Engineered Escherichia coli for the in situ secretion of therapeutic nanobodies in the gut

Summary

Drug platforms that enable the directed delivery of therapeutics to sites of disease to maximize efficacy and limit off-target effects are needed. Here, we report the development of PROT₃EcT, a suite of commensal Escherichia coli engineered to secrete proteins directly into their surroundings. These bacteria consist of three modular components: a modified bacterial protein secretion system, the associated regulatable transcriptional activator, and a secreted therapeutic payload. PROT₃EcT secrete functional single-domain antibodies, nanobodies (Nbs), and stably colonize and maintain an active secretion system within the intestines of mice. Furthermore, a single prophylactic dose of a variant of PROT₃EcT that secretes a tumor necrosis factor alpha (TNFα) neutralizing Nb is sufficient to ablate proinflammatory TNF levels and prevent the development of injury and inflammation in a chemically-induced model of colitis. This work lays the foundation for developing PROT₃EcT as a platform for potential treatment of gastrointestinal-based diseases.

Graphical Abstract

eTOC

Lynch et al. describe the development of PROT₃EcT, a suite of E. coli outfitted with a modified bacterial secretion system that enables the secretion of therapeutic payloads into their surroundings. Pretreating with a TNFα-neutralizing nanobody-secreting PROT3EcT prevents colitis in a preclinical IBD model, providing proof-of-concept for further platform development.

Introduction

Microbe-based therapeutics are emerging as a platform for the treatment of a variety of diseases, particularly those with etiologies linked to the gut^1,2. Efforts are underway to identify cocktails of beneficial natural isolates, while synthetic biology-based approaches are ongoing to engineer microbes with additional therapeutic capabilities, including the targeted deposition of therapeutic payloads to sites of disease^3–5. Due to their ease of production, administration, and natural capacity to synthesize and deliver complex biologics, engineered microorganisms hold enormous potential as affordable alternatives to traditional biologic therapies^3–6. Microbes outfitted to deliver high-specificity payloads to sites of disease provide a platform for developing interventions with improved efficacy and limited off-target effects.

Escherichia coli Nissle 1917 (EcN), a probiotic with GRAS (generally recognized as safe) status⁷, is gaining traction as a chassis for synthetic biology⁶. EcN has inherent antibacterial and anti-inflammatory activities and is genetically tractable and has been used for over a century for the treatment of a variety of intestinal diseases, including inflammatory bowel disease⁶. A variety of strategies are being pursued to enhance EcN’s therapeutic potential. For example, variants with enhanced metabolic capabilities are being investigated for the removal of toxic intermediates associated with metabolic diseases^8–10.

Efforts are also underway to engineer EcN and other commensal bacteria to deliver therapeutic payloads to sites of disease. In the case of Gram-positive bacteria, the native general secretion (Sec) and twin-arginine translocation (Tat) pathways have been repurposed to secrete therapeutic agents into their surroundings. For example, Lactococcus that secrete IL-10 or anti-TNF single-domain antibodies (also known as nanobodies or VHH) has been demonstrated to ameliorate gut inflammation in animal models^11,12. However, for Gram-negative bacteria, the engineering of these conserved secretion systems is limited as they primarily function to deliver proteins into the periplasmic component of their outer envelope (for review, see¹³). Given the difficulty of engineering Gram-negative bacteria to secrete proteins into their surroundings, work has primarily focused on developing E. coli variants programmed to lyse and release intracellular cargo^14–16 or to display proteins of therapeutic potential on their outer surface via fusion to outer membrane adhesins¹⁷ or Curli^18,19.

Numerous Gram-negative bacterial pathogens utilize complex nanomachines, including type III secretions systems (T3SSs), to transport bacterial proteins referred to as effectors directly into the cytosol of host cells. The fully assembled type III secretion machines (apparatuses) (T3SAs) are embedded within the outer envelope of the bacterium with a needle-like extension that docks onto and forms pores in host cell membranes. We and others have previously established that the T3SAs of genetically related Gram-negative pathogens, including Shigella flexneri and enteropathogenic E. coli, are functional when introduced into laboratory strains of E. coli^20–24.

