The S. pneumoniae strains used in this study are serotype 15C clinical isolates 60B (SN69534) and 60CSF (SN69531) (isolated from blood and CSF, respectively), D39 (serotype 2 ST595), and Rx1 (unencapsulated). Cells were routinely grown on Columbia agar supplemented with 5% (vol/vol) horse blood (BA), with or without gentamicin (40 µg/mL), at 37°C in 5% CO₂ overnight (Mclean et al., 2020). Growth assays were performed with pneumococci grown in a chemically defined medium (CDM) (Kloosterman et al., 2006) comprising SILAC RPMI 1640 Flex Media, no glucose, no phenol red (Sigma), supplemented with amino acids (Table 1A), vitamins (Table 1B), uracil (0.01 mg/mL), adenine (0.01 mg/mL), choline chloride (0.005 mg/mL) and catalase (10 U/mL), and either 0.5% glucose, galactose, lactose, raffinose or melibiose.

To conduct MLST analysis, the aroE, gdh, gki, recP, spi, xpt and ddl genes of the blood strains were PCR amplified, purified using the QiaQuick Purification Kit (Qiagen), and sequenced as described previously (Enright and Spratt, 1998). Sanger sequencing was conducted by the Australian Genome Research Facility (AGRF). The sequence types (ST) of the pneumococci strains were determined by searching the MLST database (https://pubmlst.org/) for matching allelic profiles.

Strains were grown in flat-bottom 96-well microtiter plates (Costar) with a final volume of 200 µL as previously described (Minhas et al., 2019). Strains were inoculated at a starting optical density at 600 nm (OD₆₀₀) of 0.05 in CDM supplemented with either 0.5% glucose, galactose, lactose, raffinose or melibiose, before being incubated at 37°C with 5% CO₂. The OD₆₀₀ was measured every 30 min for a total of 24 h in a SpectroSTAR Omega spectrophotometer (BMG Labtech). Assays were conducted in triplicate with at least three repeated independent experiments.

Carbon phenotype microarray (PM) analysis using the PM microplates PM1 and PM2a (Biolog Inc.) was performed on the strains to test for the catabolism of 190 different carbon sources as previously described (Minhas et al., 2019). Every well of the microarrays contained a different carbon source. Briefly, cells were inoculated to a final OD₅₉₀ of 0.06 in the buffer provided, according to the manufacturer’s instructions. This suspension was added in 100 µL aliquots to the wells, and the plate was incubated at 37°C, 5% CO₂. The OD₅₉₀ was measured every 15 min for 24 h. Catabolic activity was measured through colorimetric analysis. During catabolism, NADH is produced which subsequently reduces a colorless tetrazolium dye, resulting in a color change. The level of metabolism for each carbon source was arbitrarily determined by comparison with the negative value.

The formation of biofilms was measured in real time using the real time cell analyzer (RTCA) xCELLigence (Agilent Technologies Inc.) instrument, as described previously for Streptococcus mutans and Staphylococcus spp. (Gutierrez et al., 2016). This instrument detects variation in the impedance signal (expressed as the arbitrary cell index, CI) as bacterial cells attach and form biofilms on the gold-microelectrodes present in the bottom of the E-plates (Agilent Technologies Inc.) Briefly, strains were grown overnight on BA plates at 37°C with 5% CO₂. Cells were harvested and resuspended in 200 µL of CDM + 0.5% glucose or galactose a final OD₆₀₀ of 0.2. To the wells of the E-plate, 150 µL CDM + 0.5% glucose or 0.5% galactose was added before placing the plate in the cradle of the RTCA-DP system, within a 37°C incubator without CO₂ supplementation. An initial baseline impedance reading was taken before the E-plates were removed and 50 µL of bacterial suspension was added to wells for a 1 in 4 dilution, bringing the starting OD₆₀₀ to 0.05. For control wells, an additional 50 µL of CDM + 0.5% glucose or 0.5% galactose was added. E-plates were locked into the cradles of the RTCA-DP platform within the incubator and biofilm formation was monitored for 24 h by recording the impedance signal (CI) every 15 min. Assays were conducted in triplicate with at least two repeated independent experiments. Statistical analysis was performed using two-tailed Student’s t test; P values < 0.05 were deemed statistically significant.

