Bacterial strains and plasmid construction
Fifteen strains of Salmonella enterica serovar Typhimurium were used throughout the experiments (Supplementary Table 1). The parental control strain (Par) is derived from an attenuated strain of Salmonella (VNP20009) and has four deletions, ΔmsbB, ΔpurI, Δxyl, and Δasd. All plasmids contained a ColE1 origin and either chloramphenicol or ampicillin resistance (Supplementary Table 2). Three additional genomic knockouts (∆flhD, ∆sifA, and ∆sseJ) were created using a modified lambda red recombination protocol62 and primers with specific homology regions (Supplementary Table 3). Plasmids (Supplementary Table 2) were inserted into these base strains to generate strains that produce GFP (Accession KP294373) after cell invasion, re-express flhDC (Accession CP001363 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL14855]), report activation of PsifA (Accession CP001363 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL14855]), and produce Lysin E (Accession AF176034) after activation of PsseJ (Accession CP001363 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL14855]). The ID Salmonella strain was transformed to express a nanobody against β-actin (Chromotek), NIPP1-CD (Accession NM_174582.2), and CT Casp-3 (Accession AY219866). Details of gene deletions and plasmid construction are described below and the primers used are in Supplementary Tables 3 and 4.
All bacterial cultures (both Salmonella and DH5α) were grown in LB (10 g/L sodium chloride, 10 g/L tryptone and 5 g/L yeast extract). Resistant strains of bacteria were grown in the presence of carbenicillin (100 µg/ml), chloramphenicol (33 µg/ml), kanamycin (50 µg/ml), and/or 100 µg/ml of diamino-pimelic acid (DAP). All assembled DNA constructs were transformed into chemically competent DH5α E. Coli (New England Biolabs, Ipswich, MA) before electroporation into Salmonella. Electroporation was performed in 1 mm cuvettes at 1800 V and 25 µF with a time constant of 5 msec. All cloning reagents, buffer reagents, and primers were from New England Biolabs, Fisher Scientific (Hampton, NH), and Invitrogen, (Carlsbad, CA), respectively, unless otherwise noted.
Cell culture and animal models
Five cancer cell lines were used: 4T1 murine breast carcinoma cells; Hepa 1–6 and BNL-MEA (BNL 1ME A.7 R.1) murine hepatocellular carcinoma cells; MCF7 human breast carcinoma cells and LS174T human colorectal carcinoma cells (ATCC, Manassas, VA). The mouse cell lines were authenticated with CO1 barcoding and the human cell lines were authenticated with short tandem repeat profiling. All cancer cells were grown and maintained in Dulbecco’s Minimal Eagle Medium (DMEM) containing 3.7 g/L sodium bicarbonate and 10% fetal bovine serum. For microscopy studies, cells were incubated in DMEM with 20 mM HEPES buffering agent and 10% FBS. To generate tumor spheroids, single-cell suspensions of LS174T cells were transferred to PMMA-coated cell culture flasks [2 g/L Poly(2-hydroxy ethyl) methylacrylate (PMMA) in 100% ethanol, dried before use].
Multiple tumor models in mice (Mus musculus) were used. Both male and female mice, aged 4–7 weeks, were used. Delivery mechanisms and treatment efficacy were determined using subcutaneous syngeneic tumors formed with (1) 4T1 murine breast cancer and (2) BNL-MEA liver cancer cells implanted in BALB/c mice, and (3) Hepa 1–6 murine liver cancer cells in C57L/J mice. Clearance was determined in orthotopic 4T1 tumors implanted in the mammary fat pad of BALB/c mice. Toxicity and biodistribution were determined in tumor-free BALB/c and C57L/J mice. The effect on metastases was determined in BALB/c mice intravenously injected with 4T1 cells. All animal procedures complied with relevant ethical regulations and protocols were approved by the UMass Institutional Animal Care and Use Committee (IACUC). Mice were housed under a 12 h light/dark cycle at controlled room temperature of 72 °F and a relative humidity of 60%.
