Rewiring capsule production by CRISPRi-based genetic oscillators demonstrates a functional role of phenotypic variation in pneumococcal-host interactions

Abstract

Phenotypic variation is the phenomenon in which clonal cells display different traits even under identical environmental conditions. This plasticity is thought to be important for processes including bacterial virulence^1-8, but direct evidence for its relevance is often lacking. For instance, variation in capsule production in the human pathogen Streptococcus pneumoniae has been linked to different clinical outcomes^9-14, but the exact relationship between variation and pathogenesis is not well understood due to complex natural regulation^15-20. In this study, we used synthetic oscillatory gene regulatory networks (GRNs) based on CRISPR interference together with live cell microscopy and cell tracking within microfluidics devices to mimic and test the biological function of bacterial phenotypic variation. We provide a universally applicable approach for engineering intricate GRNs using only two components: dCas9 and extended sgRNAs (ext-sgRNAs). Our findings demonstrate that variation in capsule production is beneficial for pneumococcal fitness in traits associated with pathogenesis providing conclusive evidence for this longstanding question.

Extended Fig. 1.

Design and testing of extended sgRNAs in S. pneumoniae. (a) Normal CRISPRi uses a sgRNA that contains a 19 or 20 nt long spacer sequence that binds to a complementary DNA target sequence if this sequence also contains a PAM (top). Extended sgRNA’s (ext-sgRNAs) have a 24nt extension at their 5’ end that includes the +1, a PAM and the orthogonal binding site (BS). In the shown example, the spacer sequence is 19nt long targeting the luc gene encoding firefly luciferase (top) while the spacer sequence targeting ext-sgRNAs are 20 nts long (bottom). (b) S. pneumoniae strains harboring constitutively expressed luc and an IPTG-inducible dcas9 together with a constitutively expressed ext-sgRNA were grown in C+Y medium at 37°C in 96-well plates and OD595nm and bioluminescence was recorded every 10 min. Averages of three replicates are shown. Relative light units (RLU) over the optical density (OD) is shown on the Y-axis, time in h on the X-axis. Out of the 6 cloned and tested ext-sgRNAs, 4 showed similar luc repression levels upon dCas9 induction compared to a normal luc-targeting sgRNA. These are ext-sgRNAs containing BS1, BS2, BS3 and BS6. (c) Schematic overview of the three used ext-sgRNAs to construct the CRISPRlator and an example of how the used fluorescent reporters were constructed with a specific BS in their 5’UTR, just downstream of the +1.

Extended Fig. 2.

Characterization of the pneumococcal CRISPRlator. (a) Expression of ext-sgRNA2 represses transcription of the mTurquoise2 reporter (blue) and ext-sgRNA3; ext-sgRNA3 represses mNeonGreen (green) and ext-sgRNA1, and ext-sgRNA1 represses mScarlet-I (red) and ext-sgRNA2, leading to oscillatory behavior (~ symbol). (b) Schematic overview of the here designed and used microfluidics device. The chip, measured from the outer edge of the inputs/output, is 14.3 × 6.8 mm. Bacteria were injected via the top inlet and C+Y media was pumped through the middle inlet at a flow rate of 0.5 ml/h. The bottom left inlet was blocked. The single waste outlet is shown on the right. Scale bar = 1.43 mm. (c) Zoom in on a mother machine lane within the microfluidic device. Mother machine channels are 1.7 µm wide, 1.25 µm high and have 3.4 µm wide side channels of 250 nm height to increase nutrient/waste flow. The main feeding/flow channel is 20 µm high and 60 µm wide. (d) Mean expression of a total lineage of cells over time in one mother machine lane shows red-green-blue oscillation over time. (e) Circular lineage tree of the same population shown in (d), where color intensity is the mean fluorescence intensity of each single cell in the lineage (left-right: mScarletI, mNeonGreen and mTurquoise2). The ancestor cell is positioned in the middle of each tree.

Extended Fig. 3.

Capsule production correlates with mScarlet-I expression in the CAPSUlator. (a) Immunostaining of the capsule shows capsule production in wild-type D39 and absence of capsule in Δcps cells. There is no (mNeonGreen, mTurquoise) to very faint (mScarlet-I) spectral overlap between the immunostaining and the channels used for visualizing the CAPSUlator. Scale bar 5 µm. (b) Fluorescence output and immunostaining of the CAPSUlator, CAPSUlator Δcps, CAPSUlator-OFF & CAPSUlator-ON (see fig. 2). The CAPSUlator shows heterogeneous production of capsule and the three fluorescent proteins, CAPSUlator Δcps shows heterogeneous production of the three fluorescent proteins, but no capsule. In the CAPSUlator-OFF, the expression of mScarlet-I and capsule are constitutively repressed, while in CAPSUlator-ON, the ext-sgRNA cycle remains intact, but uncoupled from the repression of mScarlet-I and capsule, leading to constant expression of mScarlet-I and capsule. Scale bar 5 µm. (c) Correlation between capsule production measured as fluorescence intensity of the immunostaining around the single cells, and the intensities of the three fluorescent signals (mScarlet-I, mNeonGreen and mTurquoise2) in single cells. mScarletI and capsule intensities correlate (R²=0.52). (d) Autocorrelation function of the fluorescence intensities in single CAPSUlator cells tracked in one mother machine lane (oscillation time: 7 ± 3 generations, 382 ± 738 minutes).

