Secreted Gaussia princeps Luciferase as a Reporter of Escherichia coli Replication in a Mouse Tissue Cage Model of Infection

Measurement of bacterial burden in animal infection models is a key component for both bacterial pathogenesis studies and therapeutic agent research. The traditional quantification means for in vivo bacterial burden requires frequent animal sacrifice and enumerating colony forming units (CFU) recovered from infection loci. To address these issues, researchers have developed a variety of luciferase-expressing bacterial reporter strains to enable bacterial detection in living animals. To date, all such luciferase-based bacterial reporters are in cell-associated form. Production of luciferase-secreting recombinant bacteria could provide the advantage of reporting CFU from both infection loci themselves and remote sampling (eg. body fluid and plasma). Toward this end, we have genetically manipulated a pathogenic Escherichia coli (E. coli) strain, ATCC25922, to secrete the marine copepod Gaussia princeps luciferase (Gluc), and assessed the use of Gluc as both an in situ and ex situ reporter for bacterial burden in mouse tissue cage infections. The E. coli expressing Gluc demonstrates in vivo imaging of bacteria in a tissue cage model of infection. Furthermore, secreted Gluc activity and bacterial CFUs recovered from tissue cage fluid (TCF) are correlated along 18 days of infection. Importantly, secreted Gluc can also be detected in plasma samples and serve as an ex situ indicator for the established tissue cage infection, once high bacterial burdens are achieved. We have demonstrated that Gluc from marine eukaryotes can be stably expressed and secreted by pathogenic E. coli in vivo to enable a facile tool for longitudinal evaluation of persistent bacterial infection.


Introduction
The detection and quantification of bacterial colony-forming units (CFU) within animal infection models is critical for both basic research of host-pathogen interactions and pre-clinical evaluation of antibacterial agents and vaccines. However, longitudinal monitoring of CFU from in vivo models usually is costly and time-consuming, and requires sacrifice of animals at different time points of infection. Using luciferase-expressing genetically engineered bacteria to enable a rapid and non-destructive reporter for infection burden has been applied for a variety of bacterial infections [1]. To date, all the luminescence producing recombinant bacteria available for in vivo studies have used two classes of luciferase, either from luminous beetles including Photinus pyralis firefly luciferase (FFLuc, 62 kDa) [2,3] and Pyrophorus plagiophtalamus click beetle luciferase (CBRluc, 62 kDa) [4] or luciferase from luminous bacteria including Photobacterium and Vibro genera (LuxCDABE, 77 kDa) [5]. On the other hand, luciferase from luminous marine eukaryotes such as Gaussia princeps has not been successfully applied for in vivo bacterial studies.
The Gaussia princeps luciferase (Gluc, 19.9 kDa) from marine copepod catalyzes the oxidation of its substrate, coelenterazine (CTZ), producing blue luminescence concomitantly. As the smallest known luciferase, Gluc in either cell-associated form or secreted form has been used for in vivo monitoring of a variety of mammalian cell behaviors including primary tumor growth [6], early tumor metastasis [7], cell apoptosis [8] and T cells trafficking [9]. Additionally, Gluc has also been used as reporter for in vivo detection of certain nonmammalian organisms such as the fungus pathogen Candida albicans [10] and alga Chlamydomonas reinhardtii [11]. Gluc demonstrates several advantages over other conventional luciferases. First, it is documented that human codonoptimized Gluc is 1000-fold more sensitive than humanized firefly luciferase (FFLuc) or humanized Renilla reniformis luciferase (RLuc, 34 kDa) [12]; Second, Gluc is naturally secreted in active form from Gaussia princeps using its native secretion signal (SS), enabling reporting from both the cells themselves and their extracelluar environment; Third, Gluc exhibits good stability under adverse conditions including low pH, hydrogen peroxide, high temperature and even b-mercaptoethanol [13], making it well suited for reporting from stress-associated in vivo environments that are expected within sites of infection. Last, unlike the firefly luciferase (FFluc) or bacterial luciferase (LuxCDABE) [14,15,16,17,18,19], Gluc does not require co-factors such as ATP or FMNH 2 for its reaction. The independence of Gluc activity from these metabolites avoids declines in detection sensitivity associated with the decreased concentration of bacterial metabolite, occuring upon stationary phase [13,18,19].
Despite the many potential advantages of Gluc-based bacterial reporter systems, successful attempts have only been demonstrated for in vitro studies, such as monitoring Mycobacterium smegmatis (M. smegmatis) in vitro growth under hostile conditions [13] or reporting transcription and secretion of virulence factors in the in vitro culture of Salmonella enterica Serovar Typhimurium (S.Typhimurium) [20]. The trials of using engineered M. smegmatis or Lactococcus lactis (L. lactis) to stably express Gluc are illustrative of the challenge for making in vivo bacterial reporter. Although Gluc was robustly expressed by M. smegmatis in vitro, in vivo imaging of M. smegmatis in mice lungs was not successful [2].The other endeavor of in vivo and ex vivo imaging of Gluc-producing L. lactis did not generate any signal above the background either [4]. In order to prepare a bacterial reporter that could benefit from the sensitivity of Gluc, the robustness of Gluc to adverse conditions, and also enable in vivo bacteria detection using ex situ sampling (eg. body fluid and plasma), we prepared an engineered pathogenic E. coli that can stably secrete Gluc and tested whether Gluc can function as a CFU reporter within a persistent tissue cage model of infection.