Gene deletions
Four genetic deletions were created (∆asd, ∆flhD, ∆sifA, and ∆sseJ) using a modified lambda red recombination protocol62. A parental strain (Par) was derived from Salmonella strain VNP20009, (ΔmsbB, ΔpurI, Δxyl) by deleting asd. Salmonella were transformed with pkd46 (Yale CGSC E. Coli stock center), centrifuged at 3000 × g and resuspended in ice-cold water. A PCR product was created to insert an in-frame deletion into the asd gene by PCR amplifying the FRT-CHLOR-FRT sequence from plasmid pkd3 (Accession AY048742.1) using primers vr266 and vr268 (Supplementary Table 3), which contain 50 basepair regions homologous to asd. This linear segment was transformed into Salmonella by electroporation. After recovery, colonies were screened for knockouts by colony PCR of the junction sites of the inserted PCR amplified products. Successful transformants were grown overnight at 43 °C to eliminate pkd46.
A similar process was used to delete flhD, sifA, and sseJ. Deletion of flhD prevents the formation and function of the hetero-oligomeric FlhDC complex63. Linear DNA segments were designed to insert in-frame deletions into the genes by amplification of the FRT-KAN-FRT sequence from plasmid pkd4 (Accession AY048743.1). Three sets of primers (vr121 and vr309 for flhD; vr432, and vr433 for sseJ; and vr434 and vr435 for sifA) added 50-basepair flanking regions that were homologous to the three genes (Supplementary Table 3). After electroporation and recovery, colonies were screened for knockouts by colony PCR. Successful transformants were plated on kanamycin plates (50 µg/ml) and grown overnight at 43 °C to remove pkd46.
Plasmid construction
Fifteen strains of Salmonella (Supplementary Table 1) were created by transforming twelve plasmids (Supplementary Table 2) into the parental strain (Par) and the gene knockout strains described above (i.e., ∆flhD, ∆sifA, and ∆sseJ). All of the plasmids contained a ColE1 origin and either chloramphenicol or ampicillin resistance (Supplementary Table 2). The intracellular-reporting strain of Salmonella was generated by transforming the parental strain (Par) with a plasmid containing PsseJ-GFP (plasmid P1; Supplementary Table 2). The construction of this plasmid was initiated by first creating a promoterless-GFP plasmid from pLacGFP and pQS-GFP64. The pQS-GFP plasmid (Accession KP294373) contains chloramphenicol resistance, the ColE1 origin of replication, and the asd gene (Accession CP001363 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL14855]). Expression of ASD is necessary in Δasd strains and creates a balanced lethal system that maintains gene expression in vivo. The Plac-GFP gene circuit (Accession KP294375) was amplified from plasmid pLacGFP with primers nd1 and nd2 (Supplementary Table 4). The PCR product and the plasmid were digested with Aat2 and Pci1 and ligated with T4 DNA ligase (NEB, catalog # M0202S). The PsseJ promoter (Accession CP001363 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL14855]) was amplified from the genome of SL1344 Salmonella using primers nd3 and nd4 (Supplementary Table 4). This PCR product and the backbone plasmid were ligated after digestion with XbaI and Pci1.
A strain that re-expresses flhDC (flhDC Sal, Supplementary Table 1) was created by transforming ∆flhD Salmonella with plasmid P2 (Supplementary Table 2). Plasmid P2 was formed from temporary plasmid P3. Plasmid P3 was formed by amplifying flhDC (Accession CP001363 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL14855]) from Salmonella genomic DNA using primers vr46 and vr47 (Supplementary Table 4), and ligating it into plasmid PBAD-his-mycA (Invitrogen; catalog # V430-01). The PCR product was digested with NcoI, XhoI and DpnI (NEB, catalog #s R0193S, R0146S and R0176L). The PBAD-his-myc plasmid (Accession X81838) was digested with NcoI and XhoI and treated with calf intestinal phosphatase (NEB, catalog # M0290) for 3 h. The PCR product was ligated into the plasmid backbone with T4 DNA ligase (NEB, catalog # M0202S).
The Plac-GFP-myc circuit was inserted into P3 by Gibson Assembly. [1] The insert (Plac-GFP-myc) was amplified from plasmid pLacGFP64 using primers vr394 and vr395 (Supplementary Table 4), which added homology regions to the backbone and added the myc tag. [2] The backbone plasmid (P3) was amplified using primers vr385 and vr386, which added homology to the insert. [3] Both PCR products were digested with DpnI for 3 h, [4] and ligated by Gibson Assembly (HiFi master mix, NEB, catalog # E2621L). The gene for aspartate semialdehyde dehydrogenase (asd) gene was inserted by Gibson Assembly by amplifying asd from genomic Salmonella DNA using primers vr424 and vr425, and amplifying the plasmid backbone with primers vr426 and vr427.
