Engineered E. coli for the targeted deposition of therapeutic payloads to sites of disease

Jason Lynch Massachusetts General Hospital Coral Coral González-Prieto Massachusetts General Hospital Analise Reeves Synlogic Urmila Powale Tufts University Neha Godbole Massachusetts General Hospital Jacqueline Tremblay Tufts Cummings School of Veterinary Florian Schmidt University of Bonn https://orcid.org/0000-0002-9979-9769 Hidde Ploegh Boston Children's Hospital https://orcid.org/0000-0002-1090-6071 Jonathan Glickman Beth Israel Deaconess Medical Center John Leong Tufts University Charles Shoemaker Tufts Cummings School of Veterinary Wendy Garrett Harvard University https://orcid.org/0000-0002-5092-0150 CAMMIE LESSER (  clesser@mgh.harvard.edu ) Massachusetts General Hospital

10 secretion system. We first monitored the levels of bacteria shed in the feces of mice orally 200 inoculated with EcN or ). After the administration of a single dose of ~10 8 201 colony forming units (CFU) via oral gavage, EcN and PROT3EcT-4 were each detected in the 202 shed feces at ~10 5 CFU/gm/day for at least 14 days, with no significant decrease in fecal 203 shedding over that period (Fig. 5b). Each of the 158 Nb-secreting PROT3EcT-4 colonies isolated 204 from a total of 4 mice at 2-, 5-, and 14-days post-inoculation secreted Nbs, indicating that their 205 T3SA remained fully functional and that the alr-Nb-expressing plasmid was maintained . No significant weight loss was observed over the course of these experiments 207 ( Fig. S4d). 208 Next, we examined the biogeography of PROT3EcT-3 and EcN within the intestines of 209 mice inoculated with variants of each that constitutively express the luciferase-producing 210 luxCDABE operon 36 . The two strains exhibited equivalent luciferase activity when grown in vitro 211 ( Fig. S5a-b). Eight days post-oral inoculation of the mice, each strain exhibited similar patterns 212 of luciferase expression in explanted sections of their ileum, cecum, and proximal colon (Fig. 213 5d). In complementary studies, CFUs of PROT3EcT-3 and EcN found in various regions of the 214 intestines of orally inoculated mice were equivalent and exceeded 5 x 10 9 CFU/g of contents 215 (Fig. S5c). 216 To confirm that the modified T3SS present in PROT3EcT is actively transcribed within 217 the intestines of mice, we developed a luxCDABE-based reporter that is only activated when the 218 Mxi-Spa operons, which encode the modified T3SA, are transcribed (Fig. S5d). Variants of 219 PROT3EcT-3, but not EcN, that contained this reporter demonstrated luciferase production in 220 vitro ( Fig. S5e-f) and within the explanted cecum, proximal colon, and ileum of inoculated mice 221 (Fig. 5e). These observations establish that the T3SS present within PROT3EcT-3 is expressed 222 within the intestines of mice and does not interfere with EcN colonization. 223

PROT3EcT can be engineered to secreted functional Nb TNF 224
To establish the use of PROT3EcT as a therapeutic platform, we focused efforts on 225 investigating its efficacy in the treatment of inflammatory bowel diseases (IBD). The etiologies of 226 ulcerative colitis and Crohn's disease, collectively termed IBD, are complex and thought to be 227 driven by host genetic, environmental, and microbiota factors. Yet, both diseases exhibit chronic 228 inflammation accompanied by increased levels of pro-inflammatory cytokines 37 . Monoclonal 229 antibodies (mAb) that target the pro-inflammatory cytokine TNFα, e.g., infliximab and 230 adalimumab, are highly efficacious in controlling severe disease and in improving the quality of 231 life of patients with IBD 38 . However, given the systemic administration of these therapeutics, 232 patients receiving these agents are immunosuppressed and at increased risk of developing life-233 threatening infections and lymphoma 39 . 234 We hypothesized that the targeted delivery of anti-TNFα Nbs via PROT3EcT to the 235 intestines could reduce intestinal inflammation. To investigate this possibility, we first isolated 236 anti-TNFα Nbs from alpacas immunized with recombinant mouse TNFα, including one that 237 binds with high affinity (EC50 0.1 nM) and neutralizes TNFα (IC50 0.1 nM) (Table S1 and Figure  238 S6). We generated both monomeric and dimeric variants of Nbs engineered with an OspC2 239 secretion sequence. The dimer was secreted much more efficiently than the monomer (Fig. 6a) 240 and PROT3EcT secreted dimeric Nbs were as effective as E. coli-purified dimeric Nbs in 241 blocking TNFα-induced death of mouse L929 cells (Fig. 6b). 242

