point prior to the development of bacteremia there was a population bottleneck in the setting of IAV that most often allowed passage of only a single bacterium (w = 1) (Table 2). A caveat of this observation is that this single event could represent more than one cell since the pneumococcus is a chain-forming organism. In the absence of IAV co-infection, more than a single mutant was usually detected in the bloodstream, suggesting a significantly looser bottleneck effect (w = 2 or 3) compared to the IAV group. The difference in the proportion of bacteremic events from a single strain with and without IAV co-infection was significant (p = 0.018, two-tailed Fisher’s exact test).

Next, the mechanisms contributing to the population bottleneck in the development of bacteremia were considered. To determine whether the bottleneck in events leading to bacteremia occurred following passage into or within the bloodstream, nasopharyngeal colonization was bypassed by systemically challenging pups with a low dose of the three strains. Since pups were too small for intravenous inoculation, this was carried out by intraperitoneal inoculation (IP), which leads to rapid bacterial uptake into the circulation. At 14 hours following IP challenge all mice were bacteremic at high levels with all three strains (Table 1 and Fig 2C). Thus, sepsis appeared to develop rapidly with bacterial invasion even through it was not observed before age 12 days following IN challenge. This suggests that the bottleneck seen in bacteremia following IN challenge is likely due to events following colonization and prior to accessing the bloodstream. The need to pass through this tight bottleneck could have contributed to the long incubation period for the development of bacteremia following IN challenge. When IN bacterial challenge was shifted from age 4 to 8 days, sepsis was still observed beginning at age 13 days, suggesting an age-dependent susceptibility to invasive infection following colonization (Fig 1D).

Because seeding of the bloodstream could occur following invasion from either the upper or lower respiratory tract, we determined whether this delay in the development of bacteremia was due to a requirement for infection of the lungs following colonization. However, regardless of IAV co-infection, when sampled at age 12 days just prior to the onset of bacteremia, detection of pneumonia by culture of lung homogenates was unusual (1 out of 10 mice).

The tight population bottleneck in the setting of IAV could be due to the selection of a subpopulation of pneumococci because of a requirement for de novo genetic adaptation. Alternatively, passage through a bottleneck could require selection among pre-exiting variants. To examine these possibilities, we examined whether a strain that had passed through the bottleneck showed increased fitness for bacteremic infection during rechallenge. Single strain blood isolates were mixed with strains containing the other two resistance markers and re-inoculated IN into pups at age 4 days. However, we found that a strain that had previously passed through the bottleneck was no more likely to cause bacteremia upon rechallenge compared to strains that had not passed through the bottleneck (representative data in S3 Fig upper panel).

It was concluded that invasive infection following colonization requires passage through a population bottleneck whose size is determined primarily by host factors and more restricted in the setting of IAV co-infection.

To evaluate whether a bottleneck occurs during a transmission event, the index pup was infected IN with an equal mixture of the three strains and returned to the litter. Since there was no detectable transmission among the PBS control group (Fig 1E), only IAV co-infection was tested. Two representative litters out of five are shown in Fig 3B. Although all three strains colonized the nasal cavity of the index pup similarly, only a single strain established colonization in most (19 of 27 acquiring the organism) contact mice, suggesting a single bacterium or bacterial chain (w = 1) was likely responsible for the transmission event (Table 2). The outcome of transmission from a single strain was significantly more likely than other outcomes (p = 0.006, two-tailed Fisher’s exact test). There was no consistent correlation between the strain colonizing the index pup at the highest density and the strain(s) detected in contact pups, suggesting that transmission is a stochastic event.