The subcutaneous tissue cage infection model [21,22,23] mimics deep skin bacterial infections, facilitating the study of bacterial virulence factors [24,25], cellular immune responses [26,27,28], the evaluations of antibiotic efficacy, and assessment of in vivo emergence of bacterial resistance during antibiotic treatment of persistent infections [29,30,31]. For the mouse tissue cage model, a sterile, perforated teflon cylindrical tube (size: 8 mm outer diameter 6 6 mm inner diameter 6 20 mm length) is implanted subcutaneously (s.c.) on the back of mice to form tissue cages wherein interstitial fluid/tissue cage fluid (TCF) accumulates. Infection is initiated by injection of bacteria into the cage cavity via percutaneous puncture (p.p.). This model enables the study of a persistent infection, for up to 3 weeks, wherein the host cellular immune responses are in effect. As long as the initial inoculum size is controlled, tissue cage infection normally does not spread systemically but instead results in an abscess milieu within the cage [32,33], characterized by a suppurative environment with low pH, high protein content, and stationary phase bacteria present at high density in TCF. In such a milieu, the robustness of Gluc to adverse conditions may be of particular importance. Additionally, this model provides a system to evaluate the stability of Gluc-based CFU reporter in infections of long duration. Longitudinal monitoring of bacteria in tissue cage requires multiple sampling of TCF along the infection. Doing so removes organisms from the active infection thus the sampling volume must be kept to a minimum to avoid altering bacterial growth kinetics in tissue cage. In addition, the abscess fluid (TCF) thickens with the progress of infections, making it difficult to sample a volume greater than 10 ml. Conventional CFU plating assays require at least 10 ml or more TCF for measuring bacterial burden in TCF. Therefore, developing a CFU reporter allowing the use of less TCF for bacterial detection is meaningful and important.
Here we report for the first time that the secretion of Gluc by pathogenic E. coli enables a convenient CFU reporter for bacterial infection in mice. Gluc empowers in vivo imaging of stationary phase bacteria in tissue cage infection. Furthermore, CFU burden in TCF and secreted Gluc activity in TCF correlate to a high degree. Lastly, stationary phase bacterial burden in tissue cage even can be indicated by secreted Gluc in the blood samples of mice bearing tissue cage infection.

Results
Secretion signal PelB derived from pectate lyase B of Erwinia carotovora (E. carotovora) efficiently promotes the secretion of Gluc by E. coli to culture supernatant A total of four Gluc-derived proteins with varied secretion signaling peptides were tested in E. coli for their ability to promote secretion in vitro. These included the full length Gluc which contains the native 16 amino acid N-terminal secretion signal (SS) or Gluc with these residues removed. Two further fusion proteins were designed from the latter construct, Gluc appended to the 21 N-terminal amino acid secretion signal from pectate lyase B (PelB), and also a construct with the 60 amino acid secretion signal of Hemolysin A (HlyA) appended at the C-terminus of the Gluc protein. The cell lysates of E. coli harboring pCOLDI based plasmid encoding these different versions were analyzed by Western-blotting ( Figure 1A and 1B). The majority of overexpressed Gluc is present in the insoluble fraction of cell lysate (comparing lane T and Lane I), yet some over-expressed Gluc is retained in the soluble fraction of cell lysate (comparing lane T and lane S). Gluc without any secretion signal is highly expressed from plasmid (comparing lane B and Lane A for Gluc in Figure 1A), but not found in the culture supernatant (Lane ''-'' in Figure 1C). Similar results were observed with Coomassie Blue staining (figure S1), indicating that Gluc without a secretion signal is not naturally secreted by E. coli. This result highlights the differing efficiency of Gluc secretion between E. coli and Mycobacterium smegmatis (M. smegmatis), noting that almost 100% of Gluc was detected in the culture supernatant of M. smegmatis expressing Gluc without a secretion signal [2]. Gluc with either PelB secretion signal or native secretion signal (SS) is secreted to the culture supernatant to a greater degree than Gluc with HlyA secretion signal (comparing Lane ''H'' with lane ''P'' or lane ''+'' in Figure 1C). However, significant growth lag times were observed for E. coli cultures expressing Gluc with native secretion signal (SS) to reach the same cell density as other E. coli strains (Data not shown), suggesting fitness cost with this construct. Additional bands for Figure 1C lane P could reflect some degradation of the secreted PelB tagged Gluc protein. Nonetheless, the construct was selected for further study due to the greater functional activity of this clone relative to the others investigated.
Integration of Gaussia princeps luciferase (gluc) gene into the E. coli chromosome In order to achieve a stable expression and secretion of Gluc from E. coli, the gluc gene with codon usage optimised for Gramnegative bacteria was integrated site-specifically into the E. coli chromosome under the control of a constitutive promoter. Three different promoters that have been reported to have varying strengths were tested [34]. Also, gluc with either the pelB or the native secretion signal (SS) were examined. The three promoter sequences each fused to tagged gluc ORFs were cloned into pSMM25, a derivative of the vector pKD4 [35], so that they were immediately 59 to a FRT-flanked kanamycin resistance gene.Thus, a matrix of six template plasmids containing either the pelB tagged or the SS tagged gluc gene driven by each of the three different promoters was generated. These plasmids were used as PCR templates for generating the DNA fragments used for site-specific recombination with the chromosome. These DNA fragments were integrated into the chromosome using the l Red recombinase [35]. In the process of cloning the gluc gene driven by either strong or intermediate promoters to make template plasmids, multiple mutations were repetitively observed in the promoter region of these template plasmids (data not shown), suggesting that constitutive expression of secreted forms of Gluc at elevated levels is toxic to E. coli. As represented in Figure 2A, the DNA fragment used for the final homologous recombination contained a moderate promoter driving either the pelB tagged gluc ORF or the SS tagged gluc ORF.
