Quorum-sensing Salmonella selectively trigger protein expression within tumors

Abstract

Salmonella that secrete anticancer proteins have the potential to eliminate tumors, but nonspecific expression causes damage to healthy tissue. We hypothesize that Salmonella, integrated with a density-dependent switch, would only express proteins in tightly packed colonies within tumors. To test this hypothesis, we cloned the lux quorum-sensing (QS) system and a GFP reporter into nonpathogenic Salmonella. Fluorescence and bacterial density were measured in culture and in a tumor-on-a-chip device to determine the critical density necessary to initiate expression. QS Salmonella were injected into 4T1 tumor-bearing mice to quantify GFP expression in vivo using immunofluorescence. At densities below 0.6 × 10(10) cfu/g in tumors, less than 3% of QS Salmonella expressed GFP. Above densities of 4.2 × 10(10) cfu/g, QS Salmonella had similar expression levels to constitutive controls. GFP expression by QS colonies was dependent upon the distance to neighboring bacteria. No colonies expressed GFP when the average distance to neighbors was greater than 155 µm. Calculations of autoinducer concentrations showed that expression was sigmoidally dependent on density and inversely dependent on average radial distance. Based on bacterial counts from excised tissue, the liver density (0.0079 × 10(10) cfu/g) was less than the critical density (0.11 × 10(10) cfu/g) necessary to initiate expression. QS Salmonella are a promising tool for cancer treatment that will target drugs to tumors while preventing damage to healthy tissue.

Keywords: Salmonella; bacterial anticancer therapy; cancer; localized drug delivery; quorum sensing.

Fig. 1.

QS bacterial drug delivery. (A) The p(luxI) promoter controls one operon consisting of genes encoding for proteins LuxR, GFP, and LuxI. LuxI produces the communication molecule 3OC6HSL. The p(luxI) promoter responds to LuxR protein bound to 3OC6HSL. As the density of bacteria increases, 3OC6HSL concentration increases within the cell, creating a positive feedback loop that increases transcription of the operon. (B) QS bacteria will only turn on expression in high-density colonies in tumor tissue. Gray and green bacteria represent uninduced and induced bacteria, respectively. Blue dots represent 3OC6HSL.

Fig. 2.

In vitro behavior of QS Salmonella. (A) QS Salmonella only expressed GFP at densities above 1 × 10⁸ cfu/mL (*P < 0.05). Constitutive controls expressed GFP at all densities. (B) Before measurement, QS cultures were grown to either 0.5 × 10⁸ cfu/mL (low density) or 5 × 10⁸ cfu/mL (high density and induced) and then diluted to 0.001 × 10⁸ cfu/mL. High dilution density cultures had greater expression (*P < 0.05). Fluorescence was normalized to constitutive controls at 6 × 10⁸ cfu/mL. (C) 38 h after injection into tumor tissue in a microfluidic tumor-on-a-chip device, QS Salmonella expressed GFP only within distinct bacterial colonies. Constitutive Salmonella expressed GFP throughout the tissue regardless of bacterial concentration. Bacteria took 10 h to colonize tissue and were not present at 0 h. (Scale bar, 100 µm.) (D) Area fraction of tissue with GFP expression was less for tissue treated with QS Salmonella compared with constitutive controls after 40 h (*P < 0.05).

Fig. 3.

Salmonella distribution within tumor-bearing mice. (A) Bacterial density was higher in tumors (solid bars) than livers (open bars) for mice administered both QS and constitutively expressing Salmonella (*P < 0.05). (Inset) Liver densities were small compared with tumor densities, but bacteria were present in most livers. (B) Salmonella (red) and GFP (green) distribution in 4T1 tumors injected with QS Salmonella or constitutive controls. Sections were counterstained with DAPI (blue). (C) No Salmonella or GFP were observed in any liver section using immunofluoresence. (Scale bar, 5 mm.) (D) Most Salmonella colonies were located in low-density regions. Local density is the number of bacteria within a 197-µm circle surrounding each colony. Colonies were groups of contiguous bacteria distinctly separate from neighbors (Inset). (Scale bar, 25 µm.)

