Dynamic Pneumococcal Genetic Adaptations Support Bacterial Growth and Inflammation during Coinfection with Influenza

ABSTRACT Streptococcus pneumoniae (pneumococcus) is one of the primary bacterial pathogens that complicates influenza virus infections. These bacterial coinfections increase influenza-associated morbidity and mortality through a number of immunological and viral-mediated mechanisms, but the specific bacterial genes that contribute to postinfluenza pathogenicity are not known. Here, we used genome-wide transposon mutagenesis (Tn-Seq) to reveal bacterial genes that confer improved fitness in influenza virus-infected hosts. The majority of the 32 genes identified are involved in bacterial metabolism, including nucleotide biosynthesis, amino acid biosynthesis, protein translation, and membrane transport. We generated mutants with single-gene deletions (SGD) of five of the genes identified, SPD1414, SPD2047 (cbiO1), SPD0058 (purD), SPD1098, and SPD0822 (proB), to investigate their effects on in vivo fitness, disease severity, and host immune responses. The growth of the SGD mutants was slightly attenuated in vitro and in vivo, but each still grew to high titers in the lungs of mock- and influenza virus-infected hosts. Despite high bacterial loads, mortality was significantly reduced or delayed with all SGD mutants. Time-dependent reductions in pulmonary neutrophils, inflammatory macrophages, and select proinflammatory cytokines and chemokines were also observed. Immunohistochemical staining further revealed altered neutrophil distribution with reduced degeneration in the lungs of influenza virus-SGD mutant-coinfected animals. These studies demonstrate a critical role for specific bacterial genes and for bacterial metabolism in driving virulence and modulating immune function during influenza-associated bacterial pneumonia.


Tn-seq Sample Preparation and Illumina Sequencing
Genomic DNA from the frozen pellets from each of the three time points (pre-selection (inoculum) (t1) and post-selection (after infection) at 12 h (t2) or 24 h (t3) pbi) was digested overnight at 37 o C with MmeI (NEB), the 5' phosphate group was removed with Calf intestinal alkaline phosphatase (NEB) after which DNA was extracted with the Geneaid Small Fragment DNA kit and dissolved in H2O. An adapter was ligated with T4 DNA ligase (NEB) onto the overhang left by MmeI after which a PCR was performed with the adapter-ligated samples as template. One primer was complementary to the mini-transposon inverted repeat sequence and one primer was to the adapter (Table S1). The resulting PCR product was 140 bp in length and was amplified with the following parameters: 95 o C for 30 sec, 26 cycles of 10 sec at 95 o C, 25 sec at 55 o C and 45 sec at 72 o C, 1 cycle of 10 min at 72 o C, and held at 4 o C. The PCR product was purified using the Agencourt AMPure XP kit, dissolved in H2O and sequenced in rapid run mode on an Illumina HiSeq 2000 according to the manufacturers protocol (Illumina). A 6-nucleotide barcode sequence was included with the adapter so that harvested libraries coud be multiplexed. Following 30 sequencing cycles, raw data is extracted, split into different samples based on the 6-nucleotidebarcode sequences and stripped from the barcode and four nucleotides of the adapter sequence.
This resulted in 5-15x10 6 pneumococcal specific reads per flow cell lane.

Fitness Calculations
Fitness calculations were performed as previously described (1, 2). Following sequencing, reads were mapped to the D39 genome using Bowtie (3). Bowtie parameters (-m1-n1-best) were set so that reads could contain a single mismatch but were only allowed if they mapped to a unique location. If mapping to multiple sites was possible, the read was excluded from the analyses.
Approximately 8% of the reads had to be discarded because they could be mapped to multiple sites such as endogenous transposon related genes or other repeated sequences (6%) or could not be mapped to anywhere and were categorized as junk sequences (2%). Insertions that mapped to a location within the first 5% or the last 10% of a gene were removed from the analysis to minimize the influence of truncated functional genes. On average, 250 reads were mapped per insertion/time point. Only insertions with >15 reads in the inoculum were included in the analyses because insertions with a low number of reads slightly fluctuate over time and can influence the data disproportionately. The data were normalized to the total number of sequenced reads per time point (normalization factors were between 0.92 and 1.06). Fitness was calculated as the change in the number of reads at a specific location over time (see Main Text). Following fitness calculations, the values were normalized against a set of 'neutral' genes. These genes have no fitness effect and consist of pseudo genes and degenerate transposon related sequences. The same factor was then used to normalize the remaining dataset and make all fitness values relative to the WT D39 background. The normalization factors used for all datasets were small and were between 0.98 and 1.09. Each insertion was used to calculate the average fitness and standard deviation of the gene. A weighted average was used to control for fitness deviations due to insertions with small numbers of reads (<50 reads). This resulted in a small increase in replicate correlation and lower standard deviation.
To determine in vivo fitness and account for random loss of mutants during inoculation, the same proportion of insertion mutants that disappeared during in vivo selection were removed from the total number of insertions for each gene. The resulting set of insertions was then reanalyzed and the fitness recalculated. The resulting fitness (Wi, see Main Text) for each gene represents the growth rate per generation, which enables direct comparisons between experiments. To determine which genes differed with statistical significance, the fitness in influenza-infected mice was compared to the fitness in PBS-infected mice. Statistical significance was established when (i) fitness was composed of at least four data points, (ii) fitness deviated by at least 20%, and (iii) a one sample t-test with Bonferroni correction had a p-value less than 0.05.

Metabolic Starvation
Fig S1 shows the growth of each SGD mutant bacteria and WT D39 in cultures supplemented with lung homogenate supernatants (s/n) from mock-or IAV-infected mice, following 5 h of metabolic starvation or from culture initiation.

Kinetics of Pulmonary Cytokines, Chemokines, and Immune Cells during Naïve infection and IAV-Coinfection with Single-gene Deletion Bacterial Mutants
The absolute log10 picograms ( Figure S7 shows the percent of the lung positive for pneumococcus or neutrophil antigen in serial sections of lungs at 24 h pbi from mice IAV-infected (75 TCID50 PR8) followed 7 d later with 10 6 CFU of wild-type or SGD bacteria.