Fig. 1.. Design and characterization of a three-node ring oscillator ‘CRISPRlator’ using extended sgRNAs.

(a) Schematic overview of the use of ext-sgRNAs to construct a three-node ring oscillator. Ext-sgRNA1 has a spacer sequence (sgRNA6) that is complementary to binding site 6 (BS6) of the non-template strand of the DNA sequence encoding ext-sgRNA2. Ext-sgRNA2 has a spacer (sgRNA3) targeting BS3 present on the DNA sequence of ext-sgRNA3. Ext-sgRNA3 has a spacer (sgRNA2) targeting BS2 present on the DNA sequence of ext-sgRNA1. For clarity, only the variable 5’ region of ext-sgRNAs sequences is shown (the dCas9-binding handle and transcription terminator are not shown). (b) Graphical representation of the three-node ring oscillator called CRISPRlator. Expression of ext-sgRNA1 represses transcription of the mScarlet-I reporter (red); ext-sgRNA2 represses mTurquoise2 (blue) and ext-sgRNA3 represses mNeonGreen (green). As the ext-sgRNAs also repress each other’s transcription (a), this GRN is expected to periodically express the three reporters in the order red->green->blue. (c) Snapshots of a typical microfluidics experiment of the CRISPRlator strain. Scale bar 2 µm. (d) Autocorrelation function (ACF) of the CRISRlator cells grown in mother machine channels. ACF calculated in one mother machine lane for individual cells (thin lines) and averaged (fat line with black outlines) shows oscillations for all three fluorescent signals. (e) Progression of cell length (left) and progression of normalized fluorescent signals (right) of one individual mother cell in the mother machine over time.

Fig. 2.

CAPSUlator design and characterization. (a) Top: capsule operon and genetic context in wild type D39V. Bottom: CAPSUlator construct placed in four genetic loci on the chromosome. (b) Gene regulatory circuits of the CAPSUlator and control strains. The ~ symbol indicates that the three-node oscillatory network is intact. The ⊣ symbol signifies repression and the _/- symbol indicates the ring-node network is interrupted. (c) The CAPSUlator shows heterogeneous production of capsule and the three fluorescent proteins. Capsule production correlates with the production of mScarletI (see also Extended Fig. 3c). (d) Left: snap-shots of a microfluidic time-lapse microscopy movie of the CAPSUlator. Scale bar 2 µm. Cells producing mScarlet-I produce capsule: the cells occupy more space, thereby fitting only one cell in the width of the channel, while cells mainly producing mNeonGreen and mTurquoise2 can fit two cells in the width of a channel and are not growing as a chain. Right: schematic representation.

Fig. 3.

Characterization of synthetic GRNs driving pneumococcal capsule production in traits associated with virulence. (a) Pneumococcal biofilm formation was measured by growing strain in microtiter plates at 34°C. After 6h, biofilm formation to the wells was quantified using crystal violet staining (see Methods). The amount of biofilm formed by each strain was compared to wild type S. pneumoniae D39V using a Wilcoxon signed rank test. (b) The ability of the engineered CAPSUlator strains to adhere to human nasopharyngeal epithelial Detroit-562 cells was tested by infecting a monolayer of cells at an MOI of 5. After 1h of incubation at 37°C the non-adherent and adherent bacteria were enumerated by plating (see Methods). The ratio of adherent vs non-adherent bacteria is shown and compared to wild type D39V using a Kruskal-Wallis test. (c) Bacterial survival during starvation was tested by resuspending exponentially growing cells in 1 x PBS followed by incubation at 25°C. Viable bacteria were quantified by plating and colony counting (see Methods). After 24h of starvation, all synthetic GRNs except for the CAPSUlator-ON strain showed significantly reduced survival compared to wild type D39V (Mann-Whitney test). (d) Heterogeneous pneumococcal capsule production is beneficial for in vivo colonization compared to homogenous capsule expression. Four-day-old mouse pups were inoculated intranasally with 10⁵ CFU of pneumococci and after 24h sacrificed and bacterial loads were enumerated in nasal lavages. There was no statistically significant difference between wild type and the CAPSUlator strain, while all other GRNs (CAPSUlator-Δcps, CAPSUlator-OFF, CAPSUlator-ON) colonized worse than the CAPSUlator (Mann-Whitney test).