To define the step in transmission with a tight population bottleneck, we analyzed exit from the donor host, entry into the recipient host, and establishment in the new host. In IAV co-infected mice, there was daily shedding of each of the three mutants (Fig 2B1) and all pups were colonized by and shed all three strains (Fig 2B2), indicating a lack of a bottleneck during the exit of bacteria (Table 1). To examine whether there is a bottleneck during entry into the recipient, the experimental conditions were changed. Three index mice, each colonized at age 4 days with one of the three mutants, were used per litter to test whether there could be multiple entry events in the same recipient. Although there was still a tight bottleneck in transmission (w = 1), 6 of 11 contact pups acquiring the organism became colonized with more than one strain (Fig 3C and Table 3). This demonstrates that multiple entry events may occur and made it unlikely that strains compete at the initial entry step. In contrast to the contact mice, none of the singly colonized index mice acquired another strain. This finding suggests that once high density colonization is established further acquisition events are excluded. To confirm this observation, the experimental conditions were changed again. All the pups in a litter were colonized with a single strain–one third with each of the three mutants. No acquisition events (0 out of 17) between pups with established colonization were detected. This finding correlated with an increase in the bacterial inoculum required for establishing colonization in the majority of previously colonized compared to naive pups (>1000 CFU vs. <10 CFU). It was concluded that the tight bottleneck in transmission occurs following exit and before establishment suggesting that on average only a single shed organism/chain prevails in entering the recipient and founding a population that once established effectively outcompetes newcomers.

A strain that successfully passed through the bottleneck in transmission, was no more likely to be transmitted following IN rechallenge (representative data in S3 Fig lower panel). It was unlikely, therefore, that de novo genetic adaptation or selection among pre-exiting stable variants was a requirement for passage through the tight bottleneck in transmission.

To test the role of inflammation in the bottleneck in transmission, experiments were repeated in IAV co-infected tlr2^-/- mice, which we previously showed shed and transmit more readily due to increased inflammation in response to IAV [14]. One tlr2^-/- pup/litter was infected with the three strain mixture and co-housed with wildtype contacts. Colonization was equivalent to levels in wildtype index mice and again similar for each of the three mutants (S4A and S4B Fig). Multiple strains were detected in the nasopharynx of the majority of the wildtype contacts acquiring the organism (Table 3). A similar result was obtained when an index tlr2^-/- pup was co-housed with tlr2^-/- contacts. IAV infected tlr2^-/- index mice shed at significantly higher levels compared to wildtype controls on days of age 9 and 10 (Fig 4A), providing an explanation for increased numbers of bacteria reaching the contacts. The size of bottleneck for tlr2^-/- index mice was estimated to be at least w = 2 regardless of the genotype of the contacts and was less tight compared to wildtype index mice (see Table 2). The proportion of colonized contact mice with a single mutant was greater for experiments with wildtype compared to tlr2^-/- index mice (two-tailed Fisher’s exact test, p = 0.05). The looser bottleneck in tlr2^-/- mice with an altered response to IAV, demonstrated that the characteristics of the host response are a determining factor in the size of the bottleneck in transmission. The effect of IAV on shedding was partially recapitulated by daily IN dosing with the prototypical viral TLR3 agonist poly-ICLC (Fig 4B). In contrast, the prototypical bacterial agonists for TLR2 (Pam3Cys) and TLR4 (LPS) failed to stimulate shedding. Bacterial shedding sufficient to promote transmission requires the effect of inflammation on secretions and a high density of colonization [14]. Among the TLR agonists tested only poly-ICLC was able to stimulate inflammation as determined by the magnitude of the neutrophil response without negatively impacting colonization density (Fig 4C and 4D). This provided further evidence that specific innate immune responses to IAV dictate key determinants of the bottleneck in transmission.

This study was conducted according to the guidelines outlined by National Science Foundation Animal Welfare Requirements and the Public Health Service Policy on the Humane Care and Use of Laboratory Animals. The Institutional Animal Care and Use Committee (IACUC) at New York University approved these animal studies. New York University's IACUC oversees the welfare, well-being and proper care and use of all vertebrate animals used for research and educational purposes at NYU Langone Medical Center and School of Medicine. New York University's IACUC Assurance Number is A3435-01. The approved protocol numbers for this project are 150216–01 and 150520–01.

P1547, a serotype 6A clinical isolate that causes sepsis in adult mice following intranasal challenge, was used as a wildtype strain [22]. Strains P2396, P2397 and P2405, respectively, were constructed by transforming P1547 with plasmid DNA encoding an immunoglobulin A1 (IgA1) protease interrupted with an antibiotic resistant cassette to erythromycin (rRNA adenine N-6-methyltransferase), spectinomycin (spectinomycin adenyltransferase) or kanamycin (3', 5'-aminoglycoside phosphotransferase type III). We confirmed the insertion of the antibiotic cassettes in IgA1 protease gene by sequencing using primers (for P2396 and P2397; 5’-TCAGTAGGACTTGTATCTGC-3’ and 5’-TGGATTTAGCAATAGACGC-3’, for P2405; 5’-CCTGTCAGATCATCTCATCG-3’ and 5’-CCATTGAATAGTAGCCATTG-3’). Mutants were selected on 5% sheep blood agar plates containing erythromycin (0.1μg/ml), spectinomycin (200μg/ml) or kanamycin (500μg/ml).