A strain that re-expresses flhDC and produces GFP after invasion (flhDC reporting, Supplementary Table 1) was created by transforming ∆flhD Salmonella with plasmid P4 (Supplementary Table 2). The PsseJ-GFP-myc genetic circuit was amplified from P1 using primers vr269 and vr270, and the backbone of plasmid P3 was amplified using primers vr271 and vr272. The two PCR products were ligated by Gibson Assembly.
To generate the PsifA intracellular promoter-reporter strain, the PsifA promoter (Accession CP001363 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL14855]) was cloned from Salmonella genomic DNA using primers nd5 and nd6 and inserted into P1 using XbaI and Pci1 creating plasmid P5. The PsifA reporter strain was created by transforming plasmid P5 into background Salmonella by electroporation. The generation of the PsseJ reporter strain is described above. To investigate lysis in Salmonella, lysis gene E (LysE) was put under control of PBAD. LysE was cloned using primers nd7 and nd8 and inserted into pBAD/Myc-His A (Invitrogen) using NcoI and KpnI to form plasmid P6.
Intracellular delivering (ID) Salmonella were created by cloning the Lysin E gene (Accession AF176034) behind the PsseJ promoter. LysE was amplified using primers nd9 and nd10 and cloned into P1 using XbaI and Aat2. The Plac-GFP circuit was added to this plasmid by cloning it from plasmid pLacGFP using primers nd11 and nd12 and inserting using SacI to create plasmid P7. This plasmid constitutively expresses myc-tagged GFP to identify bacteria in both live-cell and fixed-cell assays.
Two strains of ID Salmonella with deletions of the sseJ and sifA genes were created by transforming the knockout strains described above (∆sifA and ∆sseJ) with plasmid P7. Similar to ID Salmonella, these two strains contain the PsseJ-LysE construct and constitutively express myc-tagged GFP. Note the distinction between the effector gene sseJ, which is necessary for vacuolar escape, and its promotor PsseJ, which activates in SCVs.
ID Salmonella that re-expresses flhDC (IDf + Sal) was created by transforming ∆flhD with plasmids P8. Plasmid P8 was created by amplifying the PsseJ-LysE gene circuit from P7 using primers vr398 and vr399, and ligating it into plasmid P2 using Gibson Assembly. The P2 backbone plasmid was amplified using primers vr396 and vr397.
A strain of ID Salmonella that constitutively expresses luciferase (ID Sal-luc; Supplementary Table 1) was created by cloning Plac-luc from pMA3160 (Addgene) using primers ch1 and ch2. The P7 plasmid backbone was amplified with primers ch3 and ch4 and the pieces were ligated by Gibson Assembly to form plasmid P9 (Supplementary Table 2).
To create ID Salmonella that express anti-b-actin nanobody (NB), PBAD-inducible nanobody was cloned in place of flhDC in plasmid P8. The actin nanobody (Chromotek, catalog # acr) was amplified using primers vr466 and vr467. The delivery plasmid backbone was amplified using primers vr448 and vr449. The two PCR products were ligated by Gibson Assembly to create plasmid P10.
To create ID Salmonella that express the central domain of NIPP1 (NIPP1-CD, Accession NM_174582), NIPP1-CD-myc was cloned into plasmid pLacGFP. NIPP1-CD-myc and the backbone plasmid were amplified using primers nd13-nd16 ligated by Gibson Assembly. The pLac-NIPP1-CD circuit was cloned using primers nd11 and nd17 (Supplementary Table 4) and inserted into P7 using SacI to create plasmid P11.
To create ID Salmonella that intracellularly deliver CT caspase-3 (CT Casp-3, Accession AY219866), parental Salmonella were transformed with plasmid P12. This plasmid was created by PCR amplifying template DNA encoding for CT caspase-3 using primers, vr450 and vr451 from the constitutively two-chain (CT) caspase-3 encoding plasmid pC3D175CT. The pC3D175CT plasmid (Hardy Lab DNA archive Box 7, line 62) was constructed similarly to the caspase-6 CT expression construct65 using Quikchange mutagenesis on a construct encoding full-length human caspase-3 in a pET23 expression vector (Addgene). Plasmid pC3D175CT encodes human caspase-3 residues 1–175, followed by a TAA stop codon, a ribosome binding sequence and the coding sequence for a start methionine and an inserted serine followed by the coding sequence for residues 176–286 with a six histidine tag appended. The backbone of plasmid P8 was PCR amplified using primers vr448 and vr449 and the PCR products were ligated as described above.
