Rhodobacter sphaeroides, a novel tumor-targeting bacteria that emits natural near-infrared fluorescence



Several optical imaging techniques have been used to monitor bacterial tropisms for cancer. Most such techniques require genetic engineering of the bacteria to express optical reporter genes. This study investigated a novel tumor-targeting strain of bacteria, Rhodobacter sphaeroides 2.4.1 (R. sphaeroides), which naturally emits near-infrared fluorescence, thereby facilitating the visualization of bacterial tropisms for cancer. To determine the penetration depth of bacterial fluorescence, various numbers of cells (from 108 to 1010 CFU) of R. sphaeroides and two types of Escherichia coli, which stably express green fluorescent protein (GFP) or red fluorescent protein (RFP), were injected s.c. or i.m. into mice. Bacterial tropism for cancer was determined after i.v. injection of R. sphaeroides (108 CFU) into mice implanted s.c. with eight types of tumors. The intensity of the fluorescence signal in deep tissue (muscle) from R. sphaeroides was much stronger than from E. coli-expressing GFP or RFP. The near-infrared fluorescence signal from R. sphaeroides was visualized clearly in all types of human or murine tumors via accumulation of bacteria. Analyses of C-reactive protein and procalcitonin concentrations and body weights indicated that i.v. injection of R. sphaeroides does not induce serious systemic immune reactions. This study suggests that R. sphaeroides could be used as a tumor-targeting microorganism for the selective delivery of drugs to tumor tissues without eliciting a systemic immune reaction and for visualizing tumors.

List of Abbreviations

C-reactive protein


days post-inoculation


enhanced green fluorescent protein


green fluorescent protein


indocyanine green


Luria-Bertani media


bacterial luciferase gene






red fluorescent protein



R. sphaeroides

Rhodobacter sphaeroides 2.4.1


signal-to-noise ratio

In recent decades, several microorganisms have been shown to accumulate and replicate selectively in tumors [1]. This feature has been exploited to develop novel approaches to the treatment of cancer using bacteria. Several bacterial strains have been assessed in mouse tumor models, including Escherichia coli [2, 3], Bifidobacterium longum [4], Clostridium novyi [5] and Salmonella typhimurium [6, 7].

Optical imaging markers are preferred for in vivo bacterial monitoring because of they are convenient and simple to manipulate. Bacteria have been labeled with fluorescence or bioluminescence to track their movements in animal models, allowing the fate of light-emitting bacteria to be monitored easily in a non-invasive and repeatable manner [2, 6-10]. To develop bioluminescent bacteria, the bacteria are usually transformed with a plasmid that encodes the bacterial luciferase (lux) operon (luxCDABE) [2, 8, 10]. The lux operon encodes all of the proteins required for bioluminescence, including bacterial luciferase, substrate and substrate-regenerating enzymes. Thus, bacteria that express the lux operon do not require an exogenous substrate to produce bioluminescence.

Another approach to imaging bacteria in intact animals is to engineer bacteria that express fluorescent proteins [9]. This technique can utilize orally administered bacteria to visualize the spatial and temporal behavior of bacterial infections in great detail [9]. However, fluorescence imaging may have limitations for tumor imaging because extremely high sensitivity measurements are required to image the small numbers of bacteria that enter tumors and replicate within them.

Rhodobacter sphaeroides 2.4.1, a rod-shaped, gram-negative, purple bacterium, uses a special type of photosynthetic chlorophyll to harvest light energy and transduce it to generate electrochemical energy [11, 12]. Some of the quantum energy generates structural and other changes in the absorbing molecules, before re-emitting light with less quantum energy. R. sphaeroides absorbs NIR light at two peaks near 800 and 853 nm and emits NIR light with a single peak located near 900 nm [11].

In this study, the natural NIR-emitting fluorescent bacteria R. sphaeroides was assessed to determine whether it could be used for the selective targeting of tumor tissues and visualization of various types of tumor. The results suggest that NIR-emitting R. sphaeroides has advantages over other types of fluorescent bacteria for visualizing cancers.


