Loss of Pde1 function acts as an evolutionary gateway to penicillin resistance in Streptococcus pneumoniae

Significance Streptococcus pneumoniae is an important bacterial pathogen responsible for many serious infections worldwide. Infections are often treated with penicillin antibiotics. This provides a selection pressure for the emergence of resistant strains over time, reducing treatment options and threatening patients. Combining lab evolutionary and comparative genomics approaches, we identify unique loss of function mutations in the Pde1 enzyme that lead to penicillin resistance in the absence of classical resistance determinants. We confirm this effect across clinical isolates and characterize the impact of natural genetic variation on Pde1 function. Characterization of end-stage penicillin resistance genes has, so far, not led to effective mitigations. Here, by characterizing the evolutionary events leading toward resistance, we open different possibilities for interventions against resistant S. pneumoniae.


Supplementary Experimental Procedures Bacterial strains and routine growth conditions.
All strains used in this study are given in Table S1. All Streptococcus pneumoniae strains used in this study were derived from D39 Δcps (1), unless otherwise stated.
S. pneumoniae strains were routinely grown in Todd Hewitt broth containing 0.5% yeast extract (THY, BD Biosciences) at 37°C in an atmosphere containing 5% CO2.
coli media contained 1.5% (w/v) agar and selective media contained ampicillin (100 µg ml -1 ). Bacterial growth was routinely monitored as optical density at a wavelength of 600 nm (OD600) measured spectrophotometrically in 1 cm light path cuvettes.

Selection for low-level ampicillin resistant mutants.
To prepare ampicillin slope plates, molten TSA agar was mixed with defibrinated horse blood (5% final concentration). Ampicillin was added to a final concentration of 0.0625 µg ml -1 and 15 ml of the mixture were added to a tilted petri dish (3 o approx.), creating a sloped bottom layer. Once the bottom layer set, each plate had a 20 ml TSAII 5% SB layer not containing any ampicillin poured over it on a flat surface. This layer covered the slope completely, resulting in an even agar plate while keeping the ≈3 o ampicillin gradient intact. For inoculation, S. pneumoniae D39 Δcps cells were grown to mid-exponential phase (OD600 = 0.2-0.8). Cultures were back diluted to an OD600 of exactly 0.2 in THY and 200 µl of the diluted culture were spread on the slope plates using glass beads. Plates were incubated for two days at 37°C in 5% CO2. For each plate, a single bacterial colony that grew beyond the minimal inhibitory ampicillin concentration was picked from the slope plates and cultured. The procedure was repeated multiple times to create 20 independently generated isolates (Sp_2167, Sp_2169-76, Sp_3073-76, Sp_4131-34, Sp_4136-8).

Genomic DNA preparation
Genomic DNA was prepared using phenol-chloroform extraction. For this, S. pneumoniae strains were grown in 20 ml THY to an OD600 between 0.2 and 0.8 and cultures were harvested by centrifugation (10 min, 5000 × g). Pellets were stored at -20°C until further use. Thawed pellets were combined with 500 µl lysis buffer (20 mM Tris-HCl [pH 7.5], 50 mM EDTA, 100 mM NaCl) and 50 µl lysozyme (20 mg ml -1 ) and incubated at 37°C for 1 hour. After incubation, 60 µl of 10% (w/v) N-lauroylsarkosine sodium salt were added and cells were lysed by mixing vigorously. 600 µl phenol were added and the solution was mixed using a vortex before centrifugation at 16000 × g for 5 min. The aqueous phase was transferred into a fresh microcentrifuge tube and mixed with 600 µl phenol:chloroform:isoamyl alcohol (25:24:1, v/v) solution. After another centrifugation step (5 min, 16000 × g), the aqueous phase was transferred into a fresh microcentrifuge tube and spun down again. 450 µl of the resulting aqueous phase was carefully removed and combined with 1/10 volume 3M NaOAc [pH 5.2] and at least 2 volumes of cold isopropanol (-20°C). The tube was inverted until precipitated genomic DNA was visible and could be pelleted by centrifugation (1 min, 16000 × g). The genomic DNA were washed with 700 µl of 70% EtOH, and again spun down by centrifugation (1 min, 16000 × g). All traces of ethanol were removed by air-drying the genomic DNA pellet, before re-suspending the pellet in TE buffer (100 mM Tris-HCl [pH 8.0], 1 mM EDTA) containing RNaseA enzyme (≈13 µg ml -1 ). These preparations were stored at 4 o C before use.