Invasion assays and immunocytochemistry
Mouse 4T1 or human MCF7 cells were grown on coverslips for fixed-cell imaging or on well plates for live-cell imaging. For fixed imaging, Salmonella were added to 4T1 cultures at a multiplicity of infection (MOI) of 10. For live-cell imaging, Salmonella added to MCF7 cultures at an MOI of 25. The bacteria were allowed to infect cells for two hours. The cultures were then washed five times and treated with 50 µg/ml gentamicin in culture medium to remove extracellular bacteria. Live cells on well plates were directly imaged microscopically.
To obtain detailed images, cells on coverslips were fixed with 10% formalin after 6 or 24 h of incubation. Fixed coverslips were blocked with staining buffer (PBS with 0.1% Tween 20, 1 mM EDTA, and 2% bovine serum albumin) for 30 min. The Tween 20 in this buffer selectively permeabilizes mammalian cell membranes, while leaving bacterial membranes intact. After permeabilization, coverslips were stained to identify Salmonella, released GFP, vacuolar membranes and/or intracellular f-actin with (1) rabbit anti-Salmonella polyclonal antibody (Abcam, catalog # ab35156; 1:200 dilution) or FITC-conjugated rabbit anti-Salmonella polyclonal antibody (Abcam, catalog # ab69253; 1:100 dilution) (2) rat anti-myc monoclonal antibody (Chromotek, catalog # 9e1-100; 1:200 dilution), (3) rabbit anti-LAMP1 polyclonal antibody (Abcam, catalog # ab24170; 1:200 dilution), and (4) Alexaflor-568-conjugated phalloidin (ThermoFisher, catalog # A12380), respectively.
Immunohistochemistry
Excised tumor sections were fixed in 10% formalin, embedded in paraffin and cut into 5 µm sections. Antigen retrieval was performed by incubating deparaffinized sections in 20 mM sodium citrate (pH 7.6) buffer for 20 min at 95 °C. Samples were rehydrated with DI water and Tris buffered saline with 0.1% Tween 20 (TBS-T). Prior to staining, tissue sections were blocked with Dako blocking buffer (Dako, catalog # X0909). Tissue sections were stained to identify Salmonella and GFP with (1) FITC-conjugated rabbit anti-Salmonella polyclonal antibody (Abcam, catalog # ab69253; 1:100 dilution), and (2) either rat anti-myc monoclonal antibody (Chromotek, catalog # 9e1-100; 1:100 dilution) or rat anti-GFP monoclonal antibody (Chromotek, catalog # 3h9-100; 1:100 dilution). Sections were incubated with Alexaflor-568 goat anti-rat secondary antibodies (ThermoFisher, catalog # A11077) and counterstained with DAPI-containing mountant (ThermoFisher, catalog # P36962).
Microscopy
Samples were imaged on a Zeiss Axio Observer Z.1 microscope. Fixed cells on coverslips were imaged with a ×100 oil immersion objective (1.4 NA). Tumor sections were imaged with ×10 and ×20 objectives (0.3 and 0.4 NA, respectively). Time-lapse fluorescence microscopy of live cells in well plates and tumor-chip devices were housed in a humidified, 37 °C environment and imaged with ×5, ×10, ×63, or ×100 objectives (0.2, 0.3, 1.4, and 1.4 NA, respectively). Fluorescence images were acquired with either 480/525 or 525/590 excitation/emission filters. All images were background subtracted and contrast was uniformly enhanced. Some image analysis was automated using computational code (MATLAB, Mathworks).
Tumor masses in microfluidic devices
Microfluidic tumor-on-a-chip devices were developed in our laboratory to quantify bacterial invasion52,66. Soft lithography was used to create a multilayer device with 12 tumor chambers. Master chips were constructed by spin coating a layer of SU-8 2050 onto a silicon wafer at 1250 RPM for 1 min for a thickness of 150 µm. The silicon wafer with an overlaid mask printed with the microfluidic designs was exposed to UV light (22 J/cm2) for 22 sec. After baking, wafers were developed in PGMEA developing solution for 10 min. PDMS (Sylgard 184) at ratios of 9:1 and 15:1 were used for the channel and valve layers, respectively. After aligning, the layers were baked for 1 h at 95 °C in order to covalently bind. The PDMS device was adhered to a glass slide by placing both in a plasma cleaner (Harrick) for 2.5 min. Prior to use, devices were sterilized with 10% bleach and 70% ethanol, and equilibrated with media (DMEM with 20 mM HEPES). Valve actuation was used to position tumor spheroids in the tumor chambers.