Pneumococcal strains were grown statically in Tryptic Soy (TS) broth (BD, Franklin Lakes, NJ) to mid-exponential phase at 37°C. When the bacterial culture reached the desired optical density at 620nm, cells were washed and diluted in sterile PBS for inoculation. Quantitative culture was performed by plating 10-fold serial dilutions in triplicate on selective plates. Plates were incubated overnight at 37°C with 5% CO₂. Bacteria were stored in 20% glycerol at -80°C.

All mice were bred in our facility and maintained as a dam with pups from the same litter (average size 6–8). Wildtype C57BL/6 mice and tlr2^-/- mice were originally obtained from The Jackson Laboratory (Bar Harbor, ME) and bred and maintained in a conventional animal facility. Polyinosinic-polycytidylic acid condensed with poly-l-lysine (poly-ICLC, Hiltonol), a synthetic analog of viral dsRNA, was supplied by Oncovir Inc. Pam3CysSerLys4 (Pam3Cys), a synthetic triacylated lipoprotein (Invivogen) and lipopolysaccharides (LPS) from Salmonella enterica (Sigma-Aldrich) were resuspended in endotoxin free water to a concentration of 5mg/ml. Where indicated, mice were given a daily IN dose of poly-ICLC (2.0 μg/pup), Pam3Cys (10 μg/pup) or LPS (10 μg/pup) or vehicle control from age 8 to 12 days.

Four day old pups were inoculated IN with a total of ~9,000 CFU of S.pneumoniae suspended in 3μl of PBS using a blunt pipette without anesthesia. On day eight of age, pups were inoculated IN either with influenza A virus/HKx31 strain (H3N2) (2×10²–2×10⁴ TCID₅₀) suspended in 3μl of PBS or an equal volume of PBS. Influenza A virus was grown in the allantoic fluid of 10-day embryonated chicken eggs (B&E Eggs) and stored at -80°C. Viral concentrations for infection were determined by titration in Madin-Darby Canine Kidney cells, as described previously [28].

At age 12 days, pups were euthanized by CO₂ asphyxiation, the upper respiratory lavaged with 200μl of sterile PBS from a needle inserted in the trachea, and fluid collected from the nares. To detect P1547 or its derivatives in lavages, aliquots were plated on 5% sheep blood agar containing neomycin (20μg/ml) or the selective antibiotics listed above to minimize contamination. The limit of detection was 33 CFU/ml unless otherwise noted.

The colonizing dose for strain P2405 was determined using ten-fold serial dilutions in PBS inoculated IN to pups (≥3/dose) given at age 8 days with colonization quantified in lavages at age 12 days. Where specified mice were either sham (PBS) inoculated or given a colonizing dose of strain P2397 at age 4 days to test the effect of pre-existing colonization.

Experiments to detect otitis media were carried out with the same infection schedule as described above. On day 12 of age, pups were euthanized to collect nasal lavages and middle ear lavages. Using a dissecting microscope, both middle ear cavities were washed with 3μl of PBS with a sharp sterile tip after myringotomy. The detection limit for quantitative culture for each sample was 3333 CFU/ml.

Experiments to detect pneumonia were carried out with the same infection schedule as described above. On day 12 of age, whole lung tissue, nasal lavages and blood were collected in order to minimize contamination between sites. The lung tissue was homogenized in 1ml of sterile PBS and aliquots serially diluted for quantitative culture.

Bacteremia was assessed following infection by two different routes. For the intranasal challenge route, four-day old pups were infected IN with S.pneumoniae, followed by either IAV or PBS on day eight of age. The pups were then observed every 12 hours after IAV or PBS inoculation for signs of sepsis (decreased motor activity, shivering or weight loss). Ill-appearing pups were immediately euthanized to obtain nasal lavages and collect blood samples via cardiac puncture for cultured (50 μl volume). The limit of detection to quantify bacteremia was 33 CFU/ml of blood. Pups without signs of sepsis were euthanized on 18 day of age and nasal lavage and blood cultures obtained. Bacteremia was assessed only at the time of sacrifice because the small size of pups precluded repeated sampling of blood for culture.