Bacterial strains and plasmids

The bacterial strains used in this study are summarized in Table 1. R. sphaeroides was grown in the dark at 30 °C for 4–5 days in yeast extract-casein-cysteine media, as described previously [13]. The OD of the cultured cells was measured at an absorbance of 600 nm (OD600). The cells were then washed and diluted with PBS to produce appropriate concentrations of bacteria for use in the experiments. The number of R. sphaeroides cells was calculated by assuming that 1 OD600 represents 109 CFU.

Table 1. Bacterial strains used in this study
StrainDescriptionReference source
Rhodobacter sphaeroides 2.4.1Wild type[9]
HJ1020gfpE. coli MG1655 containing pEGFPThis study
HJ1020rfpE. coli MG1655 containing pRFPThis study

E. coli asd::kan strain (HJ1020) was constructed from E. coli MG1655 by the linear DNA transformation method, as described previously [14]. To construct a plasmid containing both EGFP and asd, the asd gene was PCR-amplified from E. coli (MG1655) genomic DNA and the 1.1 kb fragment cloned into pGEM-T Easy (Promega, Madison, WI, USA), digested with EcoRI, and ligated into the same site in pEGFP (Clontech, Mountain View, CA, USA), thereby generating the construct Asd+ pEGFP, as reported previously [2]. To construct a plasmid containing RFP gene, the PCR-amplified DsRed gene was cloned into the BamHI/XbaI site of pTrc99A, thereby generating pTrc99A-BnRed. E. coli MG1655 was transformed with pTrc99A-BnRed. The asd gene was also ligated in a similar manner.

Transformed E. coli (HJ1020gfp or HJ1020rfp) was grown in LB medium (Difco, Franklin Lakes, NJ, USA) containing appropriate antibiotics and diaminopimelic acid (Sigma, St Louis, MO, USA) at 37 °C for 12–16 hr. The following day, 2% of the bacterial suspension was used to produce a fresh culture in new LB/antibiotic liquid media containing 0.2% glucose. The bacteria were harvested during the early stationary phase (OD600: 2–2.5), centrifuged (2300 g, 10 min), washed with PBS, titrated using a spectrophotometer, and diluted in PBS to produce appropriate concentrations of bacteria for use in the experiments. The bacterial numbers were calculated by assuming that 1 OD600 represents 0.8 × 109 CFU.

Cell lines

HeLa human cervix adenocarcinoma cells (CCL-2), C6 rat brain glioma cells (CCL-107), CT26 mouse colon carcinoma cells (CRL-2638), 4T1 mouse breast cancer cells (CRL-2539), AsPC-1 human pancreas cancer cells (CRL-1682), MCF7 human breast adenocarcinoma cells (HTB-22) and MDA-MB-435 human breast ductal carcinoma cells (HTB-129) were obtained from the ATCC (Rockville, MD, USA). SNU-C5 human colon cancer cells (KCLB-0000C5) were obtained from the Korean Cell Line Bank (Seoul, Korea). The HeLa, CT26, and 4T1 cells were grown in high-glucose Dulbecco's-modified Eagle medium supplemented with 10% FBS and 1% antibiotic/antimycotic mixture (Gibco, Gaithersburg, MD, USA; Invitrogen, Carlsbad, CA, USA). The AsPC-1, C6, SNU-C5, MCF7 and MDA-MB-435 cells were grown in RPMI1640 supplemented with 10% FBS and 1% penicillin–streptomycin.

Animal models

Five- to six-week-old male BALB/c or BALB/c athymic nu−/nu− mice (20–30 g body weight) were purchased from Orient Bio (Seongnam, Korea). Animal care, experiments, and killing of mice were performed in accordance with protocols approved by the Chonnam National University Animal Research Committee (Gwangju, Korea). The animals were anesthetized using isoflurane (2%) during imaging or with a mixture of ketamine (200 mg/kg) and xylazine (10 mg/kg) during surgery. Subcutaneous tumors were generated in mice as follows. The tumor cells were harvested and then suspended in 50 µL PBS. According to the tumor type, the tumor cells were injected into the right thighs of BALB/c or BALB/c nu−/nu− mice.

