Modelling invasive group A streptococcal disease using bioluminescence

Background The development of vaccines and evaluation of novel treatment strategies for invasive group A streptococcal (iGAS) disease requires suitable models of human infection that can be monitored longitudinally and are preferably non-invasive. Bio-photonic imaging provides an opportunity to reduce use of animals in infection modelling and refine the information that can be obtained, however the range of bioluminescent GAS strains available is limited. In this study we set out to develop bioluminescent iGAS strains for use in in vivo pneumonia and soft tissue disease models. Results Using clinical emm1, emm3, and emm89 GAS strains that were transformed with constructs carrying the luxABCDE operon, growth and bioluminescence of transformed strains were characterised in vitro and in vivo. Emm3 and emm89 strains expressed detectable bioluminescence when transformed with a replicating plasmid and light production correlated with viable bacterial counts in vitro, however plasmid instability precluded use in the absence of antimicrobial pressure. Emm89 GAS transformed with an integrating construct demonstrated stable bioluminescence that was maintained in the absence of antibiotics. Bioluminescence of the emm89 strain correlated with viable bacterial counts both in vitro and immediately following infection in vivo. Although bioluminescence conferred a detectable fitness burden to the emm89 strain during soft tissue infection in vivo, it did not prevent dissemination to distant tissues. Conclusion Development of stably bioluminescent GAS for use in vitro and in vivo models of infection should facilitate development of novel therapeutics and vaccines while also increasing our understanding of infection progression and transmission routes. Electronic supplementary material The online version of this article (10.1186/s12866-018-1200-1) contains supplementary material, which is available to authorized users.


Background
Group A streptococcus (GAS) is responsible for a wide spectrum of diseases ranging from the mild, such as pharyngitis, to severe, including pneumonia and necrotizing fasciitis [1]. It is estimated that at least 663,000 cases of invasive GAS disease (iGAS) arise globally per year resulting in 163,000 deaths [2], with rates increasing in more recent years.
Mouse models are commonly used to enhance the understanding of iGAS pathogenesis and to assess different therapeutics and vaccines, although the scientific limitations of such models are well-recognised. Recently bio-photonic imaging (BPI) has been introduced to refine models and reduce the numbers of animals used in experiments, providing additional valuable information on the development of disease [3]. Models using bioluminescent bacteria must be carefully characterized, as bioluminescent bacteria can be attenuated following transformation with constructs containing the lux operon. BPI has been used to model nasopharyngeal GAS disease [4][5][6], although there has been limited use in invasive disease [7].
To refine animal models of iGAS using BPI, bioluminescent derivatives of dominant iGAS serotypes are required which produce enough light to be detected in vivo, demonstrate stability, and are not attenuated in comparison to the parent wildtype strains [4][5][6]. Although a commercially available emm49 bioluminescent GAS strain exists [8], the main emm types causing iGAS in the UK are emm1, emm3 and emm89 [9]. In this work we set out to develop strains to model iGAS using bioluminescent derivatives of these leading invasive emm types emm1, emm3, and emm89, employing a molecular strategy similar to that previously applied to emm75 GAS [6].