Whole genome sequencing and variant detection
Whole-genome sequencing (WGS) was performed by MicrobesNG To identify large deletions >50 bp in length mapped reads were searched for low coverage regions. Typically a low coverage threshold of 10 was applied, but often identified deletions contained 0 coverage. The expected large deletion of the WT 'Δcps' capsule locus (cps2A-csp2H, covering ref ≈314640-322145 bp) was identified in each case. This served as an internal control and was removed from the analysis.

S. pneumoniae MLST typing
For determination of sequence types (ST) of clinical isolates, the Illumina short reads were assembled into draft assemblies using SPAdes (version 3.15.5, 2) and then the ST was determined using the command line MLST (unpublished software https://github.com/tseemann/mlst) which scans contigs for the seven MLST genes found in the pubMLST scheme (3,4).

Strain construction and molecular cloning
All plasmids and primer sequences used in this study are listed in Table S1.

S. pneumoniae deletion and integration strains
All S. pneumoniae deletion strains were generated using linear PCR fragments as described previously (5). D39 genome sequence information was retrieved using the xBASE database (6,7). Briefly, two ∼1-kb flanking regions of each gene were amplified and an antibiotic resistance marker placed between them using isothermal assembly (8). Assembled PCR products were transformed directly into S. pneumoniae. In all cases, deletion primers were given the typical name: "genedesignation"_5FLANK_F/R for 5′ regions and "gene-designation"_3FLANK_F/R for 3′ regions, antibiotic markers were amplified from ΔbgaA strains using the AntibioticMarker_F/R primers (strains Sp_0005-9). Extracted gDNA from putative deletion strains was confirmed by diagnostic PCR using the AntibioticMarker_R primer in conjunction with a primer binding ≈100 bp 5′ of the disrupted gene; these primers were given the typical name: "gene-designation"_seq_F. Typically ≈200 ng of gDNA was used to generate multiple gene deletions in the D39 Δcps strain background. However, for clinical isolates 2 µg of genomic DNA from previously created marked gene deletions in S. pneumoniae strains D39 Δcps and correct insertion confirmed by streaks on selective media or by diagnostic PCR.
Pfucose::pde1 complementation strains and pde1 point mutations For construction of pCMK20501, plasmid pAKF205 (5) was amplified by PCRs using Pfuc_XhoI_R and Pfuc_BamHI_F. The pde1 open reading frame was amplified from S. pneumoniae strains D39 Δcps genomic DNA with pde1_XhoI_optRBS_F and pde1_ORF_BamHI_R. This generated a ≈20 bp overlap between backbone and insert, which were joined using a NEBuilder® HiFi DNA Assembly Kit. E. coli DH5α competent cells were transformed with the assembled plasmid and selected colonies were screened for correct insertion by colony PCR using primer pairs: pLEM019_seq_F and pLEM019_seq_R, or pLEM023_F and pLEM023_R. Next, the plasmid was tested by restriction digest (XhoI, BamHI) before final confirmation of the pde1 ORF insert by Sanger sequencing.
Single point mutations were introduced into the pde1 ORF by site-directed mutagenesis. For each new plasmid (pCMK20505-10), the template pCMK20501 was amplified by PCR using overlapping primer pairs that introduce the mutation of interest resulting in a circular product. Generally, these primers were designated pde1_'aminoacidchange'_F and pde1_'aminoacidchange'_R. Where necessary PCR products were digested with DpnI to decrease the number of false positives caused by the pCMK2051 template plasmid. PCR products were transformed into E. coli DH5α competent cells. Resultant candidate plasmids were integrity was tested by restriction digest (XhoI, BamHI) and point mutations in pde1 were subsequently confirmed by Sanger sequencing the whole pde1 ORF.
All plasmids were inserted into the bgaA ectopic locus of the S. pneumoniae Δpde1 genome by transformation. Site specific insertion of each construct was confirmed by PCR using primers bgaA_PznDiagnostic_F and pLEM023_F, and by streaking on selective media.

