Intracellular delivery of protein drugs with an autonomously lysing bacterial system reduces tumor growth and metastases

Abstract

Critical cancer pathways often cannot be targeted because of limited efficiency crossing cell membranes. Here we report the development of a Salmonella-based intracellular delivery system to address this challenge. We engineer genetic circuits that (1) activate the regulator flhDC to drive invasion and (2) induce lysis to release proteins into tumor cells. Released protein drugs diffuse from Salmonella containing vacuoles into the cellular cytoplasm where they interact with their therapeutic targets. Control of invasion with flhDC increases delivery over 500 times. The autonomous triggering of lysis after invasion makes the platform self-limiting and prevents drug release in healthy organs. Bacterial delivery of constitutively active caspase-3 blocks the growth of hepatocellular carcinoma and lung metastases, and increases survival in mice. This success in targeted killing of cancer cells provides critical evidence that this approach will be applicable to a wide range of protein drugs for the treatment of solid tumors.

Fig. 1. The intracellular lifestyle of Salmonella is controlled by flhDC.

a The design goals were to genetically engineer a bacterial vehicle that (1) synthesizes a protein drug (yellow/purple), (2) actively invades into cancer cells, and (3) releases drug, which escapes Salmonella vacuoles (SCVs, red). b, c Ninety-six hours after intratumoral injection of 2 × 10⁶ CFU of intracellular-reporting Salmonella into subcutaneous 4T1 tumors in BALB/c mice, more bacteria (red) were intracellular (green; black arrows) than extracellular (white arrows; P < 0.0001; n = 1258 bacteria in 4 mice). d) In monolayer culture, Salmonella (light blue, arrows) invade cancer cells (red). Extracellular bacteria were removed with gentamicin (an invasion assay). e Knockout ΔflhD Salmonella were transformed with PBAD-flhDC. Uninduced bacteria (flhDC-) minimally invaded cancer cells. Induction with 20 mM arabinose (flhDC+) promoted invasion (black arrows). f Re-expression of flhDC significantly increased invasion (P = 0.0012; n = 3 independent biological samples). g In three-dimensional tumor-on-a-chip devices (top left), flhDC+ Salmonella, with a green invasion reporter (top right), invaded more cells than flhDC− controls (P < 0.0001; n = 6 chambers for flhDC+ and n = 4 for flhDC-). Data are shown as means ± SEM. Statistical comparisons in c, f, and g, and are two-tailed, unpaired Student’s t-tests with asterisks indicating significant differences (**P < 0.01; ****P < 0.0001). Images in b, d, and e are representative of 4, 52, and 6 independent biological samples. Scale bars in b, d, and e are 10 µm, and in g the scale bars are 100 µm.

Fig. 2. Design of ID Salmonella to release protein into cells.

a Salmonella with the PsifA-GFP or PsseJ-GFP reporter constructs both expressed GFP after invasion (white arrows). Extracellular expression (black arrows) from PsseJ-GFP was less than PsifA-GFP (P = 0.005; n = 27 for PsifA and n = 9 for PsseJ). The intracellular activity of the PsseJ promoter was four times greater than extracellular activity (P < 0.0001; n = 20). b Induction of PBAD-LysE at 96 h (arrow) induced bacteria lysis (n = 3). c When administered to MCF7 cancer cells, Salmonella (red, white arrow) with PsseJ-LysE and Plac-GFP delivered GFP (green, black arrow) into the cellular cytoplasm. Only released, and not intrabacterial GFP was stained. d Most intracellular Salmonella with PsseJ-LysE lysed, which was significantly greater than PsseJ-GFP controls (P < 0.0001; n = 4). e Lysis of intracellular PsseJ-LysE Salmonella occurred for 10 h after invasion (n = 3). The fraction of MCF7 cancer cells in which Salmonella lysed (disappeared) per hour (bars) was relatively constant. The cumulative fraction of cells with intact bacteria (open circles) decreased exponentially. f In liquid culture, PBAD-lysE and PsseJ-LysE Salmonella grew at similar rates as nontransformed controls (white bars). When intracellular, PsseJ-LysE Salmonella, lysed at a similar rate as induced PBAD-lysE Salmonella in culture (black bars; n = 3 for all conditions except n = 6 for uninduced PBAD-lysE). g The EGFP content of PsseJ-LysE Salmonella was determined by immunoblot. Lysed bacteria at 10⁶, 10⁷, 10⁸, and 10⁹ colony forming units (CFU) per well (four right lanes) were compared to EGFP standards at three concentrations: 1, 10, and 100 ng/well (three left lanes). h Ninety-six hours after intratumoral injection of 2 × 10⁶ bacteria/mouse to BALB/c mice with subcutaneous 4T1 tumors (n = 3), ID Salmonella (which utilizes PsseJ-LysE) delivered GFP into cancer cells (arrows). i In the same mice, delivered GFP was present in extracts from tumors (T), but not livers (L) or spleens (S). Actin was used as a loading control. j Anti-actin nanobody (NB) and GFP (Ctr) was delivered into 4T1 cancer cells with ID Salmonella. Beta-actin was immuno-precipitated with delivered nanobody and was enriched 2.5 times compared to controls. Data are shown as means ± SEM. Statistical comparisons in a and d are two-tailed, unpaired Student’s t-tests with asterisks indicating significance (**P < 0.01; ****P < 0.0001). Images in a, left and c are representative of 20 and 10 independent biological samples. The immunoblot in j is from a single experiment and the immunoblots in g and i are each representative of two independent experiments with similar results. Scale bars in a, c, and h are 10 µm.