The lacZ locus within the genome was selected as the integration site of the gluc gene based on two considerations. First, LacZ is not required for E. coli replication and pathogenesis, avoiding the fitness cost caused by the substitution of lacZ with gluc. Second, the substitution of lacZ with gluc enables a fast X-gal Blue/White screening for targeted integration of gluc in lacZ locus. Integration of the gluc gene into lacZ locus was performed in both the opposite orientation of the lacZ ORF ( Figure 2B) or same orientation as the lacZ ORF ( Figure 2C). The clones obtained exhibit the lacZ ORF substituted with the gluc gene accompanied with a kanamycin drug selection cassette.
Targeted integration of pelB-tagged gluc into the lacZ locus of E. coli yields stable secretion of Gluc The secretion of Gluc resulting from the pelB tagged gluc gene integrated in the same orientation as the original lacZ ORF is four orders of magnitude higher than the signal from the parental strain ATCC25922 and two orders of magnitude higher than the secretion from the gluc gene integrated in the opposite orientation ( Figure 3 and Figure S2). This is presumably due to transcription from the Lac promoter driving expression of Gluc, in addition to the promoter fused to the gluc gene. Integration of the SS tagged gluc gene into lacZ locus was only accomplished when the orientation of gluc is opposite to the direction of the lacZ ORF ( Figure 3 and S2), despite repeated attempts to integrate this fusion in the same orientation as lacZ transcription. This is consistent with the observation that E. coli expressing SS tagged Gluc from a plasmid grew extremely slowly in liquid culture, suggesting a significant fitness cost of high SS-Gluc expression and/or secretion. All the clones with the SS tagged gluc gene integrated in the orientation opposite to lacZ have a low level of secretion of Gluc (Figure 3 and S2).
We investigated the in vitro stability of selected integrants of Gluc tagged with PelB. First, E. coli clones which were shown to secrete Gluc to a high extent ( Figure 3) were analyzed by in vitro passage experiments. Clones were grown in LB medium without kanamycin selection for one passage until an OD 600 of about three, then diluted 1000-fold for a second passage. This protocol was followed for a total of six passages. Comparing the secreted Gluc activity in the culture supernatant of passage number one with the activity in the supernatant of sixth passage ( Figure 4A,4B and S3) reveals that there is no reduction of Gluc secretion after 6 passages. Moreover, the Gluc secretion for passage number 6 is not dependent on the presence of kanamycin selection in media (Grey bar and black bar in figure 4B and S3B). Hence, the stability of the targeted integration of gluc gene in lacZ locus is independent of kanamycin selection.
The kanamycin resistance gene was then removed from three clones (R8, 618 and 619) showing stable integration of gluc in lacZ locus using the FLP recombinase to generate 6 clones as shown in Figure 4C. In vitro passage experiments were performed on the six clones (as described above) to confirm the stability of the gluc integration after elimination of the kanamycin resistance gene. These studies demonstrated that elimination of the kanamycin resistance cassette did not affect the stability of chromosomal integration of the gluc gene in the lacZ locus (passage 1 and passage 6, grey bar and black bar in figure 4C). Furthermore, the replication of the E. coli recombinant clone 6189 (ML6189) in either LB medium (data not shown) or in Tryptic Soy Broth (TSB) was compared against the parental strain ATCC25922 ( Figure  S4A). No evidence of a fitness cost was observed in these studies. Additionally, Gluc bioluminescence correlates well with bacterial density during both exponential and stationary phase of in vitro growth ( Figure S4B).

Optimization of the luminescence reaction catalyzed by the secreted Gluc
In addition to the in vitro passage test for the stability of chromosomal integration of the gluc gene, we also examined the stability of the gluc integration after in vivo growth of the recombinant E. coli strain. Comparison of the Gluc secretion levels of the recombinant E. coli clone (ML6189) before and after in vivo thigh infection was performed using a Gluc assay kit from New England Biolab (NEB kit, white bars in Figure 5A). Colonies recovered after in vivo infection produced the same level of secreted Gluc activity in the culture supernatant as seen prior to in vivo infection ( Figure 5A), verifying the stability of integration of the gluc gene in lacZ locus during the course of infection.
To maximize the luminescence signal produced by bacterial secreted Gluc, the coelenterazine (CTZ) substrate was assessed over four different concentrations up to the solubility limit (118 mM) of CTZ in biological samples relative to the NEB kit conditions ( Figure 5A). Using a standard integration time of 1000 millisecond (ms), a linear dependence of luminescence produced by secreted Gluc from culture supernatant was observed with substrate concentration. Based on these results, the CTZ substrate concentration was varied as needed to obtain sufficient signal above noise in subsequent. The variation of the integration time from 200ms to 2000ms had no significant effect on luminescence production ( Figure 5B). Further studies using integration time longer than 2000 ms or shorter than 200 ms did not show higher signal intensity than using 1000 ms either (data not shown). These observations in terms of integration time were independent of CTZ concentration (data not shown). Accordingly, an integration time of 1000 ms was chosen for all subsequent Gluc in vitro or ex vivo assays.