Results
Light expression, growth and stability of emm1, emm3 and emm89 GAS isolates transformed with a replicative plasmid containing the lux operon Transformation of GAS with the lux operon on a replicative plasmid (pLux) resulted in light production by emm3 (M3pLux) and emm89 (M89pLux) strains, but not emm1 (M1pLux) (Fig. 1a-c). We observed that M1pLux maintained a functional lux operon despite not producing any light, as we were able to extract the plasmid, transform, and successfully express the operon in Escherichia coli (data not shown).
In contrast, M3pLux and M89pLux showed maximal light production after 5 and 6 h growth, respectively, with M3pLux peaking at approx. 3 × 10 5 RLU and M89pLux peaking at approx. 4 × 10 5 RLU (Fig. 1b, c). For both M3pLux and M89pLux, expression of light significantly correlated with optical density during in vitro growth over 8 h, although the relationship was strongest during exponential growth (Pearson's r = 0.82) compared with the overall 8 h time period of culture (r = 0.63) for M3pLux and also for M89pLux (r = 0.96 during exponential growth compared with r = 0.49 overall). Expression of light attenuated in vitro growth by M89pLux but not M3pLux (p = 0.0003 and p = 0.4332, two-tailed paired t-test, respectively).
We observed differences in the in vitro stability of pLux in the absence of antibiotic selection (Two way ANOVA with Bonferroni Post-Test, Fig. 2a-c). Despite not conferring bioluminescence to emm1 GAS, plasmid pLux demonstrated instability in M1pLux GAS in the absence of antibiotic. There was no loss of plasmid by M3pLux within the first 48 h, although all plasmid was lost by 72 h. Plasmid pLux was stable in M89pLux for the first 24 h, but after this time the number of antibiotic-resistant cells declined slowly but steadily to a 4-log difference by the end of the experiment. Taken together, these results suggested that M3pLux and M89pLux might provide useful bioluminescence in vivo in the absence of selection for short (24-48 h) experiments. :lux (red lines) was compared with appropriate wild type parental strains (blue lines) over 8 h and a logistic curve was fitted to the growth data. Light was measured using a luminometer and recorded as relative light units (RLU) and is shown in green. Median and range of 3 biological replicate cultures are shown M3pLux and M89pLux GAS can be visualised noninvasively during lower respiratory tract infection, but are unstable Female CD1 mice age 6 to 8 weeks were infected intra-nasally with each plasmid-based bioluminescent strain to investigate ability to produce a lower respiratory tract infection (Fig. 3a-c). Mice were imaged at 2 and 4 h post infection, and experiments were stopped at 5 h post infection, as the mice reached defined sepsis end points. Visible light signals were seen from mice infected with M3pLux and M89pLux but not M1pLux, consistent with the in vitro findings. In one mouse infected with M89pLux, bacteria were seen by bioluminescence to have disseminated from the lungs (Fig. 3c), and bacterial counts were subsequently recovered from the liver (data not shown). All three GAS strains, M1pLux, M3pLux, and M89pLux resulted in a productive infection of the lower respiratory tract; lungs of infected mice demonstrated an inflammatory response 2 h following infection, with infiltration of the alveoli with neutrophils, macrophages and proteinaceous fluid ( Fig. 3d-f ).
Importantly, however, despite the observed 24-48 h in vitro stability of pLux in the absence of antibiotic, only 60-70% and 50-70% of the M3pLux GAS colonies recovered from lung homogenates retained the pLux plasmid after 2 and 4 h, respectively. The plasmid was even less stable in M89pLux, and was present in only 17-23% and 0.5-27% of the M89pLux GAS colonies recovered from lung homogenates after 2 and 4 h, respectively.

Development and in vitro characterization of stable bioluminescent M89 GAS
To circumvent the problems inherent in plasmid loss, chromosomal integration of the lux operon was employed as an alternate strategy. Targeted integration of the pICL18-Lux plasmid into gene spy0535 was confirmed in M89::Lux :Lux were passaged daily in THB with kanamycin (blue line) and without kanamycin (dashed blue line) to determine stability of the replicative plasmid pLux in vitro and the integrated plasmid pICL::lux. There was a significant reduction in the number of viable bacteria retaining the plasmid when passaged without kanamycin in comparison to the number of bacteria retaining the plasmid with kanamycin. Stability was determined by serial plating onto selective solid media. Median and range of three replicates per time point are shown using PCR as previously described [5]. Primers LF and RR amplified a product of 10 kB in M89::Lux and a smaller product of 839 bp equivalent to the size of spy0535 in the parent emm89 GAS strain (Additional file 1: Figure S1).
We observed that M89::lux exhibited maximal light production (approx. 3 × 10 4 RLU) at 5.5 h (Fig. 1d), but light levels were 10-fold lower than the same strain expressing the lux operon from plasmid pLux. Expression of light significantly correlated with optical density during in vitro growth over 8 h, although the relationship was stronger during exponential growth (Pearson's r = 0.45 [p = 0.0189] over 8 h and r = 0.98 [p = < 0.0001] during exponential growth). Similar to transformation with pLux, chromosomal expression of the lux operon significantly attenuated in vitro growth (p = < 0.0001, two-tailed paired t-test). In contrast to emm89 GAS transformed with pLux, there was no loss of the lux operon by M89::Lux after in vitro growth without antibiotic selection for 168 h (Fig. 2d). Targeted disruption of the Spy0535 locus using plasmid lacking the lux operon had limited effect on growth in contrast to pICL18 Lux (Additional file 2: Figure S2).
Use of stably bioluminescent M89::Lux GAS strain in lower respiratory tract infection M89::lux was administered to mice intra-nasally to produce lower respiratory tract infection. Despite using the same bacterial inoculum as used in earlier studies, a bioluminescent signal was not observed in vivo. This was despite the fact that the infection led to histopathological evidence of lung inflammation (Additional file 3: Figure S3A) and that all M89::Lux GAS retained antibiotic resistance when tested (not shown). However, following a direct intrathoracic infection post mortem using the same inoculum dose, it was possible to detect a bioluminescent signal in lungs that were dissected (Total Flux 182,400 photons/second) thus we concluded that the signal was insufficiently strong to traverse the thoracic cage, despite the burden of bacterial infection at that time point being 2 × 10 8 cfu/ lung (Additional file 3: Figure S3B).