Fig. 3. Release of delivered proteins from SCVs.

a Selective permeabilization of cancer cell membranes enabled detection of released GFP. After administration to cells at an MOI of 10, ID Salmonella (top) released GFP (green, black arrows). Intact bacteria (red, white arrows) are easily discernable from the membranes of lysed bacteria (faint red, black arrows). In cells administered nonlysing Salmonella that produced GFP (bottom), only intracellular bacteria (red, white arrows) and no GFP was detected. b The amount of GFP detected in cultures administered ID Salmonella was fifty times greater than nonlysing controls (P < 0.0001; n = 9). c After invasion and before lysis, Salmonella (light blue, white arrow) are in LAMP1-stained SCVs (red, yellow arrows). d After lysis, GFP (green, black arrow) is retained within the membranes of SCVs (red, yellow arrow). e In phalloidin-stained cancer cells (red), released GFP (green, black arrows) moved from SCVs near the nucleus (blue) to throughout the cytoplasm. f From 6 to 24 h after invasion, the percentage of released GFP in the cytosol increased from 25 to 75% (P < 0.0001; n = 21 at 6 h and n = 6 at 24 h). g The lysis of individual ID Salmonella released GFP that diffused through the cytosol of a cancer cell (see Supplementary Movie 1). h Temporal profiles of GFP intensity, centered on the lysed bacteria. i Salmonella in LAMP1-stained SCVs (red, yellow arrow) lysed and released GFP (green, black arrow). Cytosolic bacteria (light blue, white arrow) did not lyse. j Most lysed bacteria, identified by colocalized staining for Salmonella and released GFP, were located within LAMP1-stained SCVs (P < 0.0001; n = 7). k Predominantly cytoplasmic ΔsifA ID Salmonella remained intact (red, white arrows) and lysed less (green, black arrows) than predominantly vacuolar ΔsseJ ID Salmonella and ID Salmonella (P < 0.0001; n = 9). Data are shown as means ± SEM. Statistical comparisons in b, f, and j are two-tailed, unpaired Student’s t-tests. The statistical comparisons in k were to a single control performed with ANOVA followed by Dunnett’s method. Two outliers were removed from ΔsifA using the ROUT method with a Q of 1%. Asterisks indicate significance (****P < 0.0001). Images in a, c, d, i, and k are representative of 9, 25, 7, 18, and 9 independent biological samples. Images in e are representative of 21 and 6 independent biological samples at 6 and 24 h, respectively. The images in g are frames from Supplementary Movie 1. Scale bars in a, c–e, i, and k are 10 µm, and in g the scale bars are 1 µm.

Fig. 4. PsseJ-LysE and flhDC are necessary for delivery to tumors.

a To control cell invasion and protein release, IDf+ Salmonella were created by transforming ΔflhD Salmonella with PsseJ-LysE, PBAD-flhDC, and Plac-GFP. b IDf+ Salmonella were administered to 4T1 cancer cells and PBAD-flhDC was induced with 20 mM arabinose. Controls either lacked PsseJ-LysE, were uninduced, or both. Invasion was detected with anti-Salmonella antibodies and delivery was detected by the presence of released GFP. c GFP (green, arrows) was only delivered from Salmonella with activated PBAD-flhDC and transformed with PsseJ-LysE. d Induced IDf+ Salmonella delivered significantly more GFP than all controls (P < 0.0001; n = 6). e IDf+ Salmonella (2 × 10⁸ CFU/mouse) were intravenously injected into BALB/c mice with subcutaneous 4T1 tumors via the tail vein (n = 3). In flhDC+ mice, 100 µg of arabinose was injected IP at 48 and 72 h. At 96 h, delivered GFP (arrows) was measured in excised tumors with immunohistochemistry. f In the transition zone of the tumors in e, induction of flhDC increased the fraction of cells with delivered GFP (P = 0.0004; n = 15). Data are shown as means ± SEM. The statistical comparisons in d were to a single condition performed with ANOVA followed by Dunnett’s method. The statistical comparison in f is a two-tailed, unpaired Student’s t-test. Asterisks indicate significance (***P < 0.001; ****P < 0.0001). The results in b are from a single experiment and the images in c are representative of 6 independent biological samples. Scale bars in c and e are 10 µm.