Bacterial transformations
S. pneumoniae transformations were performed as described in (5). Briefly, S. pneumoniae strains were grown to mid-exponential phase in THY and diluted to an OD600 of 0.03. Competence was induced with 500 pg ml −1 competence stimulating peptide (CSP-1), 0.2% BSA and 1 mM CaCl2. Typically, after 15 min, 1 ml of culture was transformed with approximately 200 ng of gDNA or plasmid DNA. Transformants were selected on TSAII overlay plates containing 5 µg ml −1 chloramphenicol, 0.2 µg ml −1 erythromycin, 250 µg ml −1 kanamycin, 200 µg ml −1 spectinomycin or 0.2 µg ml −1 tetracycline as appropriate. Genomic DNA was prepared for mutant strains and correct insertion of the construct was confirmed by PCR.
E. coli competent cells were thawed on ice, carefully mixed with a ligation preparation or isothermal assembly mix and chilled on ice for 30 min. Competent cells were heat shocked at 37°C for 2 min and chilled on ice for 5 min. 450 µl SOC medium (2% tryptone, 0.5% yeast extract, 10 mM NaCl, 2.5 mM KCl, 10 mM MgCl2, 10 mM MgSO4, and 20 mM glucose) were added and cells were recovered for 1 h at 37°C, before plating on LB agar containing 50 µg ml -1 ampicillin.

Spot dilutions for antibiotic resistance testing
For antimicrobial resistance testing at low antibiotic concentrations, spot dilution assays were performed. Inoculated to an OD600 ≈ 0.025 from an overnight culture, S. pneumoniae strains were grown to mid-exponential phase (OD600 = 0.2 -0.8).
Cultures were diluted to an OD600 of exactly 0.2 and serially diluted (1:10) to a final dilution of 10 -5 in THY. 5 µl spots of each dilution were spotted onto a range of fresh TSAII 5% SB plates, containing appropriate antibiotic concentrations. Plates were incubated at 37°C in an atmosphere containing 5% CO2 for ≈ 24 h.

Etest MIC testing
For additional antimicrobial resistance testing using commercially available Etest strips (Biomérieux), S. pneumoniae strains were inoculated to an OD600 ≈ 0.025 from an overnight culture and grown to mid-exponential phase (OD600 = 0.2 -0.8). These cultures were diluted to an OD600 of exactly 0.2 and 200 µl spread on a TSAII 5% SB plate. Once dried, an antibiotic Etest strip was applied, and the plate incubated at 37°C in an atmosphere containing 5% CO2 for ≈ 24 h.

Growth curves
Starter cultures were prepared by inoculating 20 ml of THY to an OD600 ≈ 0.025 from an overnight culture and resulting S. pneumoniae strains were grown to midexponential phase (OD600 = 0.2-0.8). To begin the growth curve, starter cultures were diluted to an OD600 of 0.025 in 20 ml THY, incubated at 37 o C in 5% CO2 and the OD600 was recorded every 25-30 min over 6 -7 hours, or until the strains have reached 'stationary' growth phase.

Image Analysis
For cell shape and morphology analysis images were analysed using the ImageJ Statistical analyses were applied to the datasets within the MIcrobeJ software, first running a Shapiro-Wilk test for normal distribution, before applying a t-test or MannWhitney test as appropriate. A p-value cut-off of 0.05 was applied as the value of significance.