Injection of bacteria into animals

To compare the intensity of the imaging signal as a function of tissue depth, varying numbers (108 to 1010 CFU with 10 different dosages) of R. sphaeroides and E. coli cells were injected separately, either s.c. or i.m. into each thigh of anesthetized BALB/c nu−/nu− mice. Fluorescence imaging was performed within several minutes of injection.

To investigate tumor targeting by NIR-emitting bacteria, a 1 mL insulin syringe was used to inject R. sphaeroides (108 CFU) cells into the lateral tail veins of mice with various types of tumor.

Fluorescence imaging using a cooled, charge-coupled, detector camera system

To image bacterial fluorescence, anesthetized animals were placed in the light-tight chamber of an IVIS100 system (Caliper, Hopkinton, MA, USA) equipped with a cooled charge-coupled detector camera and scanned using several different filters (ICG, GFP, and DsRed) to image various fluorescent spectra. Pseudocolor images representing the photon counts were overlaid onto photographs of the mice using Living Image software v.2.25 (Caliper).

To compare the signal intensity with different bacterial species, and to evaluate the proliferation rate of R. sphaeroides in tumor tissue, an ROI was selected manually in the collected signal. The area of the ROI was kept constant and the intensity recorded as the maximum radiance within an ROI.

Toxicity assay

Systemic or local inflammation and infection after administration of R. sphaeroides were determined by using a mouse CRP enzyme-linked immunosorbent assay kit (Life Diagnostics, West Chester, PA, USA) to measure mouse plasma concentrations of CRP and a BRAHMS PCT-Q kit (BRAHMS Diagnostica, Hennigsdorf, Germany) to measure plasma concentrations of PCT, according to the manufacturers’ protocols. The body weights of mice were also measured to assess differences in general toxicity between the control and R. sphaeroides-injected groups.

Statistical analyses

The Mann–Whitney U-test was used to assess differences between the R. sphaeroides- and E. coli-injected groups in signal intensity and toxicity. P < 0.05 was considered statistically significant for all analyses. All data are expressed as the mean ± SD. The statistical analyses were performed using SPSS 18.0 (SPSS, Chicago, IL, USA).


Comparison of fluorescent signal intensities of various bacterial species

To evaluate the penetration depth of each fluorescent bacterium, various amounts (108–1010 CFU) of three fluorescent bacterial strains (R. sphaeroides, HJ1020gfp and HJ1020rfp) were injected s.c. or i.m. into BALB/c nu−/nu− mice and the fluorescent signals in each mouse analyzed. After the bacteria had been injected s.c., R. sphaeroides and HJ1020rfp could be detected by fluorescence imaging for all CFU doses, whereas HJ1020gfp could not be detected with 108 CFU (Fig. 1a). When the bacteria had been injected i.m., fluorescence signals from R. sphaeroides and HJ1020rfp could be detected for 109 and 1010 CFU, whereas the fluorescence signal from HJ1020gfp was detected only weakly for 1010 CFU (Fig. 1b). Quantitative assessment of the signal intensity showed that, after s.c. injection, the signal from R. sphaeroides was approximately 2.3-fold higher for 109 CFU and 2.5-fold higher for 1010 CFU compared with that from HJ1020gfp, and approximately 2.3-fold higher for 109 CFU and 2.6-fold higher for 1010 CFU compared with that from HJ1020rfp (Fig. 1c). On the other hand, after i.m. injection the signal intensity of R. sphaeroides was approximately 4.3-fold higher for 109 CFU and 6.6-fold higher for 1010 CFU compared with that from HJ1020gfp, and approximately 1.9-fold higher for 109 CFU and 2.9-fold higher for 1010 CFU compared with that from HJ1020rfp (Fig. 1d). Therefore, the fluorescent signal from R. sphaeroides would be detectable in deeper tissues than that from E. coli expressing GFP or RFP.

Figure 1.

Detection of light-emitting bacteria according to the depth of the bacterial injection site.