In vivo imaging of Gluc generated luminescence
To assess whether the Gluc secreting capability of E. coli recombinant clone could endure a typical tissue cage infection course (greater than a week), mice with tissue cages infected by strain ML6189 were imaged at both 11 and 18 days post-infection, when CFU burdens are fully established at $10 8 CFU/ml tissue cage fluid (TCF). The background level of luminescence signal was determined by imaging mice infected with ATCC25922 parental strain.
Three routes of substrate administration including intraperitoneal (i.p.), tail vein/intravenous (i.v.) and intra-tissue cage (i.t.) were tested using a high concentration of CTZ (4mg/kg body weight), while avoiding solubility issues of CTZ in physiological fluid. Neither i.p. nor i.v. injection of substrate produced sufficient luminescence signal for imaging detection in tissue cages (Data not shown). In contrast, i.t. injection gave rise to prominent luminescence signal immediately after substrate administration ( Figure 6A, t = 30s). The intensity differences that resulted from different routes of substrate administration most likely reflect the impediment to perfusion of CTZ from the periphery to the tissue cage environment.
The time dependence of the luminescence signal was assessed. The most vivid image was observed immediately following substrate administration (t = 30 sec post CTZ injection), followed by a reduced level plateau stage (2-6 min post CTZ injection) that gradually decreases over time (6-30 min post CTZ injection) where the signal decays by ,30-70% with signal in some mice returning to background level ( Figure 6 and data not shown). The rapid decay of luminescence hampers the quantitation of imaging results acquired immediately after CTZ administration due to the time requirement for positioning mice under the camera. Imaging at the plateau of the luminescence reaction (2-5 min post CTZ administration) can decrease the variability of signals among different mice but results in a 2-10 fold reduction in peak signal ( Figure 6B and data not shown), thus reducing the sensitivity of luminescence imaging. In addition, imaging at plateau stage generated a signal to background ratio of one log ( Figure 6B), a window which maybe too small to differentiate tissue cages with log scale difference in terms of CFU burden. Based upon the above considerations, we next measured the secreted Gluc activity ex vivo by taking samples of fluid from the tissue cage.
Secreted Gluc activity in tissue cage fluid (TCF) and CFU burden of the recombinant E. coli are correlated To test whether Gluc secretion induced a fitness cost in vivo, the growth rate and time dependency of CFU burden were assessed for the Gluc secreting strain ML6189 and the parental strain ATCC25922 over 18 days in the tissue cage model. In addition, both CFU in TCF and the secreted Gluc activity present in TCF were monitored simultaneously over 18 days of tissue cage infection to profile the correlation between CFU burden and detection of secreted Gluc. For tissue cages inoculated with 10 5 CFU, both strains demonstrated the same CFU increase (about 3 logs) within the first 3 days and the maximal CFU burden was reached by day 3 (59 hr) post-infection ($10 8 CFU/ml TCF). At this time point, there was no statistically significant difference between the two strains in establishing the infection (ML6189: 13/ 14 mice, 93%; ATCC25922: 6/6 mice, 100%; Figure 7A and 7B). More importantly, the established CFU burden of ML6189 remained at a constant density until day 18, showing that the infection persistence was not impacted by the expression and secretion of Gluc. Similar results were obtained in experiments with a lower initial inoculum (10 3 CFU per tissue cage), where the infections in most mice were fully established with both strains by day 3 (Figure S5A and S5B). Further, both strains had a persistent E. coli burden for the remaining course (18 days) of the tissue cage infection ( Figure S5A and S5B). In all cases, the infections remained localized to the tissue cage and there was no evidence of systemic dissemination (data not shown). Thus, in vivo replication of the Gluc secreting strain is not compromised relative to the parental strain.
For ex vivo measurement of the secreted Gluc in TCF, the equivalent of as little as 1 ml of TCF can be used for quantitation of Gluc activity. In these studies we regularly sampled 20 mL of  TCF to enable simultaneous CFU plating, however luminescence measurements involved sample dilution equating to 1 mL of TCF sample per test well. The background level of Gluc activity in TCF was determined by measuring TCF from mice infected with ATCC25922 parental strain ( Figure 7D). For tissue cage infections initiated with 10 5 CFU, the increase of Gluc activity in TCF correlates with the growth of CFU ( Figure 7A and 7C). First, there is 3 logs increase of Gluc activity within the first 3 days (59 hr), the same magnitude of increase as the CFU density. Second, Gluc activity levels off at day 4 while the growth, as measured by CFU, levels off at day 3. Importantly, Gluc activity persists from the 83 hr time point throughout the course of the infection (18 days). It should be pointed out that the mice that did not show Gluc activity above background across all time points are the same mice which failed to establish a robust infection of $10 8 CFU/mL TCF. For tissue cage infections initiated with 10 3 CFU, a correlation between secreted Gluc activity in TCF and CFU burden in TCF was also observed ( Figure S5A and C) except that the Gluc activity leveled off later than 83 hours post inoculation. Careful investigation of paired CFU and luminescence measurements from individual samples taken during the first three time points (#31 h, Figure 7 and S5) showed that Gluc activity in TCF were above the background level when CFU reaches the density of about 10 6 CFU/mL TCF. The tissue cage used in these studies holds a maximum of 300 ml TCF in vivo, suggesting that the minimal number of bacteria in TCF enabling CFU detection via secreted Gluc is less than or equal to 3610 5 CFU per tissue cage. Furthermore, the ex vivo luminescence signal rises significantly above background after the infection burden is fully established ($ 10 8 CFU/ml TCF or 3610 7 CFU per tissue cage).