Intracellular Salmonella in cells and tumors
To observe invasion into cancer cells, Salmonella were administered to mouse 4T1 breast cancer cells on coverslips using an invasion assay. The cells and bacteria were stained with phalloidin and anti-Salmonella antibodies and imaged with ×100 oil immersion microscopy. To measure invasion into cells in tumors, BALB/c mice with subcutaneous 4T1 tumors were administered 2 × 106 CFU of PsseJ-GFP Salmonella (Supplementary Table 1) by intratumoral injection. Ninety-six hours after injection, tumors were excised and stained to identify Salmonella and the GFP reporter produced by intracellular Salmonella. The fraction of intracellular Salmonella was determined by identifying Salmonella (n = 1258) in four images and determining the number that colocalize with GFP.
Effect of flhDC on invasion into cells and tumor masses
To determine the effect of expressing flhDC on invasion, 4T1 cells were grown on glass coverslips and administered flhDC+ and flhDC− Salmonella at an MOI of 10. Prior to administration, flhDC+ Salmonella were grown in LB with 20 mM arabinose to induce flhDC expression. For flhDC+ bacteria, 20 mM arabinose was also added to the mammalian co-cultures to maintain gene expression. Control (flhDC−) bacteria were grown without arabinose. Eighteen hours after invasion, the cells were stained to identify intracellular Salmonella. Invasion was quantified in six images from three coverslips per condition by randomly identifying 20 cancer cells from the DAPI channel and determining if there was Salmonella staining within 10 µm of the nucleus.
To quantify invasion into tumor masses, flhDC-inducible, intracellular-reporting Salmonella (Supplementary Table 1) were administered to tumor-on-a-chip devices. Bacteria-containing medium (DMEM with 20 mM HEPES) was perfused through the devices for one hour at 3 µm/min for a total delivery of 2 × 106 CFU per device. Two conditions (flhDC+ and flhDC−; n = 6 chambers each) were compared. Similar to monolayer culture, flhDC+ Salmonella were grown in LB with 20 mM arabinose prior to administration, and 20 mM arabinose was added to the co-culture medium to maintain gene expression. Bacterial administration was followed by bacteria-free media (with 20 mM HEPES) for 48 h. Devices were imaged at 30 min intervals. Invasion was quantified at 31 h by measuring GFP expression by invaded bacteria.
Design of ID Salmonella
To determine the intracellular activation of the PsifA and PsseJ promoters, Salmonella with GFP-reporting constructs (Supplementary Table 1) were administered to MCF7 cancer cells at an MOI of 25. Extracellular promoter activity was determined as the average fluorescence intensity of bacteria from three wells, and normalized to the average intensity of PsseJ bacteria. The increase in promoter activity following invasion was determined by comparing the average intensity of bacteria in cells to extracellular bacteria. To determine bacterial death caused by lysin E, Salmonella strain PBAD-LysE (Supplementary Table 1) was grown in LB to an OD of 0.25 and induced with 10 mM arabinose. Growth and death rates were determined by fitting exponential functions to bacterial density.
To visualize and quantify triggered intracellular lysis and GFP delivery, ID Salmonella were administered to MCF7 cancer cells at an MOI of 25. Cultures were washed five times and treated with 50 µg/ml gentamicin to remove extracellular bacteria. Transmitted and fluorescent images were acquired at ×20 every 30 min for 10 h. Two hundred cancer cells were randomly selected and scored based on bacterial invasion and lysis. Times of lysis for individual bacteria (within the cancer cells) were determined as the moment of disappearance from the fluorescent time-lapse images of intracellular GFP-expressing bacteria. The lysis fraction was the number of cancer cells with lysed bacteria over the total number of observed cells. The rate of intracellular lysis was determined by fitting an exponential function to the cumulative fraction of cells with lysed bacteria. To generate images of bacterial lysis and GFP delivery, ID Salmonella were administered to 4T1 cancer cells at an MOI of 10. Coverslips were fixed and stained for Salmonella and released GFP. To quantify bacterial protein content, ID Salmonella were suspended at four densities: 106, 107, 108, and 109 bacteria and compared to a GFP standard at three concentrations: 1, 10, and 100 ng per 40 µl Laemmli buffer. GFP was identified in immunoblots with rat anti-GFP monoclonal antibody (Chromotek, catalog # 3h9-100; 1:1000 dilution).