(a) Comparison of signal intensity from various bacterial species after s.c. injection. Various cell numbers of R. sphaeroides, GFP-expressing E. coli (HJ1020gfp), and RFP-expressing E. coli (HJ1020rfp) were injected s.c. into anesthetized BALB/c nu−/nu− mice (n = 3 per group). Fluorescence imaging was performed immediately after injection, imaging signals collected over various time periods and different filters used to the fluorescent spectra (R. sphaeroides: 10 s with ICG; HJ1020gfp: 1 s with GFP; HJ1020rfp: 0.5 s with DsRed). (b) Comparison of signal intensity from various bacterial species after i.m. injection. Various cell numbers of R. sphaeroides, HJ1020gfp, and HJ1020rfp were injected i.m. into anesthetized BALB/c nu−/nu− mice (n = 3 per group). Fluorescence imaging was performed immediately after injection, imaging signals collected over various time periods and different filters used to image the fluorescent spectra (R. sphaeroides: 10 s with ICG; HJ1020gfp: 1 s with GFP; HJ1020rfp: 1 s with DsRed). (c, d) Signal intensity was assessed quantitatively by measuring the maximum radiance (photons s/cm2/sr) from ROI in the injected site and in contralateral soft tissue (background), and by calculating the SNR. The y-axis indicates SNR (maximum radiance in the injected site/maximum radiance in contralateral soft tissue). Each bar represents mean of SNR ± SD. (c) Comparison of SNRs after s.c. injection. *, P = 0.05 (R. sphaeroides vs. HJ1020gfp or HJ1020rfp with the same injected dose of bacteria). (d) Comparison of SNRs after i.m. injection. *, P = 0.05 (R. sphaeroides vs. HJ1020gfp with 109 CFU); **, P = 0.05 (R. sphaeroides vs. HJ1020gfp or HJ1020rfp with 1010 CFU).

Tumor targeting by R. sphaeroides 2.4.1

The success of tumor targeting by R. sphaeroides was assessed using in vivo fluorescence imaging. R. sphaeroides was injected i.v. 28 days after HeLa cells had been implanted in the right lateral thighs of mice. During the first few hours, NIR signals were detected mainly in the mice spleens and livers. However, within 24 hr the signals from R. sphaeroides in the liver and spleen had diminished, after which they were detected exclusively in the tumor region (Fig. 2a). The proliferation rate was assessed quantitatively by measuring photon flux from the tumors (Fig. 2b). When the tumor-targeting ability of R. sphaeroides was investigated using various additional types of tumors, it was found that R. sphaeroides could target and proliferate in 4T1, CT26, C6, SNU-C5, MCF7, MDA-MB-435 and AsPC-1 tumor cells implanted in the right lateral thighs of mice (Fig. 2c).

Figure 2.

Visualization of NIR fluorescence-emitting Rhodobacter sphaeroides 2.4.1 (R. sphaeroides) in tumor tissues.

(a) Approximately 107 HeLa cells were implanted in the right lateral thighs of BALB/c nu−/nu− mice and R. sphaeroides injected i.v. 28 days after implantation. Tumors are indicated by arrows. Fluorescence imaging was performed five times at intervals of 7 days after bacterial injection. The imaging signals were collected for 30 s with an ICG filter. The y-axis indicates photons × 106 s/cm2/sr. (b) Signal intensities were assessed quantitatively in the tumor regions by measuring the maximum radiance (photons s/cm2/sr) from the ROI, which was selected manually within each tumor site, and plotted as a function of time after injection of R. sphaeroides. (c) Various types of tumor cells were implanted in the right lateral thighs of mice, in the same manner as HeLa cells. The CT26 and 4T1 cells were injected into BALB/c mice whereas the other cells were injected into BALB/c nu−/nu− mice. R. sphaeroides was injected i.v. when the tumor measured approximately 150 mm3 and fluorescence imaging was performed 15–20 days later. Imaging signals were collected for 30 s with an ICG filter.