Ex vivo measurement of secreted Gluc for established tissue cage infection presented a signal to background ratio of three logs, a quantification window larger than the one generated by imaging. More importantly, measuring samples in vitro with a luminometer equipped with a substrate injector avoids the timing complications encountered in whole-animal imaging, thereby aiding sensitivity of detection. Lastly, ex vivo measurements required much less CTZ substrate than the amount infused into whole animal for imaging experiment. In summary, ex vivo measurements of Gluc in TCF requires sampling of TCF, but provides a sensitive method of detecting secreted Gluc as a reporter for E. coli replication.

Secreted Gluc can be detected in plasma samples and serve as an ex situ indicator for the localized tissue cage infection
We sought to test whether the Gluc secreted by E. coli could be detected in the circulating blood and used for quantification of CFU burden in tissue cage infection loci. Tissue cage infections initiated with either 10 3 CFU ( Figure 8A) or 10 5 CFU ( Figure 8B) were measured for plasmatic Gluc activity at four time points during a tissue cage infection (details seen Experimental Procedures). At an early time point of infection (12hr), there was no significant difference between ML6189 (Gluc secreting strain) infection and ATCC25922 (parental strain) infection in terms of detectable Gluc activity in plasma. However, once the tissue cage infections proceeded into stationary phase at day 3 (59hr), the mean value of plasmatic Gluc activity was about 4-fold higher for the Gluc secreting strain ( Figure 8A, B). The detection of Gluc activity in plasma samples of tissue cage infection proved that bacterial secreted Gluc can serve as a blood indicator for a wellestablished localized infection. For the persistent stage of infection from day 3 to day 18, the plasmatic Gluc activity correspondingly remained above the background.
As expected, the Gluc signals in plasma samples are lower than the ones in TCF for the same time points (Figure 7, 8 and S5). Although this limits the use of blood Gluc activity as a quantitative reporter for bacterial replication in tissue cage, it is useful as a qualitative blood indicator for advanced stages of tissue cage infection. Overall our results demonstrated that there are three ways of using Gluc secreted by E. coli to monitor bacterial replication within a tissue cage, including in vivo imaging, as well as ex vivo measuring of Gluc activity in either tissue cage fluid or plasma samples.

Discussion
Despite the extensive use of Gluc expression in mammalian cells for monitoring in vivo cell behavior [6,7,8,9,36], successful in vivo applications of this reporter for prokaryotic organisms have not been reported to date. Other researchers have explored M. smegmatis and L. lactis systems expressing Gluc, but did not observe luminescence signals above the background for in vivo imaging of bacteria [2,4]. Additionally, for these systems, no Gluc signal was detected using ex vivo assays such as measurement of secreted Gluc in blood samples of the infected animal [2,4]. Herein we report that GLuc expressed and secreted by pathogenic E. coli can be used as a reporter for monitoring localized E. coli tissue cage infection via either in vivo luminescence imaging on whole animals or ex vivo measurement of mouse interstitial fluid samples and blood samples. Compared to the traditional method of CFU quantitation, which requires outgrowth in media over long periods (hours-days), the potential for real time quantitation of in vivo bacterial burden using this luciferase reporter is especially useful.
To produce an engineered E. coli that stably secretes Gluc, it was necessary to optimize both the expression promoter and the secretion tag. Our trials of making plasmids expressing Gluc driven by strong promoters repeatedly induced multiple mutations in the promoter region. In addition, high expression and or secretion of Gluc with its native (eukaryotic) secretion signal by pCOLDI expression plasmid severely affected the growth of E. coil in culture. These results are in contrast with what seen in tumor cells where Gluc with its native secretion signal driven by the strong CMV promoter demonstrated a consistent, high expression and secretion of Gluc [12,37]. The discrepancy between tumor cells and bacteria suggested that bacteria cannot tolerate the high levels of constitutive Gluc expression, particularly when Gluc has its native secretion signal. To find the balance between bacterial fitness and sufficient Gluc expression for in vivo detection, a promoter reported to have moderate strength [34] was selected for driving expression of the E. coli codon-optimized gluc gene. In addition, it was found that plasmid encoding Gluc in Lactobacillus plantarum (L. plantarum) was not stable [4], albeit the decent Hence, in contrast to expression of Gluc from plasmid in either M. smegmatis or L. lactis [2,4], the gluc gene was stably integrated on the E. coli chromosome in the non-essential lacZ locus (Figure 2, 3 and  4). Recent reports indicate that the Gluc native secretion signal negatively affects Gluc protein production and/or activity in E. coli [38,39]. Therefore, the conventional bacterial PelB secretion signal was used for promoting Gluc secretion by E. coli (Figure1).