TNF-PROT3EcT inhibits the development of disease in a mouse model of colitis. 243
To investigate the utility of PROT3EcT as a live biotherapeutic for the treatment of 244 intestinal inflammation, we interrogated the therapeutic efficacy of TNF-PROT3EcT-4 245 (PROT3EcT that constitutively secrete the anti-TNF dimeric Nb) in the suppression of TNBS 246 (2,4,6-trinitrobenzene sulfonic acid)-induced colitis. In this preclinical model of IBD, a mixture of 247 TNBS, a hapten, and ethanol, which disrupts the mucosal barrier, is instilled into the colon via 248 rectal administration. TNBS bound to colonic tissue proteins subsequently induces inflammation 249 driven by pro-inflammatory cytokines, including TNFα 40 . As previously reported 41 , mice treated 250 intraperitoneally with a neutralizing anti-TNF monoclonal antibody (1-day prior and 2-and 4-251 days post-administration of TNBS) were protected from weight loss, colon shortening, and 252 histologic evidence of colitis (Fig. S7). 253 To test the therapeutic efficacy of TNF-PROT3EcT, we orally administered 10 8 CFU of 254 TNF-PROT3EcT-4, PROT3EcT-4, or PBS to mice one day before as well as two and four days 255 after they received TNBS (Fig. 6c). Animals that received bacteria shed equivalent levels of 10 5 -256 10 6 CFU/g of TNF-PROT3EcT-4 and PROT3EcT-4 in their feces (Fig. 6d). Treatment with TNF-257 PROT3EcT-4 significantly reduced weight loss, blunted colon shortening, and decreased or 258 completely abrogated epithelial injury and inflammation in the mucosa, including less 259 polymorphonuclear and mononuclear cell infiltration ( Fig. 6e-h). In contrast, PROT3EcT-4 did 260 not provide any protection as assessed by each of these metrics, demonstrating that the 261 therapeutic efficacy afforded by TNF-PROT3EcT is due to the secreted TNF-neutralizing Nb and 262 not EcN intrinsic. 263 Given that enemas are commonly used for drug delivery for patients with IBD, we also 264 investigated the efficacy of intrarectally delivered TNF-PROT3EcT-4 in limiting TNBS-induced 265 colitis using the same dosing strategy as described above (Fig. S8a). As with oral delivery, 266 intrarectally delivered TNF-PROT3EcT-4, but not PROT3EcT-4, ameliorated weight loss, colon 267 shortening, and colitis ( Fig. S8b-e). Thus, when delivered either orally or intrarectally, TNF-268 PROT3EcT-4 provides protection against TNFα-driven inflammation in the TNBS model of 269 colitis. 270 13 Therapeutic strains that cannot compete for and establish a replicative niche within the 271 colon may be useful if administered repeatedly to patients. To assess whether treatment with 272 similarly engineered DH10b E. coli also suppresses colonic inflammation, we developed T3EcT, 273 a variant of mT3Ec-Mxi-Spa engineered with the chromosomally encoded PJ23119 VirB gene 274 cassette, and a variant of T3EcT that secretes Nb TNF , TNF-T3EcT. After establishing that TNF-275 T3EcT constitutively secreted Nb TNF into its surroundings (Fig. S8f), TNF-T3EcT and T3EcT were 276 administered orally or intrarectally to mice using the strategy outline above (Fig. S8a). Orally 277 administered TNF-T3EcT provided no protection, likely due to its inability to colonize the 278 intestines as assessed by fecal shedding (Fig. S8b). In contrast, treatment with rectally 279 delivered TNF-T3EcT, but not T3EcT, suppressed colitis ( Fig. S8-c-e), likely reflecting repeated 280 transient deposition of anti-TNF Nbs in the colon. These observations demonstrate that the Ipa-281 Mxi secretion-based platform can be extended to additional E. coli strains. 282 To address whether bacterial secreted Nbs are restricted to the gut, we measured Nb 283 levels in the colonic contents, colon tissue homogenates, and serum of mice treated with each 284 strain across each of the TNBS experiments. To measure the anti-TNF Nb, we used a direct 285 ELISA, which is also capable of detecting the anti-TNF mAb. In mice administered the anti-TNF 286 mAb via an intraperitoneal route, we detected mAb in the serum of 50% of the mice (Fig. S9a). 287 By contrast, levels of serum anti-TNF Nb were below the level of detection in all mice orally 288 inoculated with TNF-PROT3EcT, and only detectable in 20% of mice treated with TNF-289 PROT3EcT or TNF-T3EcT via enema. We did not detect evidence of anti-TNF Nbs in the colonic 290 contents or homogenates (Fig. S9b-c). 291

