Design of combination therapy for engineered bacterial therapeutics in non-small cell lung cancer

Synthetic biology enables the engineering of bacteria to safely deliver potent payloads to tumors for effective anti-cancer therapies. However, a central challenge for translation is determining ideal bacterial therapy candidates for specific cancers and integrating them with other drug treatment strategies to maximize efficacy. To address this, we designed a screening and evaluation pipeline for characterization of bacterial therapies in lung cancer models. We screened 10 engineered bacterial toxins across 6 non-small cell lung cancer patient-derived cell lines and identified theta toxin as a promising therapeutic candidate. Using a bacteria-spheroid co-culture system (BSCC), analysis of differentially expressed transcripts and gene set enrichment revealed significant changes in at least 10 signaling pathways with bacteria-producing theta toxin. We assessed combinatorial treatment of small molecule pharmaceutical inhibitors targeting 5 signaling molecules and of 2 chemotherapy drugs along with bacterially-produced theta toxin and showed improved dose-dependent response. This combination strategy was further tested and confirmed, with AKT signaling as an example, in a mouse model of lung cancer. In summary, we developed a pipeline to rapidly characterize bacterial therapies and integrate them with current targeted therapies for lung cancer.

www.nature.com/scientificreports/ EHL1301 11,12 . We induced therapeutic production in S. typhymurium and applied bacterial lysates to cells, which identified θ toxin as having the highest effect on viability among all bacterial toxins tested (Fig. 1b, Supplementary  Fig. 1). θ toxin is a pore forming toxin from C. perfringens 13 and was previously shown to have efficacy against murine colon cancer cells 11 . To model more physiologic conditions that occur within tumors containing poorly vascularized hypoxic regions where therapeutic efficacy is thought to be reduced 14 , we tested the effect of θ toxin on spheroids derived from a subset of our cell lines. We found that lysate of S. typhymurium expressing θ toxin 2D monolayer screen using MTT viability assay to study the response of 6 NSCLC lines to 10 engineered bacterially secreted toxins. Fresh lysates of engineered S. typhimurium EHL1301 were prepared and were normalized for optical density before adding to the NSCLC cultures grown in 96-well flat bottom plates. The heatmap represents the median of percent viability (n = 8). (c) Viability of 2 NSCLC spheroids under 3 treatment conditions-serially concentrated lysate of S. typhimurium producing: (1) GFP and θ toxin (in presence of AHL) abbr. "Stθ induced", (2) GFP and low amount of θ toxin due to leaky LuxR promoter (in absence of AHL) abbr. "Stθ uninduced", and, (3) GFP only (n = 4) abbr. "St", assessed by Cell Titer Glo 3D. Error bars represent standard deviation (n = 4). (d) Representative fluorescence microscopy images of 4-day old co-culture of 6 NSCLC spheroids (grey) with S. typhimurium producing GFP (green) and stained with hypoxia dye (red). Scale bar = 200 μM. (e) Viability of NSCLC-Salmonella co-culture spheroids using Cell Titer Glo 3D assay at day 7 (n = 6). Live S. typhimurium expressing θ toxin (Stθ) upon induction with AHL is abbreviated as "Stθ. " Significant change (**** = p < 0.0001, *** = p < 0.001, ** = p < 0.01, * = p < 0.5, n.s. = not significant) was determined by paired, two-tail t test, and error bars represent standard deviation. www.nature.com/scientificreports/ (Stθ) induced significantly higher cell death (p < 0.001) within the spheroids in a dose-dependent manner, as compared to lysate of non-induced S. typhymurium or lysate of S. typhymurium alone (Fig. 1c). Furthermore, these spheroids responded to purified θ toxin as well (ATCC, BTX-100) above 0.5 μg/mL concentration (Supplementary Fig. 2). Taken together, we identified θ toxin as a promising therapeutic to be delivered by bacteria for several NSCLCs. To selectively deliver therapeutics to tumor cells, bacteria need to be able to efficiently colonize tumors in spite of the production of payloads. Previous studies have established that live S. typhymurium can colonize murine colon cancer spheroids due to tumor-specific signatures, such as hypoxia, acidic pH and lactate [15][16][17] . We first assayed if live S. typhymurium were capable of colonizing hypoxic regions within the NSCLC spheroids. We found that the GFP-labeled bacteria colonized near dye-labeled hypoxic regions within the spheroid core ( Fig. 1d), where they persisted for the duration of one week ( Supplementary Fig. 3). Next, we assayed the viability of NSCLC spheroids containing live S. typhymurium when θ toxin expression was induced with AHL. We found that live Stθ significantly reduced viability of the majority of NSCLC spheroids, as compared to colonized live S. typhymurium alone (Fig. 1e). An extended analysis demonstrated that reduced viability was achieved in all spheroids within 2 weeks of Stθ exposure ( Supplementary Fig. 4). The ability of live Stθ to persistently colonize the spheroid core, while robustly reducing viability of heterogenous NSCLC cell types, encouraged the extension of this approach to in vivo conditions. We next explored the efficacy of Stθ in a mouse tumor model, which would locally deliver