Preparation of whole-cell lysates for cyclic di-AMP quantification
To prepare whole-cell lysates the desired S. pneumoniae strains were grown to an OD600 of exactly 0.5. 20 ml of this culture was harvested at 3900 ×g. The pellet was resuspended in 1 ml of THY medium and 20 µl serially diluted to 10 -6 and 100 µl plated on TSA blood agar (in triplicate) to determine the colony forming unit (CFU) count of each sample. To create whole-cell lysates themselves, the remaining cell suspension was spun down again (16000 ×g) and resuspended in 15 min incubation at room temperature, cleared lysates were checked for complete lysis and for surviving cells by CFU count and by microscopy. Total protein concentrations were determined using a Bradford assay (Bio-Rad), following the manufacturer's instructions. Protein concentrations were calculated using a BSA standard curve using the GraphPad Prism software (version 9.3.1).

Cyclic di-AMP quantification
The c-di-AMP concentrations in cell lysates were tested using the 'Cyclic-di-AMP  To investigate the clonal inheritance of variation identified within Pde1 the phylogenetic ancestry of the isolates was reconstructed using the Interactive tree of life (iTOL) tool (20). The S. pneumoniae PubMLST core genome MLST scheme was used to generate an alignment of concatenated nucleotide sequences, and this was used to construct a Neighbor-Joining (NJ) phylogenetic tree.
The full resolution of the tree can be found here: To demonstrate the independent emergence of the resistance associated mutations across the phylogeny the ancestral character states at the four mutation sites were reconstructed using the R package, Phangorn (21). The mostparsimonious reconstruction (MPR) attempts to predict the ancestral states within a phylogeny that minimises the total number of character state changes that are required to describe the states observed at the tips of the phylogeny. The MPR for the four mutation sites was predicted using the accelerated reconstruction method (ACCTRAN) (22) and are shown in Figure S8 A-D. The parsimony score, that is the minimum number of changes required to describe the data for a given phylogeny, for the mutations A78S, Q339H, R549C, and T594I are 4145, 4999, 4386, 5351, respectively.

PBP allele diversity measurements
The Genome Comparator plugin in PubMLST (3, 4) was used to identify alleles for the six PBP loci in each S. pneumoniae isolate (n = 7169) previously analysed. For each of the four characterised pde1 mutation hotspots, the isolates were divided into two groups depending on whether they possess the wild-type residue or variant at that position. The number of unique alleles counted in each group was divided by the group size giving the number of alleles per isolate as a measure of genetic diversity. Incomplete or missing sequences were not included in the analysis.

Figure S1
No.          Spot dilution of wild type D39 Δcps and its derivatives on plates containing antibiotics as indicated. Strains were grown to mid-exponential phase, diluted to an OD 600 of 0.2 (0) and further serially diluted to 10 -5 (-5). 5 µl of each dilution were spotted onto TSA plates containing 5% horse blood and the antibiotic concentration indicated in the figure. Plates were incubated for 20-24 h at 37°C in 5% CO 2 . The displayed plates are representative for n ≥ 4 biological repeats with the same effect.
(A) Spot dilutions indicate Pde1 or Pde2 loss of function increases resistance to the cell wall targeting glycopeptide antibiotic vancomycin compared to WT. Note, as for ampicillin (Fig. 1), the effect of pde2 deletion on viability counts were less than the pde1 deletion strain.  The schematic displays the pde1 and pde2 loci with adjacent genes, respectively. Grey and white arrows display coding sequences of the genes indicated above and black arrows indicate putative promoter regions. For all spot dilutions, strains were grown to mid-exponential phase, diluted to an OD 600 of 0.2 (0) and further serially diluted to 10 -5 (-5). 5 µl of each dilution were spotted onto TSA plates containing 5% horse blood and the antibiotic concentration indicated in the figure.
Plates were incubated for 20-24 h at 37°C in 5% CO 2 . The displayed plates are representative for n ≥ 4 biological repeats with the same effect.
(B) Deletion of rplI and rpmE2 does not affect cell morphology. Cells were grown to mid-exponential phase and labelled with TADA for 5 min before imaging on 2% agarose pads. Scale bar: 2 µm. n ≥ 3.
(C) Deletion of rplI and rpmE2 does not affect growth. Strains were grown to mid-exponential phase, diluted to an OD 600 of 0.025 and OD 600 measurements were taken every~25 min until all strains reached stationary growth phase. The growth curves are representative for n ≥ 3 independent repeats.