Fig. 4.

Density dependence of GFP expression by QS Salmonella. (A) Colonies of QS and constitutive Salmonella (red) in 4T1 tumors and associated GFP (green) for areas of low and high density. Yellow indicates areas of colocalization between Salmonella and GFP. (Scale bar, 100 µm.) (B) GFP expression dependence on Salmonella density. The relationship between expression and density was linear for constitutive controls (n = 133,305 colonies) and sigmoidal (red line) for QS Salmonella (n = 84,213 colonies). At all densities less than 4.2 × 10¹⁰ cfu/g, the fraction of GFP-expressing QS colonies was less than the constitutive control colonies at the same density (*P < 0.05). Below 4.2 × 10¹⁰ cfu/g, the expressing fraction of QS colonies was less than control colonies at the lowest density, 0.3 × 10¹⁰ cfu/g (* and **P < 0.05). (C) The ratio of QS to control expressing fractions was significantly greater at densities above 3.0 × 10¹⁰ cfu/g compared with the densities below this threshold (*P < 0.05). Above 3.0 × 10¹⁰ cfu/g the QS:control ratio was close to 1.

Fig. 5.

Dependence on spatial distribution. (A) A QS Salmonella colony (Center) was more likely to express GFP if the average radial distance to its neighbors was shorter. Red, green, and blue bacteria represent uninduced, induced, and neighboring colonies, respectively. Density and radial distance was measured within circles of radius 197 µm (150 pixels) around colonies. (B) GFP expression was different for two colonies at the same density (1.2 × 10¹⁰ cfu/g) with different spatial distributions. A colony with an average radial distance to its neighbors of 74 µm expressed GFP (B, i), but a colony with an average radial distance of 145 µm did not (B, ii). (Scale bars, 100 µm.) (C) The fraction of colonies expressing GFP was greater for colonies with close compared with far neighbors, and with densities greater than 0.11 × 10¹⁰ cfu/g (n = 50,145 colonies). The expressing fraction was greater than the average for colonies with average radial distances less than 58 µm (*P < 0.05) and less than the average for distances greater than 87 µm (*P < 0.05). At an average distance of 155 µm, the fraction expressing was zero.

Fig. 6.

Calculated 3OC6HSL concentration predicts GFP expression. (A) The fraction of colonies (n = 84,213) that expressed GFP was greater (more red) at high densities (Top) and short radial distances (Left). (B) Compared with an uninduced colony (B, i, red), a target colony was more likely to express GFP (green) if (B, ii) it was surrounded by more source colonies (blue) or (B, iii) the distance to source colonies was shorter. (C) The fraction of expressing colonies and the predicted probability of induction (α), were lower for colonies with lower calculated concentrations of 3OC6HSL (P < 1 × 10⁻¹⁵). The critical 3OC6HSL concentration (C_crit) was ${\overline{C}}_t$ = 0.38. (D) The predicted fraction of expressing colonies, α, was higher at high density and low average radial distance, matching colony measurements in tumor sections (A). (E and F) Predicted α-fractions were lower at longer average radial distances (E) and higher at higher densities (F). Numbers to the right are density (×10¹⁰ cfu/g) for E and G, and radius (µm) for F, H, and I. (G and H) The predicted 3OC6HSL concentration at target colonies (${\overline{C}}_t$) was lower at higher average radial distances (G), and higher at higher densities (H). (I) Expanded range of H. At low densities below ρ_crit = 0.11 × 10¹⁰ cfu/g, the predicted 3OC6HSL concentration was zero. At the maximum possible bacterial density in liver (ρ_liv = 0.00792 × 10¹⁰ cfu/g), no 3OC6HSL was produced, regardless of spatial distribution.