Here, we report the development of the PROT₃EcT (PRObiotic Type 3 secretion E. coli Therapeutic) platform, a suite of E. coli engineered to express a Shigella T3SA modified to secrete proteins into its surroundings rather than directly into eukaryotic cells. PROT₃EcT can secrete a variety of heterologous proteins which have been outfitted with an N-terminal type III secretion sequence. PROT₃EcT are modular in design and composed of three genetic elements: (1) the operons encoding the modified T3SA, (2) their master transcriptional regulator (VirB), and (3) a therapeutic payload. PROT₃EcT-4, a variant engineered to maintain all elements in the absence of antibiotic selection, is unimpaired for growth in vitro or in the intestines of mice. As proof-of-concept of the therapeutic potential of PROT₃EcT, we provide evidence that TNF-PROT₃EcT, a variant that secretes an anti-TNFα Nb, is as effective as systemically administered anti-TNFα monoclonal antibody in suppressing the development of inflammation in a chemically induced preclinical model of inflammatory bowel disease. Together, these studies provide the foundation for the continued development of PROT₃EcT as a versatile therapeutic platform.

Results

Development of PROT₃EcT

The genes that encode the ~20 components that form the Shigella T3SA are contained within the ipa, mxi, and spa operons located adjacent to each other on a large virulence plasmid^25,26. The mxi and spa operons encode all the structural components needed to form the T3SA. The ipa operon includes those that form the tip complex that holds the machine in an OFF configuration and, upon contact, forms a pore complex in the host cell membrane, which enables the injection of proteins directly into the cytosol of targeted mammalian cells^27–29. We previously established a recombineering-based platform to transfer large regions of this virulence plasmid into engineered synthetic loci on the E. coli chromosome^20–22,30. Using this platform, we developed laboratory strains of E. coli that encode the ipa, mxi, and spa operons, which we established can inject heterologous proteins into mammalian cells²³.

With the goal of developing E. coli that secrete proteins into their surroundings, we compared the secretory activity of DH10b E. coli that contain the ipa, mxi, and spa operons vs. the mxi, and spa operons. In each case, the operons were inserted at the same chromosomal locus and a low-copy number plasmid that expresses VirB, the transcription factor that controls the expression of all three operons (Figure S1A), under the control of the IPTG (isopropyl β-d-1-thiogalactopyranoside)-inducible Ptrc promoter, was introduced. The resulting strains are referred to here as mT3Ec_Ipa-Mxi-Spa (Figure S1A) and mT3Ec_Mxi-Spa (Figure S1B).

When grown under conditions that promote expression of the operons and exposed to Congo red, a dye that triggers secretion in the absence of host cells³¹, mT3Ec_Ipa-Mxi-Spa and mT3Ec_Mxi-Spa, secreted similar levels of Ptrc-regulated OspC2 (a native Shigella type III secreted effector), demonstrating that the ipa operon plays no role in secretory activity (Figure 1A). GroEL, a highly abundant cytosolic protein, and reporter of host cell lysis, was not detected in the supernatant fractions (Figure 1A), providing evidence that proteins were secreted via the T3SA. In the absence of Congo red, OspC2 was abundantly secreted by mT3Ec_Mxi-Spa, but not mT3Ec_Ipa-Mxi-Spa, demonstrating that mT3Ec_Mxi-Spa constitutively secretes proteins into its surroundings without the need of host cell contact (Figure 1B). When we examined all proteins present in the supernatant of mT3Ec_Mxi-Spa, OspC2 was the most abundant protein, thus demonstrating that the introduction of the mxi-spa operons and VirB was sufficient to outfit DH10b E. coli with a robust Ptrc-regulated secretion system (Figure 1C).