Adherence assays were carried out on the Detroit 562 human pharyngeal cell line as previously described (Trappetti et al., 2011). Briefly, cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum (FCS), 100 U/mL penicillin and 100 µg/mL streptomycin at 37°C in 5% CO₂. Wells of 24-well tissue culture trays were seeded with 2 × 10⁵ Detroit cells in DMEM with 10% FCS and grown for 24 h before inoculation with pneumococci. These strains were grown overnight on BA plates, before being resuspended in CDM + 0.5% glucose or CDM + 0.5% galactose at a final OD₆₀₀ of 0.2. 500 µL of each bacterial suspension was added to the washed Detroit cell monolayers. As a control, each bacterial suspension was added in the same volume to empty wells. After incubation for 2 h at 37°C, wells were washed 3 times with PBS before cells were detached from the plate by treatment with 100 µL 0.25% trypsin-0.02% EDTA and 400 μL of 0.1% Triton X-100 (Sigma). Samples were plated on BA to determine the number of adherent bacteria. Assays were conducted in triplicate with at least two repeated independent experiments. Statistical analysis was performed using two-tailed Student’s t test; P values < 0.05 were deemed statistically significant. Data are presented as percentage of adherent bacteria, calculated using the colony forming units per mL (CFU/mL) of adherent bacteria and the CFU/mL of bacteria from the control wells that underwent the 2 h incubation. Control wells were also used to monitor the growth of bacteria during the incubation time, ensuring that each strain had similar rates of growth.

A FITC-dextran exclusion assay was used to determine the capsule thickness of strains based on previous methods (Hathaway et al., 2012), using FITC-Dextran 2000 kDa (Sigma). Briefly, pneumococci were grown overnight on BA plates in 37°C at 5% CO₂. Cells were washed in PBS, before being resuspended to a final OD₆₀₀ of 0.6 in 1 mL. A volume of 80 µL bacterial suspension was mixed with 20 µL FITC-dextran (10 mg/ml in MilliQ water). Each bacterial suspension was pipetted at a volume of 10 µL onto a microscope slide and a coverslip applied securely. Imaging was conducted on 3 separate days with freshly prepared bacterial suspensions each time. The slides were viewed and imaged using an Olympus FV3000 Laser Scanning microscope with a 60× objective (Adelaide Microscopy). The images were analyzed using Zeiss Zen imaging software. Statistical analysis was performed using the Kruskal-Wallis test; P values < 0.05 were deemed statistically significant.

Selected genes of the raffinose operon from the strains underwent Sanger sequencing through AGRF (Adelaide, Australia). Briefly, samples were prepared for sequencing by PCR amplification of the region of interest using the primers listed in Table 2. PCR products were analyzed by gel electrophoresis, and successfully amplified products were purified using the QIAquick PCR Purification Kit (Qiagen) as per the manufacturer’s instructions. Purified PCR products were evaluated for DNA quantity and quality using a Nanodrop (Thermofisher). Samples were diluted in ultrapure nuclease-free water to a final concentration of either 18 ng/µL or 60 ng/µL depending on the size of the product, according to the Australian Genome Research Facility (AGRF) guidelines.

S. pneumoniae strains were grown to mid-exponential phase in HI broth. Genomic DNA (gDNA) was extracted using the Promega Wizard^® Genomic DNA Purification Kit according to the manufacturer’s instructions, except the DNA pellet was rehydrated in ultrapure nuclease-free water. The lytic enzymes used were lysosome (30 mg/mL), 10 units mutanolysin and 0.1% sodium deoxycholate (DOC). The gDNA was sequenced at the South Australian Genomics Centre (South Australian Health & Medical Research Institute, Adelaide, South Australia) on an Illumina MiSeq (250-bp paired-end reads). Genome assemblies were generated using shovill v1.1.0 (Seemann, 2020a), a bacterial isolate de novo genome assembly pipeline based on SPAdes (Prjibelski et al., 2020). The resulting genome assemblies were annotated using Prokka v1.14.6 (Seemann, 2014). Variant analysis was conducted using Snippy v4.6.0 (Seemann, 2020b) by performing all pairwise combination of aligning reads from one sample to the genome assembly of another sample. Gene presence/absence analysis was determined with a reciprocal-best-hit approach applied to the Prokka-derived amino acid sequences using mmseqs2 v13.45111 (Steinegger and Söding, 2017). A 3-step iterative searching strategy was applied using the easy-rbh workflow of mmseqs2 using sensitivities of 1, 3 and 7 (command line arguments: –start-sens 1 –sens-steps 3 -s 7) using a k-mer length of 6 (command line argument: -k 6).