Toxicity of R. sphaeroides 2.4.1 infection

Proposals to administer live R. sphaeroides raise concerns about potential toxicity. Thus, the acute and short-term toxicity of i.v. injections of R. sphaeroides in mice were assessed by monitoring plasma CRP and PCT concentrations and body weight (Fig. 3). Plasma CRP concentrations were significantly lower in mice treated with R. sphaeroides than in positive controls (P < 0.01), whereas there was no significant difference in CRP concentrations between mice that received R. sphaeroides and those that received PBS (Fig. 3a). Plasma PCT concentrations were <0.5 ng/mL after administration of R. sphaeroides. These findings suggest that there was no systemic infection (or sepsis) (Fig. 3b). In addition, there were no significant differences in the body weights of infected and non-infected mice (Fig. 3c). Thus, R. sphaeroides likely induces no significant short-term host toxicity.

Figure 3.

Toxicity assay using R. sphaeroides in mice.

(a) Plasma concentrations of CRP were measured 1, 5 and 10 days after i.v. injection of PBS or R. sphaeroides into tumor-bearing mice (n = 3 per group). The data are expressed as average CRP concentrations on each dpi. Mice (n = 3) were injected i.v. with LPS (5 mg/kg over 3 min) as positive controls and blood drawn 4 hr after injection. The boxes represent the quartiles and the whiskers indicate the 10th and 90th percentiles. *, P < 0.01 (R. sphaeroides vs. positive control). (b) Immunochromatographic analysis of mouse plasma PCT concentrations after R. sphaeroides injection (1 × 108 CFU). Mice (n = 3) were injected i.v. with LPS (5 mg/kg over 3 min) as positive controls. (c) Changes in body weight after i.v. injection of R. sphaeroides. The body weights of mice were measured before and after injection of R. sphaeroides (1 × 108 CFU) or PBS.


Although many advances in the optical imaging of living subjects have been made in recent years, imaging technology still has many limitations. These include the undesirable absorption of visible light by hemoglobin and the fact that high background signals produced by autofluorescence lead to limited tissue penetration [15]. Because hemoglobin absorbs less light in this wavelength band than in the visible light range, the NIR spectral region (i.e., a window of 650–950 nm) has attractive properties for optical imaging. This wavelength band facilitates photon detection even after propagation through several centimeters of tissue, for example, through >10 cm of human breast tissue (16) and >3–4 cm of muscle tissue; it also allows assessment of brain function [17]. By contrast, only a few millimeters of tissue can be visualized using similar SNR parameters in the blue, green, or red light regions; however, the exact penetration depth is highly dependent on the wavelength selected. Thus, NIR imaging maximizes tissue penetration and minimizes the noise from nontarget biological samples [18].

R. sphaeroides, a photosynthetic and nonpathogenic bacterium, emits NIR fluorescence [11, 12]. The absorption spectrum of R. sphaeroides has two peaks near 800 and 853 nm, and the fluorescence spectrum, after excitation at 803 nm, has a peak near 900 nm [11]. Therefore, provided R. sphaeroides can target tumors successfully, it could be a powerful tool for tumor imaging. In the present study, in vivo fluorescence imaging was used to demonstrate the tropism of R. sphaeroides for various types of mouse tumors. Stronger signals from deeper tissue (muscle) were detected from R. sphaeroides than from E. coli-expressing GFP or RFP (Fig. 1b).

The tumor-targeting mechanisms of bacteria are not well understood; however, they may differ depending on oxygen tolerance [19]. Obligate anaerobes such as Clostridium and Bifidobacterium cannot survive in oxygen; thus, injected spores of these bacteria can germinate only in the anoxic regions of tumors [20]. Because completely deoxygenated tissue is only found in tumors and not in other parts of the body, obligate anaerobes reliably accumulate in the hypoxic regions of tumors.