Delivery to tumors
To identify and quantify GFP delivery to tumor cells, BALB/c mice with subcutaneous 4T1 tumors were administered 2 × 106 CFU of ID Salmonella by intratumoral injection. Ninety-six hours after bacterial injection, tumors, liver and spleens were excised. Tumor sections were stained with anti-GFP antibody (Abcam, catalog # ab6556; 1:100 dilution). To compare the amount of delivered protein in the organs of these mice, tumors, livers and spleens were snap-frozen in liquid nitrogen and treated with a buffer containing 50 mM Tris-HCl, 0.3% Triton-X 100, 0.1% NP-40 and 0.3 M NaCl to lyse mammalian cells but not bacterial membranes. Immunoblotting was performed with anti-GPF (Abcam, catalog # ab6673; 1:1000 dilution) and anti-β-actin (GeneTex, catalog # GTX26276, clone AC-15; 1:1000 dilution). To quantify the amount of protein delivered to tumors, the tumor lysates were run on a similar immunoblot and compared to a GFP standard at 0.43, 1.3, and 3.9 pmols. The amount of GFP per tumor was determined as the lysate concentration multiplied by the lysate volume, normalized by the tumor mass.
To measure the delivery of anti-actin nanobodies, NB and ID Salmonella were administered to 4T1 cancer cells at an MOI of 10. The extent of binding to β-actin was determined by immunoprecipitation. Twenty-four hours after invasion, cells were harvested and centrifuged at 600 × g for 10 min. The cell pellet was lysed, homogenized, and incubated overnight with 50 µl of anti-FLAG purification resin (Biolegend, catalog # 651502). Beads were boiled for 5 min and loaded onto SDS-PAGE gels. β-actin was identified with mouse anti-actin monoclonal antibody (Cell Signaling Technology, catalog # 8H10D10; 1:1000 dilution).
Protein release from Salmonella and SCVs
In order to quantify GFP release from vacuoles, ID Salmonella were administered to 4T1 cancer cells on coverslips at an MOI of 10. At 6 and 24 h, one set of coverslips were fixed, permeabilized and stained with using anti-Salmonella, anti-myc, and anti-LAMP1 antibodies. Acquired images were analyzed to quantify (1) the location of released GFP and (2) the location of Salmonella lysis. The fraction of vacuolar GFP was determined as the area of released GFP that was colocalized with LAMP1, normalized by the total area of released GFP. The location of bacteria lysis was determined by identifying all bacteria in seven 86.7 × 66.0 µm regions based on anti-Salmonella staining. Lysed bacteria were identified as those that colocalized with released GFP. Each lysed bacterium was classified as either vacuolar or cytoplasmic by its colocalization with LAMP1. To visualize the localization of released GFP in cells, a second set of fixed coverslips were stained with anti-Salmonella and anti-myc antibodies, and phalloidin, to visualize cell structures and boundaries.
To measure the rate of GFP dispersion through cells after lysis, MCF7 cancer cells were grown on 96-well plates with coverslip glass bottoms (ThermoFisher, catalog #160376). ID Salmonella were administered at an MOI of 25. After removing extracellular bacteria with gentamycin, transmitted and fluorescence images were acquired at ×63 every minute for 14 h. Intensities were measured on lines passing through bacterial centers starting when bacteria were intact until diffusion was complete. Cytosolic diffusivity was determined by fitting the spatiotemporal intensity profiles to the radial diffusion equation.
To determine the dependence of protein release on residence in SCVs, 4T1 cancer cells were grown on coverslips and infected with ΔsifA, ΔsseJ, or ID Salmonella at an MOI of 10 (n = 3 for each condition). All three of these strains contained the PsseJ-LysE and Plac-GFP-myc gene circuits (Supplementary Table 1). At 6 h after invasion, the cancer cells were fixed, permeabilized and stained for Salmonella and released GFP. The lysis fraction was calculated in MATLAB as number of lysis pixels (GFP positive) divided by the total (GFP or Salmonella positive).