Traditionally, the N-terminal 22 amino acid leader sequence of PelB directs the protein to the periplasmic space, after which this signal sequence is cleaved. We found that the PelB secretion signal could promote the delivery of Gluc across the E. coli outer membrane to the culture supernatant while retaining high activity of Gluc ( Figure 1C, Figure 3 and 4). The above combination of correct promoter strength, bacterial secretion signal and stable chromosomal integration enabled the isolation of an engineered virulent E. coli strain that stably expresses Gluc within animal infection models.
Our results showed that Gluc expressed by E. coli could serve as a reporter for monitoring tissue cage infection by in vivo luminescence imaging (Figure 6). While this result represents a significant advance, the practical limitations of the imaging experimental procedures, including frequent substrate injection and anesthesia, present challenges in monitoring a large cohort of animals at short time intervals. In addition, it is well-known that the luminescence propagation through living tissue can cause compromised sensitivity, reduced spatial resolution and accuracy [40]. The ex vivo assessment of secreted Gluc in TCF and blood samples as reporter for bacterial burden addresses the above limitations and therefore has significant advantages (Figure 7 and  8). Importantly, the window of signal to background ratio provided by ex vivo measurement of Gluc activity in TCF (3.5-4 logs) is not only bigger than the range (1-1.5 logs) provided by Gluc in vivo imaging ( Figure 6, 7 and S5), but also wider than the window showed by in vivo imaging of FFluc expressing M. smegmatis (2-2.5 logs) [2].
Secreted Gluc generates bright in vivo and ex vivo luminescence during the plateau stage of tissue cage infection ( Figure 6, 7 and S5), in other words, there was no evidence of decreased Gluc luminescence for in vitro ( Figure S4B) and in vivo cultures (Figure 7, S5) that achieved stationary phase, supporting the idea that Gluc is well suited for monitoring infection models focusing on stationary phase bacteria due to the independence of Gluc from bacterial exponential phase metabolites (ATP or FMNH2). As expected, the Gluc reporter enabled CFU quantification with the equivalent of as low as 1 mL of TCF (Figure 7 and S5), demonstrating the required sample volume for Gluc measurement is small, which should minimize the impact on bacterial growth kinetics upon repetitive TCF sampling. Additionally, the difficulty of sampling viscous TCF can be overcome by minimized sampling volumes.
The minimal density of E. coli capable of being detected by ex vivo assessment of Gluc in TCF is around 10 6 CFU/ml TCF, corresponding to an absolute number of #3610 5 CFU per tissue cage (Figure 7 and S5). This detection sensitivity compares favorably to other non-Gluc based luciferase-derived bacterial reporters. For example, it has been shown that the minimal density of Click Beetle luciferase (CBRluc) expressing L. plantarum required for ex vivo detection of CBRluc in feces is 10 5 CFU/100 mg feces [4]. Additionally, the minimal number of FFluc expressing M. smegmatis enabling in vivo imaging is around 6.6610 6 ,1.4610 7 CFU per lung [2].
Our results advance the understanding of using Gluc as a reporter for in vivo bacterial infection. However, some concerns should be resolved to broaden the applications. Reporting CFU via secreted Gluc in TCF requires a burden of $10 6 CFU/mL TCF to enable detection above the background luminescence. Therefore, the CFU burden at 14 hr post infection (hpi) was not well represented by Gluc activity in TCF for all animals sampled (Figure 7 and S5). Individual animals at this time point with Gluc signals above the background corresponded to those animals with $10 6 CFU/mL TCF. This effect is more prominent for the infections initiated with 10 3 CFU than the ones with 10 5 CFU, consistent with the requirement of the lower innocula to grow further to achieve 10 6 CFU/mL. Further optimization of the detection of Gluc in biological samples may be able to address this issue. However, Gluc activity in TCF indeed showed the similar magnitude of increase with CFU burden in TCF for the later time points (59 hr and later, Figure 7 and S5), indicating that the current detection limit does not pose a problem for the established phase of infection, which has frequently been the main focus of tissue cage infection model.
All previous successful trials using secreted Gluc present in blood as reporter for in vivo processes are from oncology studies. Published work on Gluc-secreting human glioma cells and breast adenocarcinoma cells have shown a dynamic range with 2-3 orders of magnitude for quantification of in vivo implanted tumor cells by blood Gluc activity [6,7]. However, such a broad dynamic range was not observed for detecting blood Gluc activity produced from bacteria in our tissue cage studies. This is not particularly surprising since the bacterial infection remained localized to the tissue cage, thereby presenting another permeation barrier for secreted Gluc diffusing from the Gluc production site to the blood stream. An example highlighting the impact of the location of Gluc production lesion on dynamic range is that Gluc secreted by human glioma cells implanted subcutaneously in mice provides a dynamic range of 4 orders of magnitude in blood, while the same cells implanted in mice brain only generates a range of 1 order of magnitude in blood [6]. Further, Gluc secretion by human breast adenocarcinoma cells implanted in mice mammary fat pad produces a dynamic range of about 1.5 orders of magnitude in blood [7]. These results suggest that the signal intensity of Gluc in blood is affected by a variety of disease parameters including the duration and location of the Gluc-producing lesion as well as the level of Gluc expression by the pathogen.