Figure 1. E. coli engineered with a modified T3SA efficiently secrete proteins into their surroundings.

(A, B, C, E, F) Secretion assays of designated strains engineered to secrete OspC2-FLAG. Supernatant (TCA precipitated) (S) and whole-cell pellet lysate (P) fractions were obtained 30-minutes (A, B, E), 6-hours (C), or at times indicated (F) post-transfer to new media. Unless otherwise indicated in (A), IPTG was present to induce expression of VirB and OspC2. Congo red (CR) was added, when noted. Immunoblots labeled with anti-FLAG or anti-GroEL antibodies (A, B, E, F) or a Coomassie blue-stained gel (C) are shown. (D) Schematic of PROT₃EcT-1 that expresses plasmid-encoded Ptrc VirB and Ptrc OspC2-FLAG. (G) Gentamicin protection assay to assess invasion of bacteria into intestinal epithelial cells (HCT8). Each dot represents a biological replicate, and the horizontal bar indicates the mean. Data were analyzed using one-way ANOVA with Tukey’s post-hoc test; for all strain comparisons to Shigella *** P < 0.0001. (H) Translocation assay to assess the injection of OspC2-FLAG into cervical epithelial cells (HeLa). The soluble and insoluble (bacteria-containing) fractions were separated and immunoblotted with anti-FLAG and anti-ß-actin. Data in each panel are representative of at least 2 independent experiments. See also Figure S1.

We next investigated whether similar modifications to two non-pathogenic human E. coli isolates, E. coli Nissle 1917 (EcN) and E. coli HS, would equip them with a functional protein secretion system. First, we developed PROT₃EcT-1, EcN engineered with the mxi-spa operons at the analogous chromosomal locus as mT3Ec_Mxi-Spa. PROT₃EcT-1 (Figure 1D) secreted OspC2, albeit at somewhat lower levels than mT3Ec_Mxi-Spa (Figure 1E). In each case, secretion was VirB-dependent (Figures 1A and 1E), indicating that the secretory activity of PROT₃EcT-1 is dependent on the expression and assembly of a functional T3SA. Similar modifications to the E. coli HS led to the generation of PROT₃EcT-2, which secreted OspC2 at levels equivalent to that of mT3Ec_Mxi-Spa (Figure S1C).

While our initial studies examined the amount of protein secreted over 30 minutes, we found that when grown in media that support their growth, over a 6-hour period, mT3Ec_Mxi-Spa and PROT₃EcT-1 accumulated OspC2 in their secreted fractions (Figure 1F), demonstrating that, when induced, they continually secrete protein into their surroundings.

Shigella are intracellular pathogens that rely on their T3SS and its secreted proteins to invade non-phagocytic epithelial cells. Given that bacteria engineered with the mxi-spa operons lack effectors and components of the tip complex, they are not expected to invade host cells or inject proteins into host cells. To confirm, first, we compared the ability of Shigella, PROT₃EcT-1, mT3Ec_Mx-Spa, DH10b E. coli, and EcN to invade HCT8 cells, a mammalian cell line of intestinal origin. As expected, Shigella but none of the other strains were protected from gentamicin, an antibiotic that kills extracellular but not intracellular bacteria (Figure 1G). Second, we compared the levels of OspC2 in the soluble cytosolic and insoluble bacterial containing fractions of mammalian HeLa cells exposed to WT Shigella, mT3Ec_Mx-Spa and PROT₃EcT-1 (Figure 1H). As expected, OspC2 was only observed in fractions of cells infected with WT Shigella. Together these studies demonstrate that mT3Ec_Mx-Spa and PROT₃EcT-1 do not invade or inject proteins into mammalian cells.