Control of invasion
To determine the dependence of protein delivery on invasion and intracellular lysis, flhDC Sal (contains PBAD-flhDC, but not PsseJ-LysE) and IDf+ Sal (contains PBAD-flhDC and PsseJ-LysE; Supplementary Table 1) were administered to 4T1 cancer cells in well plates and on coverslips infected at an MOI of 10. Prior to invasion, flhDC expression was induced in the flhDC+ cultures with 20 mM arabinose. Expression was maintained in the subsequent mammalian co-cultures with addition of 20 mM arabinose. Control (flhDC-) bacteria were grown without arabinose. For flow cytometry, cells were processed into a single-cell suspension with 0.05% trypsin (ThermoFisher, catalog # 25300-054) and fixed with 5% formaldehyde in PBS and 1 mM EDTA. After permeabilization with 0.1% Tween, cells were stained with FITC-conjugated anti-Salmonella antibody (Abcam, catalog # ab69253; 1:100 dilution), and anti-myc monoclonal antibody (Chromotek, catalog # 9e1-100; 1:100 dilution), followed by Dylight 755 secondary antibody (Thermofisher, # SA5-10031; 1:100 dilution) staining against the primary anti-myc antibody. Fluorescence minus one (FMO) of each sample were used as gating controls for each fluorophore. Samples were analyzed on a custom-built flow cytometer (dual LSRFortessa 5-laser, BD; with BD FACS Diva software). All fluorophores were compensated with compensation beads (BD, catalog # 552845) and did not carry more than 2% bleed over into any other channel. Cells were gated to exclude debris, isolate single cells, and quantify the percentage with intracellular Salmonella and delivered protein (Supplementary Fig. 7). For microscopy, coverslips were fixed, permeabilized and stained for released GFP. Protein (GFP) delivery was quantified in MATLAB as the number of pixels stained for GFP-myc normalized by total in the PsseJ-LysE−, flhDC− condition.
To determine the effect of flhDC on protein delivery, BALB/c mice with subcutaneous 4T1 tumors were injected with 2 × 106 CFU of IDf+ Salmonella (flhDC+) and ∆flhD Salmonella (flhDC−) via the tail vein. Prior to injection, Salmonella were grown without arabinose to prevent flhDC expression until after tumor colonization. At 48 and 72 h after bacterial injection, 100 µg of arabinose in 400 µl of PBS was injected intraperitoneally (IP) into flhDC+ mice to induce expression. Ninety-six hours after injection, tumors were excised, sectioned and stained with primary anti-GFP antibody (Chromotek, catalog # 3h9-100; 1:100 dilution) and followed by anti-rat secondary (Life Technologies, catalog # A11077). Delivery was quantified at 20 random points in the transition zone of each acquired tumor image. A point was scored as positive if a cell within 20 µm contained delivered GFP.
Temporal colonization, biodistribution, and toxicity of ID Salmonella
To determine the bacterial density in tumors over time, 2 × 107 CFU ID Salmonella that express firefly luciferase (ID Sal-luc, Supplementary Table 1) were intravenously injected into BALB/c mice with orthotopic 4T1 tumors in the mammary fat pad. At 24, 48, 72, 168, 336 h after bacterial injection, mice were injected IP with 100 µl of 30 mg/ml D-luciferin in sterile PBS and imaged with an IVIS animal imager (PerkinElmer, SpectrumCT). Bacterial density was determined as the photon flux. After acquiring the final image at 14 days, the bacterial density was measured by excising tumors and culturing the homogenized tissue on agar plates.
To measure the biodistribution, tumor-free BALB/c mice were injected with 1 × 107 ID Salmonella via the tail vein. Control mice were injected with sterile saline. After 14 days, six organs were excised and weighed: spleen, liver, lung, kidney, heart, and brain. Organs were minced and cultured on agar plates. A second experiment was performed to determine the biodistribution at earlier times. Tumor-free C57L/J mice were intravenously injected with either saline (control), 1 × 107 ID Salmonella, or 1 × 107 CT Casp-3 Salmonella. At two timepoints (6 h and 7 days), tissues were excised and weighed from two separate groups of mice. Bacterial densities were determined by mincing organs and culturing on agar plates.
To measure the toxicity of ID Salmonella, four tumor-free BALB/c mice were injected with 1 × 107 ID Salmonella via the tail vein. After 14 days, whole blood was isolated by percutaneous cardiac puncture. A second experiment was performed to determine the toxicity of bacterial delivery of CT Casp-3. Tumor-free C57L/J mice were intravenously injected with saline (control), 1 × 107 ID Salmonella, or 1 × 107 CT Casp-3 Salmonella. Sera were collected 7 days after injection. Chemistry profiling and comprehensive hematology was conducted on the serum and whole blood samples by IDEXX Laboratories (Grafton, MA).