θ toxin in vivo while reducing systemic toxic effects. Specifically, we administered live bacteria intratumorally in tumor xenografts grown from a NSCLC cell line (H460) and induced the production of θ toxin 1 day after bacteria injection ( Fig. 2a). This assay resulted in a 2.5-fold reduction in tumor growth within a week (Fig. 2b, top, p-value < 0.001, Supplementary Fig. 5). Importantly, tumor control was achieved without inducing systemic toxicity after administration of high concentration of Stθ (4.5 × 10 8 CFU/mL and 40 μL per tumor), as assessed by weight change (Fig. 2b, bottom), and the absence of detectable S. typhymurium via IHC assay from peripheral organs at the endpoint (Fig. 2c, top). In addition, none of the peripheral organs showed signs of apoptosis, as indicated by the lack of cleaved caspase 3 signal, compared to the tumors (Fig. 2c, bottom, Supplementary Fig. 6). As immortalized Human Bronchial Epithelial Cells (HBECs) showed moderate level of response to the toxins ( Supplementary  Fig. 7), we routinely examined behavioral points of the mice such as activity, aggression, bite reflex, posture, presence of straub tail, seizures, and overall signs of morbidity upon Stθ exposure. Our experiments demonstrated lack of these behavioral changes. In summary, our studies point to the safety of this approach, reducing concerns regarding potential systemic toxicity associated with intratumoral use of live S. typhymurium in vivo.
Can we improve Stθ by combining with current standard of care chemotherapies as well as small molecule inhibitors being tested in clinical trials? To narrow down potential drugs to combine with Stθ treatment, we aimed to gain mechanistic insight into cellular pathways altered in two of the Stθ-responder spheroids (H460 and H1819, Fig. 3a). We compared genome-wide transcriptional profiling (RNA-seq) of tumor cells derived from spheroids housing live S. typhymurium in their hypoxic cores in presence or absence of producing θ toxin. Gene  www.nature.com/scientificreports/ Set Enrichment Analysis (GSEA) of Next-Gen sequencing revealed significant changes in at least 10 signaling pathways shared by both NSCLC cell lines upon Stθ treatment ( Fig. 3b and Supplementary Fig. 8, normalized p-value < 0.5 and falls discovery rate q-value < 0.5). Specifically, cell cycle checkpoint, PI3K/AKT/mTOR signaling and DNA repair pathways were among the most significantly enriched gene sets in both H460 and H1819. As top 50 differentially expressed genes from H1819 and H460 did not show modest overlap ( Supplementary  Fig. 9), we selected 7 small molecule inhibitors specific to lung cancer therapeutic landscape that are well known (NCT01294306, NCT00744900, NCT03392246) 18,19 to target the significantly enriched pathways in our analysis. We hypothesized-if under Stθ treatment, H1819 and H460 are dependent on the signaling through these pathways for their survival, a combinatorial strategy with each drug and Stθ would eliminate more cancer cells than the single treatments. Indeed, when combinatorically treated with each drug and Stθ, 4 out of 7 combinations showed robust efficacy across both spheroids (Fig. 3c) compared to drug-only or lysate of Salmonella expressing θ toxin. Notably, H1819 showed more resistance to 3 out of 7 combination therapy, and may represent more drug-resistant disease, as it was derived from a patient previously treated with both chemotherapy and radiation (ATCC). Interestingly, this benefit was not observed when the combinatorial treatment was tested on spheroids derived from mouse lung cancer cells with genetically modified TP53 and KRAS ( Supplementary  Fig. 10) suggesting this improved efficacy is specific to H1819 and H460 as indicated by their transcriptional profiling. Taking AKT signaling as an example, we observed high expression of phopho-AKT (Ser473) upon treatment with bacterial θ toxin and this signal was reduced in presence of MK2206 drug, (Fig. 4a,b) which is known to block AKT-phosphorylation at Ser473. This data suggests biological changes induced by the θ toxin treatment in NSCLC cells can be exploited to design new combination therapies. Specifically, NSCLC cells may upregulate AKT phosphorylation upon θ toxin treatment and blocking AKT phosphorylation with MK2206 achieves improved efficacy. Lastly, to test the efficacy of this combination therapy in vivo, we treated tumor xenografts with Stθ and MK2206 drug. For effective comparison with MK2206 + Stθ treatment, we designed multiple control cohorts: vehicle only, MK2206 alone, Stθ alone, live S. typhymurium alone, and, MK2206 + live S. typhymurium. Additionally, to observe the benefit of combination therapy at lower dosage and earlier time points, we intratumorally dosed bacteria (4.5 × 10 7 CFU/mL and 20 μL per tumor) and started treatment when tumors reached ~ 150 mm 3 . The cohort treated with MK2206 alone or with live bacteria producing no toxin showed no difference in tumor growth from the cohort treated with only vehicle (Fig. 4c). However, mice that received MK2206 + Stθ treatment showed significantly reduced tumor growth (p-value < 0.001) compared to other control cohorts. No significant reduction in total body weight was detected in any treatment groups (Fig. 4d). Taken together, Stθ showed improved efficacy and no additional toxicity when combined with MK2206 targeted therapy in vivo.