PROT₃EcT can be engineered to secrete heterologous proteins

We next investigated whether PROT₃EcT-1 recognizes heterologous proteins as secreted substrates. In prior studies, we had found that the fusion of the first 50 residues of multiple type III effectors fused to the N-termini of heterologous proteins was sufficient to generate secreted variants²³. These residues encode the elements that define type III secretion sequences (SSs)³². As previously observed with mT3Ec_Ipa-Mxi-Spa, we found that PROT₃EcT-1 recognized similarly modified heterologous proteins, including MyoD and Mef2c, two mammalian transcription factors, TALE, a transcription activator-like effector, and 60 and 70 kDa fragments of APC, a mammalian tumor suppressor protein (Figure S1D).

Nbs, the ~15kDa variable domains of heavy chain-only antibodies, are ideal substrates for our bacterial secretion system as they are small stable proteins that generally exhibit strong antigen-binding affinity. Monomeric Nbs were previously engineered to be recognized as secreted substrates of the T3SS of enteropathogenic E. coli³³. Thus, we screened for modifications to a representative monomeric Nb that would enable its secretion by PROT₃EcT. Nb^{ASC 34} was fused to the secretion sequences of nine Shigella effectors (OspC2, IpaH4.5, IpaH7.8, IpaH9.8, OspE, OspF, OspD3, VirA, and OspG)^21,23. Variants fused to the OspC2 and OspG secretion sequences exhibited the highest level of secretion (Figure 2A). Fusion of the OspC2 sequence also enabled the secretion of monomeric Nb^{PD-L1 35}, Nb^{CTLA4 36}, Nb^{NPI 37}, and Nb^{Stx2 38} (Figure 2B) and single-chain heterodimeric and heterotrimeric Nb^Stx2 (Figure 2C). As observed with full-length OspC2 (Figure 1C), heterodimeric SS^OspC2-Nb^Stx2 was the most abundant protein present in the supernatants of PROT₃EcT-1 (Figure 2D) and exhibited binding and neutralizing activity against Shiga toxin, its target antigen (Figures 2E and 2F). This was an important finding as it demonstrated that the Nb, which like other type III secreted substrates is unfolded when loaded into the secretion apparatus, once secreted correctly refolds.

Figure 2. PROT₃EcT can be engineered to secrete nanobodies.

Secretion assays of (A) Nb^ASC fused to designated N-terminal type III secretion signals and a C-terminal HA-tag, (B) Nb^ASC, -Nb^PD-L1, -Nb^CTLA-4 and -Nb^NP1 fused to the N-terminal OspC2 secretion signal (SS^OspC2) and C-terminal HA-tag, (C) monomeric (1x), heterodimeric (2x) and heterotrimeric (3x) Nb^Stx2, fused to an N-terminal SS^OspC2and a C-terminal 3xFLAG tag, (D) Nb^Stx2 dimer modified with an N-terminal SS^OspC2 (SS^OspC2-Nb^2x) and C-terminal HA tag. (A, B, C) were obtained at 6-hours. Immunoblots labeled with anti-GroEL as well as anti-HA (A, B) or anti-FLAG (C) antibodies or (D) a Coomassie blue-stained gel are shown. (E) Stx2 ELISA. (F) Stx2 Vero killing assay. For (E-F), each dot represents an individual biological replicate. Data in each panel are representative of at least 2 independent experiments. See also Figure S1.

Development of constitutively active Nb-secreting PROT₃EcT

Before studying the behavior of PROT₃EcT in mice, we sought to generate a constitutively active variant that maintains all its genetically engineered components in the absence of antibiotic selection. First, to enable constitutive expression of the secretion system, we replaced the inducible Ptrc promoter of VirB, its master transcriptional regulator, with several predicted to be constitutively active in the gut, e.g., Shigella PvirF³⁹, E. coli PompC⁴⁰, and two synthetic promoters, PJ23115 and PJ23119⁴¹. PROT₃EcT-1 variants that carried plasmids that express VirB under the control of each of these promoters secreted equivalent levels of Nbs (Figure S1E). Thus, we introduced the PJ23119-virB expression cassette into the chromosome of PROT₃EcT-1 to generate PROT₃EcT-3 (Figures S1F–G).