Strains were initially grown overnight on BA plates at 37°C with 5% CO₂. Cells were harvested, washed, and resuspended in 1 mL of CDM + 0.5% glucose or raffinose to a final OD₆₀₀ of 0.2. Suspensions were incubated at 37°C with 5% CO₂ for 30 min. RNA was extracted using a Qiagen RNeasy Minikit as per the manufacturer’s instructions. Differences in levels of gene expression were determined using one-step relative real-time qRT-PCR in a Roche LC480 real-time cycler, as previously described (Minhas et al., 2019). The specific primers used for the different genes are listed in Table 2 and were used at a final concentration of 200 nM per reaction. Primers specific for gyrA mRNA were used an internal control. Amplification data were analyzed using the comparative critical threshold (2^-∆∆CT) method (Livak and Schmittgen, 2001). Assays were performed in triplicate with a minimum of two independent experiments. Statistical analyses were performed using two-tailed Student’s t test; P values < 0.05 were defined as statistically significant.

Animal experiments were approved by the University of Adelaide Animal Ethics Committee. Female outbred 4- to 6-week-old CD-1 (Swiss) mice were anaesthetized by intraperitoneal injection of ketamine (8 mg/mL) and xylazine (0.8 mg/mL), and were challenged intranasally with 50 µL of bacterial suspension containing 1× 10⁸ CFU in SB as previously described (Minhas et al., 2019). The challenge dose was retrospectively confirmed by serial dilution and plating on BA. At 24 h and 48 h, groups of 6 or 8 mice were euthanized by CO₂ asphyxiation before harvesting the blood, lungs, nasal tissue, ears, and brain. Tissues were homogenized and pneumococci enumerated as previously described by serial dilution and plating on BA plates containing 5 µg/mL gentamicin (Trappetti et al., 2011). Statistical analyses of log-transformed CFU data were performed using two-tailed Student’s t test; P values < 0.05 were deemed statistically significant.

We began characterization of serotype 15C blood and CSF isolates from the same patient, denoted 60B and 60CSF, by performing multi-locus sequence typing (MLST) to determine the sequence type (ST). Both 60B and 60CSF have the sequence type 8711 (ST8711). Additional information provided by the German National Reference Center for Streptococci included the antibiotic sensitivity of the isolates. Both isolates, 60B and 60CSF had the same MIC values for all the antibiotics tested, with all MICs falling below the clinical breakpoint (European Committee on Antimicrobial Susceptibility Testing; EUCAST), indicating full sensitivity for both the isolates (data not shown).

As previous research undertaken in our lab has shown that isolates of the same serotype/ST potentially metabolize the carbohydrate raffinose differentially, we investigated the carbohydrate metabolism capabilities of 60B and 60CSF. Two types of phenotypic microarray plates, PM1 and PM2a, with a total of 190 carbon sources were used to test the isolates’ metabolic ability. A total of 29 of the carbon sources are metabolized by both 60B and 60CSF, 10 carbon sources were metabolized by just 60CSF and the remaining 151 carbon sources were unable to be metabolized (Table 3). In line with prior research implicating an role for raffinose metabolism in tissue tropism, raffinose, and one of its derivatives, melibiose, were metabolized by 60CSF but not 60B. Of the carbon sources that were differentially metabolized by 60B and 60CSF, we analysed the growth of these strains in chemically defined medium (CDM) with the clinically relevant sugars: raffinose, melibiose, lactose and galactose. The growth of 60B and 60CSF were comparable in CDM + glucose (Figure 1), and the growth pattern was consistent in CDM + galactose and CDM + lactose (data not shown). Altought the phenotypic microarray showed that 60CSF was able to metabolize melibiose, there was no growth for either strain in CDM + melibiose, implying that melibiose may not be able to support growth of pneumococci as the sole carbon source in the present conditions (data not shown). Despite this, 60CSF was able to grow in CDM + raffinose, whilst 60B displayed an inability to grow in this medium (Figure 1). This indicates that despite the strains coming from the same patient and being closely related, there is a marked difference in their capacity to utilize raffinose as a carbon source.