Facultative anaerobes such as Salmonella and E. coli use a complex mechanism to target tumors. When injected systemically, Salmonella attach to the walls of tumor vasculature in small but measurable numbers [21]. Accumulation of Salmonella in a tumor is associated with an influx of blood into the tumor; this is caused by increased blood concentrations of TNF-α as part of the immunological response to this organism [22]. In addition, chemotaxis may be one of the mechanisms that control accumulation of Salmonella in tumors. Using in vitro tumor models, Salmonella has been shown to identify and penetrate tumors by detecting and moving chemotactically toward small molecular gradients of serine, aspartate and ribose [23]. R. sphaeroides has three operons that potentially encode complete chemosensory pathways [24], as well as four CheA proteins (three of which are essential for chemotaxis), CheY and CheB proteins [25]. Tumors provide relatively immuno-privileged environments in which bacteria can replicate unimpeded by the macrophages and neutrophils that normally try to eliminate them [26]. Thus, both the immune system and chemotaxis may be involved in bacterial tumor-targeting mechanisms.

The fluorescence signal intensity depended on the number of R. sphaeroides. As shown in Figure 1, the signal intensity of bacterial fluorescence correlated well with the number of bacteria injected either s.c. or i.m. Therefore, we speculate that the increased signal intensity of bacterial fluorescence in HeLa tumors for several days (Fig. 2a) indicates active proliferation of R. sphaeroides in tumor tissue after colonization. In our previous study [10], direct comparison between the photon flux and bacterial counts revealed an excellent correlation when the bioluminescence gene was stably maintained by a balanced-lethal host-vector system. Because live R. sphaeroides naturally emit fluorescence, the signal intensity is thought to indicate their proliferation rate.

In this study, R. sphaeroides started to accumulate in HeLa tumor tissue on Day 7 after injection (Fig. 2a). This suggests that R. sphaeroides is less capable of providing rapid visualization of tumors than other optical probes such as copolymer delivery-quenched NIR-fluorescence probes [27]. The signal intensity may depend on the growth rate of R. sphaeroides in different tumor types, that is, the delayed visualization of tumors with R. sphaeroides may be attributable to the bacteria's slow growth rate. Further studies are required to ascertain the growth rate of R. sphaeroides in larger animal models.

No evidence of serious local or systemic inflammatory reactions was noted after i.v. administration of R. sphaeroides, and the plasma concentrations of CRP and PCT did not change significantly (Fig. 3). In the acute-phase response of most, although not all diseases, circulating CRP concentrations reflect ongoing inflammation and/or tissue damage much more accurately than do other laboratory variables. The LPS obtained from R. sphaeroides is reportedly nontoxic [28]. In particular, nontoxic diphosphoryl lipid A of R. sphaeroides has been reported to inhibit LPS-induced macrophage TNF secretion [29] and to down-regulate LPS responsiveness through the induction and actions of corticosterone [30]. Therefore, bacterial colonization of tumors may not induce serious systemic inflammation and infection. Further study is required to clarify the influence of R. sphaeroides on the host immune system.

In the field of bacteria-mediated cancer therapy, R. sphaeroides, which is nonpathogenic and may evade immune reaction, may have limited applications as a therapeutic agent. We observed no evidence of tumor-suppression after i.v. injection of R. sphaeroides in various tumor-model mice in which Salmonella treatment reportedly has antitumor effects [6]. However, R. sphaeroides has a potential role in photodynamic therapy because, during the photosynthetic process, bacterial chlorophyll generates reactive oxygen species, which can destroy tumor tissue [31, 32]. In addition, R. sphaeroides could be used to produce and secrete antitumor agents in a similar manner to Salmonella or E. coli.

In conclusion, this study demonstrated that R. sphaeroides has a tumor-targeting capacity in various types of tumor-bearing mice. The natural NIR fluorescence produced by R. sphaeroides had a significantly greater tissue penetration depth than the range of visible light produced by E. coli expressing GFP or RFP. Systemic injection of live bacteria did not elicit serious systemic inflammation or infection, which may facilitate translation of this strategy into clinical applications, such as intraoperative observation of extent of tumor. Further sophisticated investigations will be required to analyze its targeting capacity in small tumors and metastatic lymph nodes, thereby elucidating the tumor-targeting mechanisms and reducing targeting time after injection of R. sphaeroides during clinical applications.


This work was supported by the National Research Foundation of Korea (No. 2012-0006072).


All authors have no conflicts of interest.