Exploration of Bacterial Bottlenecks and Streptococcus pneumoniae Pathogenesis by CRISPRi-Seq

Abstract

Streptococcus pneumoniae is an opportunistic human pathogen that causes invasive diseases, including pneumonia, with greater health risks upon influenza A virus (IAV) co-infection. To facilitate pathogenesis studies in vivo, we developed an inducible CRISPR interference system that enables genome-wide fitness testing in one sequencing step (CRISPRi-seq). We applied CRISPRi-seq to assess bottlenecks and identify pneumococcal genes important in a murine pneumonia model. A critical bottleneck occurs at 48 h with few bacteria causing systemic infection. This bottleneck is not present during IAV superinfection, facilitating identification of pneumococcal pathogenesis-related genes. Top in vivo essential genes included purA, encoding adenylsuccinate synthetase, and the cps operon required for capsule production. Surprisingly, CRISPRi-seq indicated no fitness-related role for pneumolysin during superinfection. Interestingly, although metK (encoding S-adenosylmethionine synthetase) was essential in vitro, it was dispensable in vivo. This highlights advantages of CRISPRi-seq over transposon-based genetic screens, as all genes, including essential genes, can be tested for pathogenesis potential.

Keywords: CRISPRi-seq; Streptococcus pneumoniae; bacterial pathogenesis; bottleneck; influenza A virus superinfection; pneumonia.

Figure 1.. A doxycycline-inducible CRISPRi system in S. pneumoniae.

(A) The two key elements, dcas9 and sgRNA, were integrated into the chromosome of S. pneumoniae D39V and driven by a doxycycline-inducible promoter (P_tet) and a constitutive promoter (P3), respectively. With addition of doxycycline, dCas9 is expressed and binds to the target under the guidance of a constitutively expressed sgRNA. The specific dCas9-sgRNA binding to the target gene acts as a transcriptional roadblock. In the absence of the inducer, the target gene is transcribed. (B) The CRISPRi system was tested by targeting luc, which encodes firefly luciferase. The system was induced with doxycycline at different concentrations. Luciferase activity (RLU/OD) and cell density (OD595) were measured every 10 minutes. Top panel shows the growth and bottom panel shows the luciferase activity. The values represent averages of three replicates with SEM. (C) Reporter strain to assess in vivo activity of the doxycycline-inducible CRISPRi system. Strain VL2351 constitutively expresses mNeonGreen and mScarlet-I, and mNeonGreen is targeted by the sgRNA. Bacteria were collected from blood of mice on control or doxy-chow at 48 hpi and imaged with confocal microscopy in both the red and green channels. (D) Bacterial load at both lung and blood was quantified. Each dot represents a single mouse. Mean with SEM was plotted. There is no significant (NS) difference between the bacterial load in control- and doxycycline-treated mice (Mann Whitney U test).

Figure 2.. Workflow for construction of the pooled doxycycline-inducible CRISPRi library.

(A) 1499 sgRNAs were selected (see STAR methods), targeting 2111 genetic elements out of the 2146 in S. pneumoniae D39V. (B) The vector for sgRNA cloning, named pPEPZ-sgRNAclone, was designed to enable high efficiency Golden Gate cloning, monitoring false positive ratio, and construction of Illumina library in a one-step PCR. SpecR is the spectinomycin resistant marker; NGS indicates key elements which allow construction of an Illumina library by one-step PCR; P is the constitutive promoter which drives the expression of sgRNA; mCherry encodes a red fluorescent protein placed in the base-pairing region of sgRNA and flanked by a BsmBI site on each end; handle and ter represent the dCas9 handle binding region and terminator of the sgRNA. E. coli with the pPEPZ-sgRNAclone form red colonies resulting from the expression of mCherry. BsmBI digestion of the vector produces ends that are compatible with the sgRNA oligo annealing in (C). (C) Forward and reverse oligos were designed for each sgRNA containing 20 bp complementary to sgRNA and 4 nt overhangs compatible with the BsmBI digested vector. The oligos were annealed and pooled together followed by 5’ phosphorylation. (D) Ligation product of the digested vector (B) with the sgRNA annealing (C) was transformed into E. coli. E. coli transformed with the vector containing the sgRNA show white colonies due to replacement of mCherry with the sgRNA. 70,000 E. coli colonies were pooled together, and plasmids were purified and serve as an sgRNA reservoir. (E) Pooled plasmid library was transformed into a S. pneumoniae.

Figure 3.. Fitness evaluation of CRISPRi targets under laboratory conditions.