Considering that raffinose was only metabolized by and capable of supporting the growth of 60CSF, and not 60B, we performed Sanger sequencing targeting the genes, that are part of the core genome, that are critical for raffinose metabolism to investigate whether they were functional in both strains (Figure 2). The raffinose uptake and utilization operon consists of eight main genes encoding an α-galactosidase (aga), transcriptional regulators (rafS and rafR), an ABC transporter with solute-binding protein and two cognate permeases (rafE, rafF and rafG), a sucrose phosphorylase (gtfA), and a protein of unknown function (rafX). Additionally, further upstream there is a gene encoding an ATP-binding protein (rafK), which is essential for import of raffinose into the bacteria (Rosenow et al., 1999) (Figure 2). The aga, rafR, rafE, rafF, rafG and rafK were sequenced. Each of these genes encode proteins with a critical role in raffinose uptake and utilization (Figure 3). The sequences of all the genes investigated were the identical between the 60B and 60CSF strains. However, for both 60B and 60CSF, there was a SNP in aga relative to the published D39 and TIGR4 genome sequences that resulted in an early stop codon, truncating the Aga protein at 213 a.a., compared with 721 a.a. for native Aga (Supplementary Figure 1). As the gene sequences involved in raffinose uptake and metabolism were the same between 60B and 60CSF, we investigated if the expression of these genes correlated with the capacity to utilize raffinose by performing quantitative real-time reverse transcription-PCR (qRT-PCR).

To determine if the difference in ability to utilize raffinose between the blood and CSF isolates corresponded with raffinose operon gene expression, 60B and 60CSF strains were grown to the same OD₆₀₀ (0.2) in CDM + Glucose and then washed and resuspended in CDM + Raffinose and incubated for a further 30 min. RNA was then extracted, and levels of rafR, aga, rafK, and rafG mRNA, representative of each of the three RafR-regulated transcriptional units, were then measured relative to gyrA mRNA by qRT-PCR. Although rafR was not expressed at detectable levels in either strain (data not shown), the expression levels of the α-galactosidase (aga) and the ATP binding protein component of the transporter (rafK) were significantly up-regulated in the CSF isolate compared to the blood isolate (P < 0.05), while the expression level of the putative permease (rafG) was comparable in both strains (Figure 4). These results indicate that, while the component of the raffinose ABC transporter gene (rafG) was equally expressed in both strains, the uptake of raffinose may be limited by the reduced transcription of the gene encoding the ATP binding protein rafK required to energize this transporter.

We next explored whether the two clinical isolates displayed any differences in disease phenotype using a mouse model of infection. Mice were intranasally challenged with 10⁸ CFU of each strain and the bacterial burden in the blood, brain, ears, lungs, and nose were assessed at 24 and 48 h post challenge. Mice challenged with either strain showed no detectable bacteria present in the blood at either time point (data not shown). At 24 h post challenge, there was no significant difference in the relative bacterial burden between the mice infected with the 60B or 60CSF isolates in the brain or lungs (Figure 5). No bacteria for the 60B isolate were detected in the ears at 24 h post challenge (Figure 5), while bacteria were present in the ears of over 85% of the mice for the 60CSF strain (Figure 5; P < 0.0001). Conversely, the bacterial burden in the nose was higher for the 60B strain compared with the 60CSF strain (Figure 5; P < 0.0001). At 48 h post challenge, there were no observed significant differences between the two strains in any niche (data not shown).