Expression and secretion of Gluc with its native secretion signal at a high level via the strong CMV promoter were tolerated by tumor cells, which may be due to the fact that tumor cells naturally lack many replication checkpoints [6,12]. In contrast, we and others have showed that expressions of Gluc at high levels or with its native secretion signal severely compromise bacterial fitness [4,39]. These results suggest that it will not be straightforward to boost the signal intensity of Gluc in blood via increasing expression of Gluc from bacteria. An alternative approach such as optimizing the detection method of Gluc in biological samples, especially in blood/plasma samples, is indicated. It was observed that the luminescence signal produced by Gluc increases with increasing substrate concentration for bacterial culture supernatant samples ( Figure 5) [2], and this trend was the same for measurements of Gluc in TCF and blood samples (Data not shown). This suggests that using high concentration of CTZ substrate reduces the interference of Gluc activity by other molecules in biological samples. However, solubility limitations of CTZ in biological samples and physiologically relevant buffers restrict the maximal CTZ concentration (e.g. 118 mM in these studies). Therefore, improvements in the physical properties of CTZ may enhance the detection of Gluc in biological samples. To this end, a new chemical variant of CTZ with improved aqueous solubility was recently reported to show enhanced light output for in vivo imaging and ex vivo blood assay, albeit it has a faster luminescence decay rate [41]. Additionally, there have been successes in engineering Gluc for improved kinetics and spectral-enhancements that may provide further sensitivity [39,42,43,44]. Taken together, these advances demonstrate the potential for further optimization of the bacterial secreted Gluc as a fast and sensitive reporter for longitudinal evaluation of bacterial infection.

Ethics Statement
This research was performed by strict following the guidelines of Public Health Service Policy on Humane Care of Laboratory Animals and the mouse research protocol used here was approved by the Institutional Animal Care and Use Committee of AstraZeneca (IACUC Permit Number: 11-01-i). Mice were anesthetized with intraperitoneal (i.p.) injection of 80 mg/kg Ketamine and 10 mg/kg Xylazine during tissue cage surgery, and all efforts were made to minimize suffering.

Bacterial strains and Culture
All strains/clones used and or generated in this study can be found in Table 1 and List S3 of File S1. For assay of Gluc luciferase secretion in vitro, E. coli clones were grown in either LB medium or Tryptic soy broth (TSB) medium for 16 hr at 200 rpm at 37uC. The culture was centrifuged for 2 min at 16,0006g and the supernatant collected. For cultivation of bacteria for mouse infections, E. coli clone ML6189 or ATCC25922 was grown in TSB medium for 16 hr at 200 rpm at 37uC. Then the overnight culture was diluted 1:10 in TSB, the OD 600 determined. The calculated volume of the overnight culture was pipetted into the appropriate volume of saline to obtain the target inoculum of either 10 3 CFU/100 ml per tissue cage, 10 5 CFU/100 ml per tissue cage or 10 6 CFU/100 ml per mouse thigh.
Western Blotting E. coli harboring pCOLDI construct was grown in LB medium at 37uC with shaking at 200 rpm until OD 600 reached 0.5. A portion of bacteria was pelleted (5,0006g, 10 min) to provide the pre-induction sample (''B'' in figure 1). The rest of the culture was cooled down on ice prior to being transferred to 18uC and induced with 1 mM IPTG overnight with shaking at 200 rpm. A second portion of culture was pelleted for the post-induction sample (''A'' in figure 1) and the supernatant saved as the post-induction supernatant. The remaining culture was pelleted, resuspended in lysis buffer (20 mM Tris pH 7.5, 1 mM EDTA, 5% glycerol,) and lysed by French press. A portion of lysate was saved as total lysate (''T'' in figure 1). The remaining sample was subjected to 15,4006g centrifugation for 30 min and the supernatant saved (sample ''S'' in figure 1). The pellet was retained as the insoluble fraction (''I'' in figure 1). Aliquots corresponding to 1610 8 cells were loaded into each well of protein gel and Gluc was detected by Western blotting using a rabbit polyclonal antibody against Gluc (NEB, Ipswich, MA). The post-induction supernatant samples were concentrated by chloroform-methanol precipitation [46]. Aliquots corresponding to the supernatant from 2610 8 cells were loaded into each well of protein gel and detected by Western as described above.
pSMM25 is the derivative of pKD4 [35], for making plasmid pSMM25 from pKD4, an Eco RI site and a Bam HI site were introduced just upstream of the 59 FRT site flanking the kanamycin resistance gene. These restriction sites were created in pKD4 by performing two rounds of mutagenesis using the QuickChange Mutagenesis kit from Stratagene (La Jolla, CA) according to the manufacturer's protocol. Briefly, first the BamHI site was created using mutagenic primers Smm105 and Smm106. The Eco RI site was subsequently introduced using primers Smm107 and Smm108. Following each round of mutagenesis, the DNA was transformed into chemically competent PirPlus DH10bpir116 cells (Open BioSystems, Huntsville, AL) to support replication of the pir-dependent R6kg replication origin of the pKD4 parent plasmid. The presence of the introduced restriction sites were verified by restriction digest of purified plasmid with Eco RI and Bam HI.