For the IP challenge route, 13 day old pups were injected with ~450 CFU of S.pneumoniae (~150 CFU of each mutant) suspended in 50μl of sterile PBS. At 14 hours post-infection, pups were euthanized and blood samples were collected by cardiac puncture for quantitative culture.

The experimental schedule was based on a previously described study [14]. The schedule was modified and transmission was assessed at age 12 rather than 14 days to minimize the possibility of transmission from a newly colonized contact mouse to another contact mouse. Except where indicated, one pup in each litter was randomly selected as the index mouse and at age 4 days infected IN as described above. The index mouse was then returned to the litter and housed with the dam and other uninfected pups (contact mice). When the litter was 8 days of age, all pups were inoculated IN either with IAV or PBS. In co-infection studies, all pups within a litter were given IAV to eliminate the potentially confounding effect of viral transmission among pups. To detect bacterial transmission from the index to contact pups, all pups were euthanized at age 12 days and nasal lavages were collected and plated on the selective medium.

The infection schedule was the same as described above for colonization experiments. From day 8–12 of age, nasal secretions were cultured in samples collected by gently tapping (10 times) the nares onto TS agar containing neomycin (20 μg/ml) and catalase (6,300U/plate) (Worthington Biochemical Corporation, NJ) [14]. The sample was then evenly spread across the plate using a sterile swab for quantitative culture. When a mixture of marked strains was used, the colonies on TS neomycin plate were individually patched on selective agar plates to determine the ratio of each mutant.

For evaluating an inflammatory status in the nasal cavity, nasal lavages were stained with the following antibodies: anti-CD11b-V450 (BD), anti-Ly6G-PerCP (BD) and anti-CD45-APC Cy7 (BD) after FcR blocking with anti-CD16/32 (BioLegend). Cells were then fixed with 4% paraformaldehyde until analysis. BD LSR II was used for flow cytometry and analyzed with FlowJo software.

Mann-Whitney U test was used for comparisons between two groups. Kruskal-Wallis test was used for comparisons between three or more groups. The time course of sepsis was compared using the Kaplan-Meier’s log-rank test. All statistical analyses were performed using Graph Pad Prism 6 software.

For calculating the size of the bottleneck, we utilized the mathematical model previously reported by Gerlini et al and Margolis et al. [25][29]. In brief, this model is an approximation of taking out balls from a box containing equal number of three different colored balls, which follows a Poisson distribution. Let k be the number of independent events, the probability that the events would occur k times is calculated by P(k)=e−λλkk!. The coefficient number λ is calculated by P(0) = e^−λ, which is approximated to the ratio of non-affected mice. Let w be the number of bacteria that established a population in each infectious site or the nasopharynx of a new host, also let i be the number of mutants detected in the target site in each experiment. P(i|w,k) follows Bernoulli distribution which is extended probability theory of binomial distribution. By considering each case of i = 0, i = 1, i = 2 and i = 3, each probability is calculated as below.

i = 0; no mutants observed; P(i = 0|w,k) = 0,

i = 1; one mutant observed; P(i=1|w,k)=(13)wk−1,

i = 2; two mutants observed; P(i=2|w,k)=(13)wk−1(2wk−2),

i = 3; three mutants observed; P(i=3|w,k)=1−P(i=0|w,k)−P(i=1|w,k)−P(i=2|w,k) The sum of the probability of k = 0,1,2,3…, P(i|w) is calculated as below. P(i=0|w)=∑k=0∞P(k)∙P(i=0|w,k)=P(k=0), P(i|w)=∑k=0∞P(k)∙P(i|w,k)=∑k=1∞P(k)∙P(i|w,k)fori=1,2,3 Because P(k) becomes negligibly smaller as k becomes larger, the cases of k>3 were excluded from the calculation as below. P(i|w)≅∑13P(k)∙P(i|w,k)fori=1,2,3 Eventually the estimated number of w = 1, w = 2, w = 3… can be calculated by multiplying the number of total mice and P(i|w). We determined the most likely value of w by comparing with the estimated number and the observed number using the Chi-square goodness-of-fit test.