In Vivo Dual RNA-Seq Analysis Reveals the Basis for Differential Tissue Tropism of Clinical Isolates of Streptococcus pneumoniae

Streptococcus pneumoniae is a genetically diverse human-adapted pathogen commonly carried asymptomatically in the nasopharynx. We have recently shown that a single nucleotide polymorphism (SNP) in the raffinose pathway regulatory gene rafR accounts for a significant difference in the capacity of clonally-related strains to cause localised versus systemic infection. Here we have used dual RNA-seq to show that this SNP extensively impacts both bacterial and host transcriptomes in infected lungs. It affects expression of bacterial genes encoding multiple sugar transporters, and fine-tunes carbohydrate metabolism, along with extensive rewiring of host transcriptional responses to infection, particularly expression of genes encoding cytokine and chemokine ligands and receptors. The dual RNA-seq data predicted a crucial role for differential neutrophil recruitment in the distinct virulence profiles of the infecting strains and single cell analysis revealed that while reduced expression of the RafR regulon driven by a single rafR SNP provides a clear advantage for pneumococci to colonize the ear, in the lung it leads to massive recruitment of neutrophils and bacterial clearance. Importantly, the observed disease outcomes were confirmed by in vivo neutrophil depletion showing that early detection of bacteria by the host in the lung environment is crucial for effective clearance. Thus, dual RNA-seq provides a powerful tool for understanding complex host-pathogen interactions and revealed how a single bacterial SNP can drive differential disease outcomes.

ABSTRACT 21 Streptococcus pneumoniae is a genetically diverse human-adapted pathogen commonly carried 22 asymptomatically in the nasopharynx. We have recently shown that a single nucleotide 23 polymorphism (SNP) in the raffinose pathway regulatory gene rafR accounts for a significant 24 difference in the capacity of clonally-related strains to cause localised versus systemic infection. 25 Here we have used dual RNA-seq to show that this SNP extensively impacts both bacterial and

INTRODUCTION
with raffinose as the sole carbon source, correlating with defective transcription of raffinose 68 pathway operons. Remarkably, in serotype 14 ST15, exchanging the rafR alleles between blood 69 and ear isolates reversed both the in vitro and in vivo phenotypes (Minhas et al., 2019). Thus, 70 the single D249G SNP in rafR appears to be the determinant of differential virulence phenotype 71 between the blood and ear isolates, which may reflect differential engagement of innate host 72 defences and/or differential bacterial nutritional fitness in distinct host niches (Figure 1). 73 Dual RNA-seq applies deep sequencing to simultaneously quantify genome-wide 74 transcriptional responses of host and pathogen (Westermann et al., 2017;Wolf et al., 2018). 75 This approach offers higher efficiency and more restricted technical bias compared to 76 conventional approaches, such as assaying single species or array-based methods. In the present 77 study, we have used dual RNA-seq analysis to examine host-pathogen transcriptional cross-talk 78 in the blood and ear isolates and rafR-swapped derivatives thereof, during the early stages of 79 infection. Our data strongly suggest that the rafR SNP interacts with the pneumococcal genetic  Our previous studies have shown that at 6 h after IN challenge with serotype 14 ST15 S. 87 pneumoniae, the numbers of blood and ear isolates (strains 4559-Blood and 9-47-Ear, 88 respectively) present in murine lungs are similar (10 6 -10 7 CFU per lung). However, by 24 h, 89 the ear isolate had been cleared from the lungs, instead spreading to the ear and brain. In 90 contrast, the blood isolate persisted in the lungs at 24 h, but did not spread to the ear or brain  In order to directly compare pathogen transcriptional responses in murine lung, we listed 116 homologous genes between the two wild type ear and blood isolates and used these genes to 117 visualize the transcriptional response in a principal component analysis (PCA) plot (Figure 2a). 118 Here, the pneumococcal transcriptional response of the ear isolate (strain 9-47-Ear, dark orange) 119 to murine lung infection diverges considerably from the response of the blood isolate (strain 120 4559-Blood, dark purple). Specifically, 76 homologous genes are significantly upregulated in 121 the ear isolate, while 40 genes are upregulated in the blood isolate in the murine lung.

122
Upregulated genes in strain 9-47-Ear include genes involved in carbohydrate metabolism, 123 general stress response and nutrient transporters, while upregulated genes in the blood isolate 124 include genes encoding permeases for small molecules and nisin biosynthesis orthologous 125 proteins.

126
Furthermore, replacing the rafR of the ear isolate 9-47-Ear with the allele from the blood 127 isolate 4559-Blood (designated strain 9-47M) dissociates its transcriptional response 128 considerably from its parental 9-47-Ear strain (Figure 2a   We also identified genes that were commonly up or down regulated between the strains 182 that persisted in murine lungs (4559-Blood and 9-47M) or the strains that were cleared from    (Figure 3a). Specifically, there was a diverging host response to the ear 209 isolate (9-47-Ear, dark orange) and blood isolate (4559-Blood, dark purple). Interestingly, rafR 210 swap in blood isolate background (4559M, light purple) mimics the lung response to the wild 211 type ear isolate (9-47-Ear, dark orange); the two strains harbor the D249 rafR allele.
212 Surprisingly, the rafR swap in the 9-47-Ear background (9-47M, light orange) which harbours 213 the G249 allele, does not drive the host response to mimic those of the wild type 4559-Blood 214 strain (dark purple) that also has the G249 allele, but rather towards a new, third position of 215 genome-wide expression.

216
Genome-wide plotting of the murine transcriptional response to S. pneumoniae strain 9-     is not known whether differential carbohydrate metabolism better adapts these strains to 429 available carbohydrate sources in these niches, or whether altered host pro-inflammatory 430 responses contribute to ascension of the Eustachian tube or penetration of the blood-brain 431 barrier. Unfortunately, the total numbers of pneumococci present in these niches are too low 432 for pathogen-host transcriptomic analyses using available technologies.

433
Intra-species variation in virulence phenotype is a common feature of pathogenic    (Table 1). Cells were routinely grown 466 in serum broth (SB) as required. Bacteria were plated on Columbia agar supplemented with 5% 467 (vol/vol) horse blood (BA) and incubated at 37°C in 5% CO2 overnight.  Blood were defined as having common coverage at least 90% and identity at least 90%. This 517 single-pass alignment was selected onto chimeric genome was selected to minimize false 518 discovery rate. However, due to this approach, we have to adjust the summarizing process, 519 taking into account the overlapping nature of bacterial genes and its organization into operon 520 structures. 521 We then analyzed host and pathogen libraries separately in R (R v3.   Groups of 8 outbred 6-week-old female Swiss (CD-1) mice (32 in total) were anesthetized and 546 challenged with the bacterial suspension as outlined above in the total RNA extraction method.