A single dose of TNF-PROT3EcT is associated with TNFα suppression and inhibition of 292 intestinal inflammation. 293
Given that the TNF-PROT3EcT-4 treated mice exhibited minimal weight-loss post-294 administration of TNBS, we tested whether pretreatment with a single dose is therapeutically 295 efficacious. Two days post-TNBS administration, mice pre-treated with a single oral dose of 296 TNF-PROT3EcT-4 ( Fig. 6i) exhibited minimal evidence of weight loss, colon shortening, and 297 colitis ( Fig. 6k-m). PROT3EcT-4 or the vehicle diluent (PBS) had no effect. As before, similar 298 levels of both strains were shed in the feces (Fig. 6j) and all colonies of shed TNF-PROT3EcT-4 299 retained the ability to secrete anti-TNF Nb (Fig S7f). For these experiments, mice were 300 sacrificed two days post-administration of TNBS, a time point at which we reproducibly detected 301 elevated proinflammatory cytokine levels within colonic tissue in controls. Significantly lower 302 levels of TNFα and IL-6 were detectable within the colonic tissue of mice pretreated with TNF-303 PROT3EcT-4 ( Fig. 6n-o), suggesting that the secreted anti-TNFα Nb sequesters its target and 304 reduces IL-6. While others have observed that TNF neutralization or EcN treatment can 305 increase IL-10 production in the gut, we observed equivalent levels of IL-10, regardless of the 306 intervention ( Fig. 6p). 307

Discussion 308
Here we describe the development of PROT3EcT, E. coli engineered for the in-situ delivery of 309 high specificity protein payloads to sites of disease. Using synthetic biology-based approaches 310 we have engineered both laboratory and non-pathogenic human E. coli isolates with a T3SA 311 modified to secrete proteins in a regulated or constitutive manner. PROT3EcT engineered with a 312 constitutively active secretion system that is maintained in the absence of antibiotic selection 313 exhibited growth in vitro at rates equivalent to unmodified EcN and can colonize the intestines of 314 mice for at least 14 days. 315 We demonstrate the ability of TNF-PROT3EcT, a variant that constitutively secretes anti-316 TNF Nbs, to suppress the development of inflammation in a preclinical mouse model of IBD. 317 Orally or rectally administered TNF-PROT3EcT was as efficacious as systemically administered 318 anti-TNF mAb in limiting the development of TNBS-induced colitis. Other groups have also 319 engineered microbes to treat gut inflammation, the most closely related being variants of 320 Lactococcus lactis that secrete IL-10, an anti-inflammatory cytokine, or an anti-TNF Nb 42, 43 . A 321 native secretion system of this Gram-positive bacterium was repurposed for the secretion of 322 therapeutic payloads. However, the strain of L. lactis used does not colonize the intestines of 323 humans or mice 44, 45 and may vary in its metabolic activity within the mammalian intestine 46 , 324 likely accounting for why it only moderately suppressed inflammation when administered on a 325 daily basis. By contrast, we observe that pre-treatment with just a single oral dose of TNF-326 PROT3EcT significantly ameliorates colonic inflammation and injury. 327 The modular design of PROT3EcT is such that it can be adapted to secrete different 328 payloads as well as to respond to environmental cues. For example, in future studies, by 329 altering the conditions that induce expression of VirB, PROT3EcT's T3SA could be endowed 330 with an 'on switch' triggered by specific signals of the gut's inflammatory milieu, e.g., reactive 331 nitrogen species 36,[47][48][49] . In terms of therapeutic payloads, we demonstrate the versatility of 332 PROT3EcT to secrete different Nbs, including Nbs that inhibit the activity of bacterial toxins 333 (Nb Stx2 ) or immune checkpoint molecules (Nb PD-L1 and Nb CTLA-4 ). While the Nbs we studied were 334 each derived from immunized alpacas, synthetic yeast-and bacterial-based Nb libraries are 335 also available that can be screened rapidly for Nbs with desired properties 50, 51 . Furthermore, 336 PROT3EcT is not limited to the secretion of Nbs, as we and others have previously established 337 that a variety of other heterologous proteins can be recognized as type III secreted substrates 52 . 338 By altering its route of administration, PROT3EcT can be expanded for the deposition of 339 therapeutics not only to the gastrointestinal tract, but also to solid tumors, as EcN home to and 340 colonize a variety of solid tumors when administered intravenously, at least in mice 53 . 341 Given its inherent anti-inflammatory and anti-microbial properties, EcN is GRAS and has 342 been used for over a century to treat various intestinal diseases and is available over-the-343 counter in some countries. However, EcN contains an operon that mediates the synthesis of a 344 colibactin, a genotoxin capable of mediating the formation of DNA crosslinks 54 . Other colibactin-345 producing E. coli promote the development of colorectal cancer (CRC) in mouse models 55 and 346 induce mutational signatures found in human CRC 56 . Whether this will turn out to be an issue 347 that limits the use of EcN-based therapeutics in humans remains to be discovered. However, 348 EcN mutants deficient in colibactin biosynthesis are not impaired in their ability to colonize the 349 intestines of at least mice 57, 58 . In future studies, we intend to test the ability of colibactin-350  PROT3EcT delivering its therapeutic payload directly to the gut lumen. 508 plasmids are referred to as pDSW206-ccdB-Nb* (* = name of Nb present in construct).