Cytotoxicity of CT Casp-3 and NIPP1-CD
To measure the efficacy of delivering protein drugs, NIPP1-CD and CT Casp-3 Salmonella were administered to Hepa 1–6 liver cancer cells at an MOI of 10. Cell death was detected with 500 ng/ml ethidium homodimer and calculated as the fraction of dead Salmonella-invaded cells over the total number of Salmonella-invaded cells. To measure cell death in tumor masses, media containing 2 × 107 CFU/ml NIPP1-CD or CT Casp-3 Salmonella and 500 ng/ml ethidium homodimer was perfused through tumor-on-a-chip devices for 1 h at 3 µm/min. Bacterial administration was followed by bacteria-free media. Transmitted and fluorescence images were acquired every 30 min for 24 h. Death was quantified as the percentage of the tumor mass stained with ethidium homodimer at 24 h.
To measure efficacy and the extent of delivering NIPP1-CD, 1 × 107 CFU/mouse of NIPP1-CD Salmonella or saline (controls) were administered to BALB/c mice with subcutaneous 4T1 tumors by intravenous injection. Tumors were measured twice a week and volumes were calculated with the formula (length)*(width2)/2. After 31 days, tumors were excised and stained for Salmonella (Abcam, catalog # ab69253; 1:100 dilution) and NIPP1-CD with antibodies to the c-terminal myc tag (Chromotek, catalog # 9e1-100; 1:100 dilution), followed by a secondary antibody to the myc specific antibody (Life Technologies, catalog # A11077; 1:100 dilution). In the acquired images, DAPI staining was used to identify regions with viable nucleated cells. The average delivery of NIPP1 was determined as the fraction of the viable cell area that positively stained for delivered NIPP1-CD.
To measure the efficacy of delivering CT Casp-3, bacteria were administered to BALB/c mice with subcutaneous 4T1 tumors, BALB/c mice with subcutaneous BNL-MEA tumors, and C57L/J mice with subcutaneous Hepa 1–6 tumors. For the 4T1 and Hepa 1–6 tumor models, groups of mice (n = 6 for 4T1; n = 3 for Hepa 1–6) received intratumoral injections of saline, 4 × 107 CFU ID Salmonella, or 4 × 107 CFU CT Casp-3 Salmonella. The ID Salmonella (bacterial) control established the baseline effect of bacteria colonization and intracellular lysis. Every 5 days, tumors were injected with bacteria or saline. For the BNL-MEA model, once tumors reached 100 mm3, mice were intravenously injected every five days with 1 × 107 CFU/mouse of CT Casp-3 Salmonella (n = 5), 10 mg/kg Sorafenib (n = 4), or saline (n = 5). Sorafenib (10 mg/kg) is the standard-of-care for liver cancer. Tumors were measured twice a week and volumes were calculated with the formula (length)*(width2)/2. Mice were sacrificed when tumors reached 1000 mm3.
To measure the efficacy of CT Casp-3 in secondary tumor sites, lung metastases were formed by injection of 5.0 × 104 luciferase-expressing 4T1 cells into the tail veins of female BALB/c mice. Relative metastasis volume was determined by injecting the mice IP with 100 µl of 30 mg/ml D-luciferin in sterile PBS and imaged with an IVIS animal imager (PerkinElmer, SpectrumCT). When lung colonization was detected, mice were injected intravenously every 5 days with 1 × 107 CFU CT Casp-3 ID Salmonella or injected IP with 10 mg/kg paclitaxel. Tumor burden was monitored weekly with BLI until study endpoint.
Statistical methods
Statistical analysis was performed in Excel (Microsoft Office Professional Plus 2016) and GraphPad Prism 9.2.0. Comparisons of two populations were made with two-tailed, unpaired Student’s t-tests. Comparisons of multiple conditions were made using ANOVA with a Bonferroni correction. Comparison of multiple conditions to a single control were performed with ANOVA followed by Dunnett’s method. In some large datasets, outliers were removed using the ROUT method, with a Q (maximum false discovery rate) of 1%. To compare survival, log-rank tests were used with Bonferroni correction. All measurements were taken from distinct samples. Values are reported as means ± standard errors (SEMs). Statistical significance was confirmed when P < 0.05.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