Use of stably bioluminescent emm89 GAS strain in soft tissue infection
Having established that a lower respiratory tract infection may be difficult to monitor non-invasively, M89::Lux was then assessed in a soft tissue model of invasive disease. Following intramuscular infection with M89::Lux a bioluminescent signal was immediately observed in the thigh of infected animals ( Fig. 4a) that demonstrated significant correlation (r 2 = 0.88, p < 0.05 logistic regression) with the actual bacterial burden quantified from the tissue (Additional file 4: Figure S4). Although light was also observed 24 h following infection, the bioluminescent signal no longer correlated with Fitness burden of bioluminescence expression by emm89 GAS In previous work [5] we determined that expression of light by GAS derivative M75::Lux conferred an overall fitness burden that was manifest in vivo as a reduced growth rate and reduced competitive index when compared with the parent strain. Accordingly, mice were infected intramuscularly with experimental inocula containing a 50:50 mix of M89::Lux and M89 wild type GAS. After 24 h, mice were euthanised and the ratio of M89::Lux to parent emm89 strain was calculated (by replica plating of colonies onto selective media) from both the site of infection, and also sites of dissemination including the local draining lymph node, blood, spleen, and liver ( Table 1). Expression of lux by M89::Lux conferred a measurable fitness burden that could be detected at most sites of infection; indeed there was limited systemic dissemination by the bioluminescent strain to the liver and blood ( Table 1). The fitness burden of the lux operon in both M89::Lux and M75::Lux GAS was also observed using larvae of the Greater wax moth Galleria mellonella as a surrogate host although did not reach significance (Additional file 5: Figure S5). The ability of different bioluminescent strains to be used in invasive models of GAS infection was assessed by comparing dissemination of M89::Lux to the ipsilateral lymph node, blood, spleen and liver, with the previously characterised M75::lux pharyngitis strain [5,6]. Female CD1 mice aged 6-8 weeks were infected intramuscularly with the bioluminescent strains M89::Lux and M75::Lux. At 24 h following infection mice were euthanised and the following organs harvested: thigh, ipsilateral inguinal lymph node, spleen, liver and blood to assess the bacterial burden in the tissues. Bioluminescence of colonies cultured on solid media from the tissues was analysed, to assess the stability of the Lux chromosomal construct in the strains. M89::Lux demonstrated increased stability at the site of infection and at sites of dissemination in comparison to the M75::Lux strain. Interestingly, in the absence of competition from the parent strains, both the bioluminescent emm89 and emm75 strains disseminated  Table 2).