(A) Workflow of CRISPRi-seq. The CRISPRi libraries were cultured in C+Y medium in the absence (CRISPRi-OFF) or in the presence (CRISPRi-ON) of 10 ng/μl doxycycline or 1 mM IPTG. Bacteria were collected after approximately 21 generations of growth. Genomic DNA was isolated and used as a template for PCR. The forward oligo binds to Illumina amplicon element read 1 and contains the Illumina P5 adapter sequence; the reverse oligo binds to read 2 and contains the P7 adapter. Index 1 and index 2 were incorporated into the forward and reverse oligos respectively, for barcoding of different samples. (B) Violin plots showing the distribution of sgRNA abundance in each sample. ‘−’ represents control samples without inducer; ‘+’ represents induced samples. The abundance of sgRNA =1499*(counts of sgRNA)/(total counts of all sgRNAs). (C) Correlation of the fitness of targets evaluated by IPTG-inducible and doxycycline-inducible libraries. The log2FC, calculated with DEseq2, represents the fold change of sgRNA frequency between the control sample and induced sample. (D) Refinement of essential and non-essential genes of S. pneumoniae D39V by CRISPRi-seq. The sgRNAs were classified according to the number of their targets. 1 gene represents the sgRNAs targeting single gene operons; 2 represents two gene operons; >=3 s represents three or more gene operons. See also Figure S1.

Figure 4.. Exploring bottleneck sizes during infection using CRISPRi-seq.

(A) Workflow of fitness cost and bottleneck evaluation in a mouse pneumonia model by CRISPRi-seq. (B) Bottleneck size of lung samples at 24 hpi. 11 mice were treated with control chow, and 10 mice were treated with doxycycline chow. The horizontal red dash line marks 14,990 bacterial cells, which is a 10-fold theoretical coverage of the CRISPRi library. The red asterisks point to mouse #7 and mouse #9 in the control group. The bottleneck size of these two mice is lower than 10-fold of the library diversity. (C) Bottleneck size in lung and blood at 48 hpi. The black asterisks point out the lung samples without successful collection of bacterial samples, which include mice without doxycycline treatment #1 and #8, mice treated with doxycycline #2-dox, #3-dox, #7-dox, and #11-dox. (D) The number of bacteria barcoded with different sgRNAs in the control group (no doxycycline treatment) was calculated according to the bacterial load and sgRNA abundance in the population. Violin plots show the distribution of bacteria number in the lung samples (left panel) and blood samples (right panel), each dot represents one bacterial variant. Notice that some mice were not shown here, because the total bacterial load was below the limit of detection and the bacterial numbers of each variant could not be calculated. See also Figure S2.

Figure 5.. CRISPRi-seq identified PurA as important for infection.

(A) Comparison of fitness cost of gene depletion by CRISPRi by different sgRNAs between the mouse lung infection model at 24 hpi and C+Y medium. The difference was shown as the log2 fold change between the two conditions by DEseq2 analysis and the P-values are adjusted by FDR. The highlighted sgRNAs were selected for follow-up studies. (B) Growth of the deletion mutants and the wild-type D39V strain in C+Y medium. Cell density was determined by measuring OD595nm every 10 minutes. The values represent averages of three technical replicates with SEM (same for panel E). (C-D) Mouse infection with individual mutants, compared to wild type D39V. Each dot represents a single mouse. Mean with SEM was plotted. (C) The mutants were tested in three batches of infection assays, for each assay the wild-type strain was tested in parallel. Significant difference between D39V and ΔpurA was tested by Sidak’s multiple comparisons test, and the adjusted p value is 0.0158. (D) Validation study of sgRNA0005 targets. The virulence of deletion mutants and complementation strains were tested and compared to wild type D39V. There was a significant difference between the wild-type and ΔpurA strain tested by Kruskal-Wallis test with Dunn’s post-analysis, and the adjusted p-value was 0.0007. Note that ectopic expression of purA complemented the phenotype of the purA deletion mutant. (E) Growth of ΔpurA in blood-like medium lacking adenine, adenosine, guanine, uracil, uridine, xanthine, and complete medium.

Figure 6.. CRISPRi-seq in the influenza A virus pulmonary pneumococcal superinfection model.

(A) Workflow of CRISPRi-seq screen. (B) The bacterial load in the lung at 24 hpi shows no impact of doxycycline. Horizontal bar indicates average. The inoculum used in this model is approximately 5×10⁴ CFU intranasally (i.n.). (C) Comparison of fitness cost of gene between the IAV superinfection model at 24 hpi and C+Y medium. The difference was shown as the log2 fold change between the two conditions by DEseq2 analysis and the P-values are adjusted by FDR. Labelled circles represent sgRNAs targeting genes previously shown to be important for virulence or confirmed in the present study by mutational analysis. (D) Workflow for the confirmation study with individual strain in the IAV superinfection model. (E) IAV superinfection with pneumolysin deletion (Δply), capsule deletion (Δcps), and metK deletion (ΔmetK) mutant, compared to wild type D39V. Each dot represents a single mouse. ** indicate significantly different bacterial loads, p<0.05 Kruskal-Wallis one-way ANOVA. Horizontal bar indicates average. (F) The biosynthetic pathway of S-adenosylmethionine (SAM) synthesis. (G) Growth of the metK deletion mutant in C+Y medium supplemented with different concentrations of SAM. Mean and SEM of three replicates were shown. See also Figure S3.