Primers, either the pair of FlacZ_ForwardL_MYL and FlacZ_ReverseL_MYL or the pair of OlacZ_ForwardL_MYL and OlacZ_ ReverseL_MYL, were used to amplify gluc (with ss tag or pelB tag) together with a kanamycin resistance cassette from pJWW3 or pJWW6. Integration of either the pelB tagged gluc or the ss tagged gluc into lacZ locus of ATCC25922 chromosome by l red recombinase strategy was performed as described previously [35]. The DNA fragment produced from the primer pair of FlacZ_ForwardL_MYL and FlacZ_ ReverseL_MYL was used to integrate gluc in the same orientation as the original lacZ ORF. The DNA fragment produced from the primer pair of OlacZ_-ForwardL_MYL and OlacZ_ReverseL_MYL was used to integrate gluc in the orientation opposite to the original lacZ ORF. Successful targeted integration events were selected for white colonies on X-gal plates, kanamycin resistance (Kan R ), and ampicillin sensitivity (Amp S for the expulsion of the l red recombinase expressing helper plasmid) and were validated by PCR tests using locus specific primers (List S1 in File S1). The above Kan R integrants were transformed with pCP20 for elimination of the kanamycin resistance selection marker on chromosome as described previously [35].

Mice and Tissue cage model
All animal protocols were conducted under approved institutional guidelines (IACUC protocol number 11-01-i). Mice were kept in a specific pathogen free (SPF) facility of AstraZeneca Boston R&D. Eight weeks old female CD1 mice (Charles River Lab, Boston, MA) were anesthetized with intraperitoneal (i.p.) injection of 80 mg/kg Ketamine and 10 mg/kg Xylazine. When Stage III anesthesia was achieved, as determined by lack of pedal reflex, buprenorphine (0.1 mg/kg) was subcutaneously (s.c.) administrated to mice. A sterile tissue cage was s.c. implanted on the mouse back using aseptic techniques [27,48]. The tissue cages were closed polytetrafluoroethylene Teflon cylinder with frequently spaced 0.2-mm holes on the wall. Tissue cages implanted had dimensions of 20 mm length by 8 mm external diameter (6 mm internal diameter) accomodating an internal volume of 600 ml. Two weeks after surgery, sterility of implanted cage was verified by the absence of colony on Tryptic Soy Agar (TSA) plates when plating about 15 ml of tissue cage fluid (TCF) on the plate.

Sampling of tissue cage fluid (TCF) and plasma from infected mice
Mice were anesthetized by inhalation of isofluorane and about 20-40 ml interstitial fluid (TCF) in tissue cage was drawn by percutaneous puncture (p.p.) using a sterile 25-gauge needle. For determination of CFU burden in TCF, 20 ml of TCF was 1:10 serially diluted in saline and plated on TSA plates. Colonies were counted after 16 hr incubation at 37uC. The remaining TCF sample was subjected to a 4uC centrifugation of 7,5006g for 15 min to yield a clear supernatant that was frozen at 280uC until use. About 50-100 ml blood was drawn from mice by submandibular bleeding [49] and transferred into K 2 EDTA containing

Gaussia luminescence in vivo imaging
Gaussia luciferase substrate, coelenterazine (CTZ-10, Gold BioTechnology, St. Louis, MO) was reconstituted in methanol to a concentration of 11.8 mM (5 mg/ml) and stored at 280uC until use [37]. Before imaging, CTZ stock was diluted in Duelbecco's-PBS (D-PBS) to 1.2 mg/ml work solution. Mice were anesthetized by inhalation of 4% isofluorane and each administrated with 100 ml CTZ work solution by either i.v., i.p., or intra-tissue cage (i.t.) route, at a dose of 4 mg/kg body weight. In vivo imaging was performed at the indicated time after CTZ administration using an IVIS Spectrum system (Caliper Life Sciences, Alameda, CA). Mice were positioned within the imaging chamber and 2.5% isofluorane in air was administered via nose cones through the IXG8 gas anesthesia system (Caliper Life Sciences). Images used for quantification analysis for i.t. route administration of CTZ were all acquired using 16 second integration times. For assessment of Gaussia luminescence intensity at a given location by the Living image software (version 4.3.1), pseudocolor images were generated and represented with different colors according to the color scale bar (from blue, least intense, to red, most intense; units as photons second 21 cm 22 steradian 21 ). Total luminescence quantification (photons second 21 ) at each time point was done by the region of interest (ROI) tool of the Living image software.

Measurement of Gluc activity in TCF and plasma
Clear supernatant of TCF or plasma samples (see the Experimental Procedure for sampling of TCF and plasma) was initially diluted 50-fold in D-PBS then further 1: 10 serially diluted. For each dilution, 50 ml sample was loaded into each well of 96well plates (serial number 3912, Corning, NY). The plates were assessed by adding 50 ml of CTZ substrate to each well at the final concentration of either 118 mM or 39 mM via a built-in substrate injector in luminometer (Model of InfiniteH 200 PRO, Tecan Systems, San Jose, CA). Gluc activity in TCF (CPS per ml TCF) or Gluc activity in plasma (CPS per ml plasma) was determined by the luminescence signal falling into the linear measurement range of the luminometer. It was found that 118 mM CTZ yield better results than 39 mM. Consequently, 118 mM CTZ was used for all final quantification analysis.

Statistic Analysis
Unpaired and two tailed Student's t-test was used for all two comparisons with one variable (indicated by the connected lines in the figure), P,0.01 was considered as statistically significant. Figure S1 Gluc without secretion signal expressed by E. coli is associated with bacterial cell pellet but not culture supernatant.