Next, to enable constitutive Nb expression, we replaced its inducible Ptrc promoter with the constitutive PJ23108 promoter (Figure S1H). In addition, to maintain flexibility in the introduction of expression circuits and to enable higher levels of payload expression, we developed a means to maintain plasmids in PROT₃EcT in the absence of antibiotic pressure. Based upon the work from Hwang and colleagues⁴², we developed PROT₃EcT-4, a derivative that lacks alr and dadX (Figure 3A). These genes encode EcN’s two alanine racemases that convert L- to D-alanine, an amino acid essential for cell wall biosynthesis that is very limited in the mammalian gastrointestinal tract⁴³. In parallel, we inserted alr onto the Nb-producing plasmid. PROT₃EcT-3 and PROT₃EcT-4 that carry this plasmid secreted similar levels of Nbs. PROT₃EcT-3 but not PROT₃EcT-4 lost its ability to express and secrete Nbs in the absence of antibiotic selection (Figures S1I–J). Remarkably, despite expressing a constitutively active secretion system, Nb^Stx2-secreting PROT₃EcT-4 (Stx2-PROT₃EcT) exhibited essentially identical growth patterns when grown in parallel (Figure 3B) or in competition (Figure S1K) with EcN, indicating that these modifications do not add a significant metabolic burden.

Figure 3. PROT₃EcT stably colonizes the gastrointestinal tract of mice but does not affect the gut microbiota.

(A) Schematic of PROT₃EcT-4, which expresses VirB via a constitutive promoter (Pc) from a chromosomal locus. The Nb is expressed via a constitutive promoter from a plasmid maintained via auxotrophic selection. (B) Growth curves of wild-type EcN and Stx2-PROT₃EcT grown in parallel. Data are presented as the mean ± SEM. (C) Study design schematic. (D) Shed bacterial titers. Data are presented as the mean ± SEM, each with 4 mice per group, and reflect at least 2 independent experiments. (E) Principal Coordinate Analysis (PCoA) plot showing Bray-Curtis dissimilarity for 16S rRNA gene amplicon fecal surveys from sham-treated (PBS) (n=12) and PROT₃EcT-treated (n=12) mice, collected on day 14 post-treatment. (F) A hierarchically clustered heatmap of the z-score abundance of bacterial family/genus (f = family, g = genus) within each sample (rows) from the same samples as in (E). Color codes indicate treatment conditions (PBS vs. PROT₃EcT). (G-H) Bioluminescent imaging of intestinal explants isolated from individual mice 8 dpi inoculated orally with designated strains transformed with (D) a constitutive bioluminescent reporter (pMM543) or (E) a T3SS-dependent reporter system (pMxiE-lux+ and pNG162-IpgC). Starting 1 day before they were administered bacteria to ensure plasmid maintenance, the mice began receiving kanamycin and spectinomycin in their drinking water. See also Figures S1–S2 and Tables S1A–B.

PROT₃EcT colonizes and maintains an active T3SA but does not perturb the gut microbiota of mice

Our initial in vivo experiments focused on investigating the ability of Nb-secreting PROT₃EcT-4 to colonize the mouse gastrointestinal tract. We monitored the titers of bacteria shed in the feces of C57BL/6 mice orally inoculated with EcN or Stx2-PROT₃EcT (Figure 3C). After administering a single dose of ~10⁸ colony forming units (CFU), we detected ~10⁵ CFU/g/day of each strain for at least 14 days (Figures 3D and S2A). The mice appeared well and demonstrated no evidence of significant weight loss throughout this time (Figure S2B).