Given the observed niche preferences for the strains at the 24 h time point, where the 60B strain was higher in the nose, while 60CSF higher in the ear, we investigated whether these strains produced varied amounts of capsule that may account for these differences. We analyzed the relative capsule abundance by measuring the FITC-Dextran exclusion area of the two isolates, as well as D39 and Rx1 as encapsulated and unencapsulated control strains, respectively. No significant difference in the mean FITC-Dextran exclusion area was observed between the 60B and 60CSF strains (Figure 6). The unencapsulated control strain Rx1 displayed a significantly lower mean capsule area compared with the D39 strain (P < 0.01), which showed a mean area comparable to that of the 60B strain, while there was a minor but significant difference compared with the 60CSF strain (Figure 6; P < 0.01). We next considered whether the difference in tissue tropism could be due to an altered capacity of the bacterial strains to adhere to the nasopharyngeal epithelium. The in vitro adherence of each strain to Detroit 562 pharyngeal cells was explored in the presence of glucose or galactose as the sole carbon source, the latter representing the prevalent sugar present in the nasopharynx (Paixão et al., 2015). No significant difference in the total number of adherent cells was observed between the strains in either carbon source (Figure 7). Together, these data suggest that the differences in tissue tropism are not merely a consequence of altered capsule expression or nasopharyngeal adherence capabilities between the strains. To determine whether the 60B and 60CSF strains displayed a difference in biofilm formation capacity, we performed an assessment of their biofilm formation with the xCELLigence instrument. An analysis of biofilm formation was performed in glucose, the preferred carbon source, and galactose, a carbon source accessible during nasopharyngeal colonization, where biofilm formation is thought to occur. No significant difference in the rate of biofilm formation over a 24h period was observed between the strains (data not shown). Furthermore, although the maximum biofilm formation in 60CSF appeared higher than in 60B, with higher biofilm levels in the galactose medium, this difference was not statistically significant (Figure 8).

As we have previously shown niche-adaptation differences between closely related isolates (Minhas et al., 2019) and isolates from the same patient (Tikhomirova et al., 2021) to be driven by a SNPs in loci responsible for uptake and utilization of raffinose, we performed whole genome sequencing of 60B and 60CSF on the Illumina platform, to determine whether the phenotypic differences observed here were a product of similar genomic events. SNP analysis comparing 60B and 60CSF sequences revealed that no SNPs were present in the well characterized raffinose locus or in any other coding regions. However, reciprocal best hit-based analysis revealed that there may be different fragments of a gene, encoding a putative glycosyl hydrolase family 36 (GH36) alpha-galactosidase, present between the blood and CSF isolates. The nucleotide and amino acid sequences shared no sequence similarity with the aga gene sequenced at the raf locus of these strains. We performed a BLASTn search using the nucleotide sequence for the largest fragment from the CSF isolate. This region showed 100% sequence identity to the latter 75% of an alpha-galactosidase gene that was present as part of the accessory genome in 68 out of 9029 other pneumococcal strains (0.75%), including the S. pneumoniae 4559, EF3030, and 947 strains (Junges et al., 2019; Minhas et al., 2019). Based on the short-read sequencing, obtained with the Illumina platform, the blood isolate appeared to possess two fragments spanning a 594 bp and 967 bp region of the alpha-galactosidase (Supplementary Figure 2), while the CSF isolate possessed one larger gene fragment encompassing 1680 bp in total (Figure 9). The sequencing did not identify regions with similarity to the first 537 bp of this full-length gene in either isolate. However, the fragments in the CSF isolate encompassed the remainder of the gene, while the contigs from the Illumina sequencing did not yield sequences that correspond to the final 132 bp of this gene for the 60B strain. We conducted further analyses of the alpha-galactosidase gene in the S. pneumoniae 4559 blood isolate. The alpha-galactosidase gene, depicted as a pink arrow (Protein ID WP_000158269.1; Figure 10), was present in a ~13 kb region of DNA that possessed a relatively low GC content (33% GC) to that of the upstream and downstream DNA (40.4% GC and 40% GC, respectively), suggesting a possible horizontal gene transfer event may have occurred. However, there were no other genes or features identified nearby that would indicate that this region was a mobile genetic element. This ~13 kb region of S. pneumoniae 4559 contained nine genes in total, including genes with putative functions such as a NPCBM/NEW2 domain-containing glycoside hydrolase family 27 protein (WP_001813616.1), an additional alpha-galactosidase (WP_000687798.1), an alpha-L-fucosidase (WP_000683280.1), two carbohydrate ABC transporter permeases, (WP_001812774.1 and WP_000057512.1), an ABC transporter substrate binding protein (WP_001036442.1), and a rhamnulokinase (WP_000388563.1) (Figure 10). In S. pneumoniae 4559, this region was situated adjacent to the zinc acquisition pathway genes, adcCBA. We then examined whether either the 60B or 60CSF contigs contained the adjacent gene sequences. Both isolates contained contigs with sequence identity to the glycoside hydrolase family 27 and alpha-galactosidase genes.