Discussion
Bioluminescence imaging has refined animal models, allowing infections to be monitored non-invasively and in real-time, reducing the number of animals required to carry out experiments and providing further information regarding the development of disease [9]. In this work, we have developed bioluminescent derivatives of invasive GAS and assessed these in different models of infection.
The bioluminescent strains developed in this work demonstrated light production that correlated with bacterial quantities in vitro, a central requirement for use of BPI in modelling infection. Although the multi copy plasmid constructs used in our work conferred greater brightness to emm3 and emm89 GAS, these derivatives were suitable only for very short-term studies due to poor stability. Emm1 GAS failed to produce light in vitro or in vivo despite confirmation that the bacterial lux operon construct was functionally intact, suggesting that the environment created by emm1 GAS might not support expression of the lux operon for reasons that are not fully understood at present. Others have also reported reduced light production in emm1 GAS using the lux operon although constructs transformed using the firefly luciferase (FFluc) can produce light in emm1 GAS [10]. FFLuc constructs require exogenous administration of a substrate, luciferin, for light production whereas the lux operon encodes the genes for the luciferase enzyme and recycling of the aldehyde substrate. As such, it could be that substrate expression or availability is reduced in lux-transformed emm1 GAS. Although use of FFLuc may circumvent this, the need for an exogenous substrate is more challenging in vivo if an injectable agent such as luciferin is required, since our own imaging studies have shown that mice infected via the respiratory tract can auto-and cross-infect such injection sites readily, simultaneously highlighting the superiority of BPI to demonstrate hitherto unrecognised transmission routes for infection.
Despite the brightness of bioluminescent strains containing the replicative plasmid, pLux, instability during passage made these strains undesirable for use in vivo. Plasmid instability can be overcome using toxin-antitoxin (T-AT) strategies that prevent loss of daughter cells that do not contain the target plasmid [11]. To circumvent problems of plasmid instability in the current work, a more stably bioluminescent emm89 strain was developed through chromosomal integration of the lux operon. As expected from previous studies [5] the strain was tenfold less bright than its counterpart strain transformed with a replicative plasmid, and showed attenuation in growth in broth compared with the parent strain and a control strain with disruption of Spy0535 alone. Previous work to create M75::Lux demonstrated that disruption of the chromosomal Spy0535 locus had no impact on growth or fitness of GAS in vitro or in vivo; although a clear fitness burden was directly related to insertion and expression of the lux operon [5]. We speculate that a similar fitness burden is conferred by the lux operon in M89::Lux although the degree to which this affects use in animal models of infection is not yet clear. M89::Lux was not bright enough to be non-invasively detected in vivo during the early stages of inflammation and was only visualised after exposure of the lungs by dissection. Although there was a significant correlation between bacterial burden and light emitted from the dissected lung, similar to that reported by other groups [11], this mode of measurement was not practical and was only found following a direct intrathoracic injection. We cannot exclude the possibility that imaging of lung inflammation might have been more successful at later time points, or if the bacterial inoculum had been prepared using bacteria at logarithmic growth, given that bioluminescence expression is greatest during late logarithmic growth.
In contrast, M89::Lux was successfully used to model infection in the thigh, and light production in vivo correlated with bacterial counts in the early stages of infection, a feature that is under-reported in the literature. Importantly however, there was a reduction in correlation between bacterial burden and light emitted after 24 h. The possible reasons for this are manifold but include conditions that limit the activity of the lux operon such as anaerobiasis in tissues, or reduced bacterial metabolic activity as a starvation response during GAS deep soft tissue infection. Perhaps unsurprisingly, the expression of the lux operon by M89::Lux conferred a measurable fitness burden in vivo that could be detected at all sites of infection; indeed there was limited systemic dissemination by the bioluminescent strain in comparison to the parent strain. The extent to which such a fitness burden matters largely depends on the model or characteristic evaluated. It seems likely that subtle genetic differences in virulence between isogenic bioluminescent strains may be obscured by such a fitness burden while larger effects such as those conferred by vaccination or antimicrobial therapy could be detected easily.
The M89::Lux and M75::Lux bioluminescent strains were compared in the invasive soft tissue infection model of iGAS. Despite chromosomal integration of the luxABCDE construct, and stability of the M75::Lux isolate in the nasopharynx, it was apparent that some loss of the construct occurred upon systemic dissemination of GAS to distant systemic sites, which was only apparent for M75::Lux and not the newly constructed M89::Lux strain. While this no doubt relates to the use of an otherwise intact plasmid to achieve chromosomal integration, as well as increased host immune pressure during bacterial systemic spread, the stability of M89::lux underlines potential differences between strains depending on the model used and provides a bioluminescent strain that can be more readily used in soft tissue infection models.

Conclusions
We used bioluminescent GAS in in vitro and in vivo models of infection to facilitate the evaluation of novel therapeutics and vaccines while also increasing our understanding of infection progression and transmission routes. Despite some attenuation, the newly derived emm89 bioluminescent GAS represents a clinically invasive strain suitable for use in soft tissue infection studies.

Bacterial strains
The GAS strains used in this study are shown in Table 3. Clinical strains H305 (emm1) and H293 (emm89) have been extensively characterised in previous laboratory studies [13,14] while strain H325, that was isolated from blood as part of routine clinical care, was selected as the most transformable clinical emm3 strain available. GAS strains were cultured in Todd Hewitt Yeast broth (THY), Todd Hewitt Agar (THA) or on Columbia Blood Agar (CBA). Plasmids were propagated in Escherichia coli strain DH5α TOP10 Competent Cells (Invitrogen) and grown in Luria Bertani (LB) broth at 37°C shaking at 200-220 rpm. For stability experiments, bioluminescent strains were passaged daily in the presence and absence of antibiotics and then replica plated onto  [14] CBA and antibiotic selective media. For E. coli, ampicillin (Sigma-Aldrich, UK) was used at 50 μg/ml, while for GAS, kanamycin (Sigma-Aldrich, UK) was used at 400 μg/ml. All strains were grown at 37°C. For animal experiments, GAS were cultured without shaking with 5% CO 2 at 37C overnight and centrifuged (Sorvall RTH 750 Rotor), washed twice in phosphate buffered saline (PBS), and re-suspended in PBS to produce an inoculum of 1 × 10 7 -1 × 10 8 colony forming units (cfu) per mouse. Numbers of viable bacteria within the different inocula were retrospectively assessed through plating serial dilutions of each inoculum onto blood and antibiotic selective agar. Bioluminescence was measured in vitro at regular time intervals as relative light units (RLU) using a bench top luminometer (Modulus).