To investigate whether the bacteria isolated from feces maintained secretory activity, we monitored the secretion of 39-40 isolates from a total of 4 mice on days 2, 5, and 14. Each of the 158 Stx2-PROT₃EcT colonies examined secreted Nbs at levels equivalent to the inoculation strain (Figures S2C). These findings demonstrate that within the gastrointestinal tract of mice, in the absence of antibiotic selection, PROT₃EcT-4 maintains a functional secretion system and the alr-Nb-expressing plasmid.

We also compared the fecal microbiome composition of C57BL/6 mice 14 days post-inoculation with PBS or ~10⁸CFU of PROT₃EcT-4 by conducting 16S rRNA gene amplicon surveys. At that time, ~10⁵CFU/gm/day of PROT₃EcT-4 were found in the feces of mice that received bacteria (Figure S2D). We observed no community-level changes in the gut microbiota composition of PROT₃EcT-4 and sham-treated mice as assessed by Bray-Curtis dissimilarity of the reads (Figure 3E, Table S1A). Similarly, hierarchical clustering at the genus-level (Figure 3F) and differential abundance testing with MaAsLin2 (Table S1B) showed no significant taxonomic changes of the gut microbiota of mice that received PROT₃EcT-4 versus PBS.

We examined the biogeography of PROT₃EcT-3 and EcN within the intestines of mice inoculated with variants that constitutively express the luciferase-producing luxCDABE operon⁴⁴, which exhibited equivalent luciferase activity when grown in vitro (Figures S2E–F). Eight days post-oral inoculation of the mice, the two strains showed similar patterns of luciferase expression in explanted sections of their ileum, cecum, and proximal colon, which had feces removed before imagining (Figure 3G).

To confirm that the modified T3SA present in PROT₃EcT is actively transcribed within the intestines of mice, we developed a luxCDABE-based reporter that is only activated when the mxi-spa operons, which encode the modified T3SA, are transcribed (Figure S2G). Variants of PROT₃EcT-3, but not EcN, that carry this reporter demonstrated luciferase production when grown in culture (Figures S2H–I) and within the explanted cecum, proximal colon, and ileum of inoculated mice (Figure 3H), demonstrating that the secretion system is expressed within the intestines of mice and does not perturb EcN colonization. For all the luciferase-based studies, we treated the mice with antibiotics to ensure maintenance of plasmids that encode the luxCDABE operon.

PROT₃EcT can be engineered to secrete functional SS^OspC2-Nb^TNF

To establish the use of PROT₃EcT as a therapeutic platform, we focused efforts on investigating its efficacy in the treatment of inflammatory bowel diseases (IBD). The etiologies of ulcerative colitis and Crohn’s disease, collectively termed IBD, are complex and thought to be driven by host genetic, environmental, and microbiota factors. Yet, both diseases exhibit chronic inflammation accompanied by increased levels of pro-inflammatory cytokines⁴⁵. Monoclonal antibodies (mAb) that target the pro-inflammatory cytokine TNFα, e.g., infliximab and adalimumab, are highly efficacious in controlling severe disease and in improving the quality of life of patients with IBD⁴⁶. However, given the systemic administration of these therapeutics, patients receiving these agents are immunosuppressed and at increased risk of developing life-threatening infections and lymphoma⁴⁷.

We hypothesized that the targeted delivery of anti-TNFα Nbs via PROT₃EcT to the intestines would reduce intestinal inflammation. To investigate this possibility, we isolated Nbs from alpacas immunized with recombinant mouse (m)TNFα, including one (JTT-B10) that bound with high affinity (EC₅₀ 0.1 nM) and neutralized mTNFα (IC₅₀ 0.1 nM) (Table S2 and Figure S3A–C). We generated monomeric and homodimeric variants of this Nb (Nb^mTNF) engineered with an SS^OspC2. The homodimer (SS^OspC2-Nb^2xmTNF) was secreted much more efficiently than the monomer (Figure 4A) and was detected at ~0.5 μg/mL in the supernatant of an 18-hour culture (Figure S3D). PROT₃EcT-secreted SS^OspC2-Nb^2xmTNF was as effective as E. coli-purified homodimeric Nb in blocking mTNFα-induced death of mouse L929 cells (Figure 4B). Using a similar approach, we also isolated an alpaca-derived human-specific anti-TNFα Nb that binds with high affinity (0.5 nM) (Figure S3E) and blocks human TNFα-induced death of mouse L929 cells (Figure S3F). When engineered with an N-terminal SS^OspC2, functional homodimers of this Nb (SS^OspC2-Nb^2xhTNF) are secreted by PROT₃EcT (Figures S3G–H).