(D)
The introduction of antibiotic markers into the S. pneumoniae genome does not affect ampicillin resistance. Strains containing all antibiotic markers used in this study were inserted into the bgaA ectopic locus. In all cases these insertions did not impact ampicillin resistance when compared to pde1 loss of function strains.  Phylogenetic data comparing sequence variation in Pde1 to genes and protein sequences found in the same pathway: Pde2 and CdaA. This acts as a comparison figure for the data show in Fig. 3A and Fig. 3B. Data for Pde1 is duplicated from Fig. 3C + Fig. 3D for clarity and easier comparison. Genome sequences of >7200 S. pneumoniae isolates were retrieved from the PubMLST data base (4), alongside their known MIC for penicillin. Isolates were split into groups based on their penicillin resistance profile according to EUCAST breakpoints. 'Resistant' was used for strains with a reported MIC of > 2 mg ml -1 (dark red), 'intermediate' for 2 mg ml -1 ≥ MIC > 0.06 mg ml -1 (red), and 'susceptible' for MICs ≤ 0.06 mg ml -1 penicillin (cyan). For A and B, all mutations found within the dataset were normalised by the number of protein sequences available for each group:       In addition to the laboratory strain D39 Δcps and its derivative R6, pde1 deletion led to increased ampicillin resistance in clinical strains including serotypes covered by the PCV-13 vaccine, serotypes currently prevalent in the world (post-PCV13), and local clinical isolates. Importantly, the effect could also be observed in isolates classified as intermediate or resistant. Serotypes for each strain are given in superscript. Where the strain is part of the Pneumococcal Molecular Epidemiology Network (PMEN), the number in the collection is indicated by a dash following the strain name. Mosaic PBPs were classified as present where one or more PBP had significant sequence variation (>1%). A dash indicates no information. Nonsynonymous mutations in the Pde1 sequence of clinical strains are indicated by single letter amino acid code. Representative spot plates (n ≥ 2) with an increase in resistance upon pde1 deletion are framed in green. Importantly where deleting pde1 did not effect resistance these genomes already contained Pde1 variants which are now known to be loss of function (compare mutations with those tested in Fig. 5).   The data shows higher allele diversity across all pbp genes when strains contain a known pde1 loss-offunction variant(s). Note the R549C variant acts as a control in these plots as it is now known to be fully functional (grey). All loss of function variants have higher degrees of diversity (black) with the greatest change shown for A78S and Q339H. The Genome Comparator was used to identify pbp allele number as a measure of sequence diversity. The PBP allele diversity measurements were split by the Pde1 variant. The allele frequency data used to generate these plots are given in Figure S13.   The D39 Δcps genotype (Δcps2A'-Δcps2H') was excluded from derivative strains for clarity. Strains are ordered as introduced in the manuscript, whilst plasmids and oligonucleotide sequences are shown in alphabetical order. Amp = ampicillin, Cam = chloramphenicol, Erm = erythromycin, Kan = kanamycin, Spec = spectinomycin, Tet = tetracycline, CEP = 'chromosomal expression platform'. S.pn = ORF amplified from S. pneumoniae D39. For oligonucleotide sequences; where restriction enzyme sites have been introduced these are underlined. Any primer containing relevant sequence from an ORF are shown in blue. Overlapping sequences used for isothermal assembly when generating gene knockout or genome integration constructs are shown in bold. Regions where codons have been altered are highlighted in red.