Plasmids and construction of bioluminescent derivatives
Two plasmids were used to introduce luxABCDE into emm1, emm3, and emm89 strains by electroporation as described previously; [12] pLux (pTHLK) [5] which is a replicative plasmid and pICL18 Lux which acts as a suicide vector following introduction of luxABCDE into the redundant Spy0535 locus by homologous recombination. Disruption of the Spy0535 locus in emm89 was achieved as previously [5] using plasmid pUCMUTΔ0535 [14].

Animal experiments
6-week-old female CD1 (Harlan, UK) were used to model infections. Mice were maintained in HEPA filtered cages with sterile bedding and free access to sterilized food and water. The number of mice per cage was dependent on weight. GLP Mini Fun Tunnels (Lillico) were provided in the cages for environmental enrichment. In vivo experiments were performed in accordance with the Animals (Scientific Procedures) Act 1986, subject to protocols that were approved by the Imperial College Ethical Review Process (ERP) panel and the UK Home Office. At the end of experiments, all the animals were culled under Schedule 1 to the Animals (Scientific Procedures) Act 1986 using CO 2 inhalation and then confirmation of death by dislocation of the neck.

Murine model of lower respiratory tract infection
Following gaseous anaesthesia with isofluorane, CD1 mice were challenged intranasally with 1 × 10 7 -1 × 10 8 cfu per mouse, administered in a volume of 25 μl. Pilot work has shown this volume and bacterial load to be sufficient to cause a reproducible lower respiratory tract infection and lung inflammation. Viable bacteria within inocula were assessed by serial dilutions onto the appropriate media. Animals were imaged at 0 h, 2 h, and 4 h after infection and all experiments terminated when animals reached planned endpoints. Bacterial burden in the nose, lung, spleen, liver, blood, muscle and inguinal lymph node (ILN) was quantified. To determine whether the chest wall attenuated the bioluminescent signal obtained from the strains, a known inoculum of M89::Lux was administered directly into the thoracic cage of CD1 mice immediately after being euthanised.

Murine intramuscular infection
Female CD1 aged 6 to 8 weeks were intramuscularly infected with approximately 1 × 10 8 cfu in a 50 μl volume by injection directly into the thigh [13]. Following infection the animal was allowed to recover and move freely in the cage. Viable bacteria within the inoculum were quantified by plating serial dilutions. Animals were observed for 24 h following infection, euthanised at the defined end point and infected muscle, inguinal lymph nodes, spleen, liver and blood were cultured to quantify GAS. To determine stability of constructs in vivo, bacterial colonies from dissected tissues were plated onto TH agar and bioluminescence quantified after 5 h of growth (Modulus Single Tube Luminometer). The proportion of colonies of S. pyogenes that were bioluminescent (or kanamycin resistant) was quantified to determine the proportion of each strain surviving in vivo.

In vivo assays of fitness burden
For competition assays, GAS strains were cultured overnight, centrifuged (1864×g for 10 min) and washed twice in PBS. The optical density of each strain at 600 nm (OD600) for each strain was adjusted to 1 (7 × 10 9 CFU ml − 1 ), and suspensions of different isolates were mixed 1:1 pairwise and used to inoculate mice intramuscularly. Bacterial colonies from dissected tissues were replica plated onto selective media. Individual colonies were inoculated into TH agar and bioluminescence quantified after 5 h of growth (Modulus Single Tube Luminometer). GAS that were either bioluminescent or kanamycin resistant were quantified to determine the proportion of each strain surviving in vivo. Competitive Indices (CI) were calculated as follows: CI = (mutant output/WT output)/(mutant input/WT input).

Galleria mellonella fitness assay
To determine fitness of GAS strains in an alternate surrogate host with a functioning innate immune system, Galleria mellonella were obtained from a single vendor (Live foods) and infected with 10 μl GAS or PBS (control) into the left hindmost proleg, as per previously described [15,16]. Following inoculation, the infected larvae were incubated at 37°C in 5% CO 2 in petri dishes. The wax worm larvae were infected in groups of 20 and observed for survival with mortality of the wax worm