Figure 4. TNF-PROT₃EcT secretes fully functional anti-TNF Nbs and colonizes mice but does not affect the gut microbiota.

(A) Secretion assays of PROT₃EcT-1 engineered with monomeric (1x) or homodimeric (2x) Nb^TNF fused to an N-terminal OspC2ss and a C-terminal FLAG. Supernatant (TCA precipitated) (S) and whole-cell pellet lysate (P) fractions were obtained at 6-hours. Immunoblots labeled with an anti-FLAG antibody are shown. (B) Viability of L929 cells following incubation with 0.2 ng/ml of murine TNFα plus either supernatant (Sup) from designated strains or purified homodimeric Nb^TNF. Data are presented as the mean ± SEM (C) Shed bacterial titers. Each dot represents an individual mouse, and the horizontal bar indicates the mean. (D) PCoA plot showing Bray-Curtis dissimilarity for 16S rRNA gene amplicon fecal surveys from PROT₃EcT (n=7) and TNF-PROT₃EcT-treated (n=8) mice. (E) A hierarchically clustered heatmap of the z-score abundance of bacterial family/genus (f = family, g = genus) within each sample (rows) from the same samples as in (D). Color codes indicate treatment group (PROT₃EcT vs. TNF-PROT₃EcT) and caging information. Data in each panel are representative of at least 2 independent experiments. See also Figures S3–S5 and Tables S2 and S3A–B.

TNF-PROT₃EcT inhibits the development of disease in a mouse model of colitis

To investigate the utility of PROT₃EcT as a live biotherapeutic for the treatment of intestinal inflammation, we chose a preclinical chemically induced mouse colitis model whereby a mixture of TNBS (2,4,6-trinitrobenzene sulfonic acid), a hapten, and ethanol, which disrupts the mucosal barrier, is rectally instilled into the colon. TNBS bound to colonic tissue proteins induces inflammation driven by pro-inflammatory cytokines, including TNFα⁴⁸. For these studies, we used BALB/c mice, which are more susceptible to TNBS than C57BL/6 mice^48,49. As previously reported⁵⁰, mice treated intraperitoneally with a neutralizing anti-TNF monoclonal antibody (1 day prior and 2- and 4-days post-administration of TNBS) were protected from weight loss, colon shortening, and histologic evidence of colitis (Figure S4A–G).

Before testing the efficacy of Nb^2xmTNF-secreting PROT₃EcT-4 (TNF-PROT₃EcT) in colitis suppression, we compared the composition of the fecal microbiota of BALB/c mice colonized with TNF-PROT₃EcT or PROT₃EcT-4. We carried out 16S rRNA gene amplicon surveys on the feces from mice 14-days post-inoculation with TNF-PROT₃EcT and PROT₃EcT-4, which like C57BL/6 mice continued to shed ~10⁵ CFU/gm/day (Figure 4C). We observed no separation in the gut microbiota at the community level between TNF-PROT₃EcT and PROT₃EcT-4 treated mice, as assessed via Bray-Curtis dissimilarity of the reads (Figure 4D, Table S3A). Consistent with this ordination analysis, hierarchical clustering at the genus level (Figure 4E) and differential abundance testing using MaAsLin2 (Table S3B) showed no significant taxonomic changes that correlate with TNF-PROT₃EcT versus PROT₃EcT-4 administration. Thus, Nb^2xmTNF secretion by PROT₃EcT does not significantly affect gut microbiota community composition as assessed.