Fig. 5. Safety, biodistribution, and clearance of ID Salmonella.

a Firefly luciferase-expressing ID Salmonella (2 × 10⁷ CFU/mouse) were intravenously injected into BALB/c mice with 4T1 tumors in the mammary fat pad (n = 4). Prior to imaging, mice were injected IP with 100 µl of 30 mg/ml D-luciferin. b The bacterial density in the tumors increased for 72 h and then decreased. c Biodistribution of bacteria in tumor-free BALB/c mice, 14 days after intravenous injection with 1 × 10⁷ ID Salmonella (n = 5). Densities were below detection in the lungs, kidneys, hearts, and brains. Measurements of zero bacteria in spleen (1 of 5) and liver (2 of 5) were not displayed. In spleens and livers, bacterial densities were more than 3000 times lower than in tumors (from the separate experiment in panel b using the same organ mincing technique (P = 0.0001, n = 3 mice, one mouse died prior to density measurement). d Comprehensive hematology of blood drawn from tumor-free BALB/c mice, 14 days after intravenous injection with 1 × 10⁷ ID Salmonella or saline (n = 4). No changes were observed in the number of any immune cells in the blood. e Chemistry profiling of the blood from the mice in d. There was no indication of liver damage, despite some liver colonization (c). Markers of liver damage are ALP alkaline phosphatase, ALT alanine transaminase, and AST aspartate transaminase. Data are shown as means ± SEM. Statistical comparisons in c were to a single condition performed with ANOVA followed by Dunnett’s method and asterisks indicate significance (***P < 0.001).

Fig. 6. Delivery of NIPP1-CD.

a ID Salmonella delivery of NIPP1-CD caused more death (red, white arrows) in Hepa 1-6 cells compared to controls (P = 0.0021, n = 6). b In microfluidic tumor masses, delivery of NIPP1-CD caused more cell death (red) than bacterial controls. The percentage of dead cells increased with time as Salmonella invaded into cells and delivered protein (P = 0.042, 0.017, 0.014, 0.017, 0.024, 0.030, and 0.039 at 6.5, 7, 7.5, 8, 8.5, 9, and 9.5 h, respectively, n = 4). c For multiple (n = 4) tumor masses, as shown in b, NIPP1-CD significantly increased cell death (P = 0.0177). d NIPP1-CD ID Salmonella was intravenously administered to BALB/c mice with subcutaneous 4T1 tumors. After 31 days, Salmonella (black arrow, green) and delivered NIPP1-CD (white arrows, red) were dispersed throughout the tissue. Data are shown as means ± SEM. Statistical comparisons in a–c are two-tailed, unpaired Student’s t-tests with asterisks indicating significance (*P < 0.05; **P < 0.01). Images in a, b, and d are representative of 6, 4, and 3 independent biological samples, respectively. Scale bars in a and b are 100 µm, and in d the scale bar is 10 µm.

Fig. 7. Delivery of CT Casp-3 reduced tumor growth and increased survival.

a Delivery of CT Casp-3 killed Hepa 1–6 cells (white arrows, red). Cells invaded with control Salmonella (green, black arrows) or not invaded (yellow arrows) did not die. b CT Casp-3 caused significantly more death than controls (P < 0.0001, n = 6). c In microfluidic tumor masses, CT Casp-3 caused more cell death (red) than bacterial controls. d The fraction of dead cells was significantly greater in masses administered CT Casp-3 compared to controls (P = 0.0072, n = 3). e In 4T1 subcutaneous mammary tumors, intratumoral delivery of 4 × 10⁷ CFU of CT Casp-3 Salmonella decreased growth compared to bacterial controls, which delivered GFP (P = 0.0199, n = 6 mice per condition). f Lung metastases were established by injection of 5 × 10⁴ luciferase-expressing 4T1 cells into the tail vein of BALB/c mice. Mice were treated with either 1 × 10⁷ CFU of CT Casp-3 Salmonella or 10 mg/kg paclitaxel (n = 6), and imaged after IP injection of 100 µl of 30 mg/ml D-luciferin. Metastases grew exponentially in some paclitaxel-treated mice, but not in any treated with CT Casp-3 Salmonella. g On average, metastases grew over 85 times after treatment with paclitaxel. Treatment with CT Casp-3 Salmonella prevented growth more than eight times the initial volume (dotted line; P = 0.027). Metastatic volume is based on bioluminescence imaging and is reported relative to initial volume (see Supplementary Fig. 5 for individual mice). h In subcutaneous BNL-MEA liver tumors, intravenous delivery of 1 × 10⁷ CFU of CT Casp-3 Salmonella decreased growth compared to 10 mg/kg Sorafenib (P = 0.045) or saline (P = 0.031; n = 5 for saline and CT Casp-3 and n = 4 for Sorafenib). i In C57L/J mice with subcutaneous Hepa 1–6 liver tumors, intratumoral injection of 4 × 10⁷ CFU of CT Casp-3 Salmonella reduced volume to 11% of bacterial controls (ID Salmonella) by 19 days and 12% of saline controls by 22 days after injection (P = 0.00083, n = 3 mice per condition). j Delivery of CT Casp-3 significantly increased survival (P = 0.0224) and cured one mouse. Data are shown as means ± SEM. Statistical comparisons are one-sided (h) and two-sided (b, d, e, g, i) unpaired Student’s t-tests, with asterisks indicating significance (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). The statistical comparisons in j are log-rank tests with Bonferroni correction. Images in a are representative of six independent biological samples per condition. Scale bars in a and c are 20 and 100 µm, respectively.