Discussion
We established a strategy to identify combination treatments for genetically-distinct solid tumors by assessing their response to bacteria therapies. Due to the rapid advances in synthetic biology, the ability to create engineered bacterial therapies far outpaces the throughput of animal-based testing, thus creating a major bottleneck for clinical translation. Here, we utilized NSCLC spheroids as a model for solid tumors which are three-dimensional and more physiological relevant in predicting bacterial therapy responses in mouse cancer models 11,12,15 . The combination with RNA-seq based identification of small molecule inhibitors enabled rapid and parallel assays for selection of effective therapies, which were then validated for safety and efficacy in a mouse model.
A limitation of our current pipeline is the low number of animals per cohort in the in vivo studies. Increasing the number of animals and assessing the overall survival upon treatment will be important, specifically if this pipeline is utilized for understanding the mechanism of resistance to bacterial therapeutics. Nevertheless, our approach addresses an impediment to the development of new combination therapies for solid tumors -namely, the narrow therapeutic window before significant systemic toxicity is observed [20][21][22] . As the toxins by themselves are not selectively detrimental to the cancer cells, selective delivery of the toxin by the live bacteria in the tumor is crucial to avoid systemic toxicity. Several species of bacteria can selectively colonize solid tumors, primarily due to reduced immune surveillance in tumor cores, and can be controlled to deliver therapeutic payloads after colonizing tumors [23][24][25] . Furthermore, attenuations of bacteria, such as those previously made to S. typhimurium reduce pathogenicity and immunogenicity, which have led to clinical trials where safety has been demonstrated 26,27 . Thus, our treatment strategy has the potential to increase therapeutic windows for NSCLC by combining the benefits of both bacteria therapies and previously developed safe pharmacological inhibitors.
Although this study focused on tumor cells in response to θ toxin treatment, going forward it can be similarly used to determine the response of immune cells to Stθ treatment to establish a comprehensive safety profile. Additionally, while we focused on NSCLC, our strategy can be expanded to other lung cancers and to the cancer of other organs. We envision fast and selective expansion of this pipeline to improve treatment efficacy and safety for solid tumors.
In vitro treatment, MTT and Cell Titer Glo 3D viability assays. For  H460 subcutaneous tumor, treatment with live bacteria and small molecule inhibitor. All animal experiments were approved by the Institutional Animal Care and Use Committee (Columbia University, protocol AC-AABQ5551). 5 × 10 6 H460 cells per mouse in 400 μL of 0.9% saline were injected subcutaneously in the right hind flank of 4-6 weeks old female NSG mice. Tumor volume was quantified using calipers to measure the length, width, and height of each tumor (V = ½ × L × W 2 ). When tumors reached an average size of 400mm 3 , mice were randomized and assigned to the different treatment groups. S. typhimurium was prepared from overnight culture, controlled for OD 0.1 and concentration of 4.5 × 10 7 0r 4.5 × 10 8 CFU/mL CFU/mL and intratumorally injected at a concentration of 5 × 10 8 cells per mL in 1X PBS with total volume of 20-40 μL per tumor 11,12 . 0.5 mL of 10 μM AHL was injected subcutaneously the day after bacterial treatment to induce therapeutic expression. MK2206 was dissolved in 30% captisol (Selleck Chemicals cat.no. S4592) in 1X PBS and injected in 240 mg/kg concentration by oral gavage, twice a week. Animal experiments were carried out independently at Danino lab and at The Oncology Precision Therapeutics and Imaging Core (OPTIC) in Columbia University. We confirm that all methods were performed in accordance with the relevant guidelines and regulations. www.nature.com/scientificreports/ Tissue collection, processing, staining and imaging. Mice were euthanized per IACUC protocol.
Ethics declarations. The authors declare no competing interests. All animal experiments were approved by the Institutional Animal Care and Use Committee (Columbia University, protocol AC-AABQ5551). All methods are reported in accordance with ARRIVE guidelines (https:// arriv eguid elines. org) for the reporting of animal experiments. For tumor-bearing animals, euthanasia was required when the tumor burden reached 2 cm in diameter or after recommendation by the veterinary staff. Euthanasia was carried out in a CO 2 chamber with a regulator and flow meter to maintain a fill rate of 30-70% of chamber volume per minute for 5 min. Gas flow was maintained for at least a 1 min after cessation of respiration. A secondary physical method of cervical dislocation is followed.