Next, to test the therapeutic efficacy of TNF-PROT₃EcT, one day before as well as two and four days after they received TNBS, we orally gavaged mice with 10⁸ CFU of TNF-PROT₃EcT, PROT₃EcT-4 or PBS (Figure 5A). All animals that received bacteria consistently shed ~10⁵ CFU/g in their feces (Figures 5B and S4H). Treatment with TNF-PROT₃EcT significantly reduced weight loss, blunted colon shortening, and decreased or completely abrogated epithelial injury and inflammation throughout the length of the colonic mucosa (i.e., the distal and proximal colon), including less polymorphonuclear and mononuclear cell infiltration (Figures 5C–F). In contrast, PROT₃EcT-4 did not provide any protection as assessed by each of these metrics, demonstrating that the therapeutic efficacy afforded by TNF-PROT₃EcT is due to the secreted SS^OspC2-Nb^2xmTNF and not EcN intrinsic.

Figure 5. TNF-PROT₃EcT inhibits the development of TNBS-induced colitis.

(A, G) Study design schematic. (B) Shed bacterial titers. (C, H) Body weight change (%). In (C), *denotes comparison to PBS group, P=0.0118; ^# denotes comparison to PROT₃EcT-4; day 1, P=0.0238; day 2, P=0.0122. In (H), # denotes comparison to PROT₃EcT-4, day 1, P=0.0003, day 2, P=0.0018, day 3, P=0.02. (D, I) Colon length. In (D), *, P=0.0219; ***, P=0.0004. In (I), *, P=0.0123. (E, J) Histologic colitis scores. In (E), vs. PBS *, P=0.0231; vs. PROT₃EcT-4 *, P=0.0141. In (J), vs. PBS, P= 0.0406 and vs. PROT₃EcT-4, P=0.0331. (F) Representative colon histology sections stained with hematoxylin and eosin. Scale bars = 100 μm. Data reflect at least 2 independent experiments, each with 3-5 mice per group, and are presented as mean ± SEM (B-C, H) or individual values ± SEM (D-E, I-J). Data were analyzed with a two-way ANOVA with Tukey’s post hoc test (B-C, H) or a Kruskal-Wallis test with Dunn’s multiple correction test (D-E, I-J). TNBS = 2,4,6-Trinitrobenzenesulfonic acid. EtOH = ethanol. See also Figures S4 and S6–7.

Given that enemas are commonly used for drug delivery for patients with IBD, we also investigated the efficacy of intrarectally delivered TNF-PROT₃EcT in limiting TNBS-induced colitis using the same dosing strategy as described above (Figure 5G). As with oral delivery, intrarectally delivered TNF-PROT₃EcT, but not PROT₃EcT-4, ameliorated weight loss, colon shortening, and colitis (Figures 5H–J). Both strains were shed at similar levels in feces (Figure S4I).

To evaluate the effectiveness of TNF-PROT₃EcT in another colitis model, we chose dextran sodium sulfate (DSS)-induced colitis in which DSS administration disrupts the colonic epithelium leading to mucosal injury and inflammation⁵¹. However, we observed that neither the administration of anti-TNF mAb nor TNF-PROT₃EcT via an oral or intrarectal route offered protection against the development of colitis as assessed by weight loss, colon shortening and histological examination of the colon (Figure S5A–N). Thus, further substantiating the specificity of TNF-PROT₃EcT in blocking TNFα-driven inflammation.

SS^OspC2-Nb^2xmTNF secretion but not E. coli colonization is required for protection against TNBS-induced colitis

Therapeutic strains that cannot compete for and establish a replicative niche (i.e., non-colonizing) within the colon may be useful if administered repeatedly to patients. To assess whether treatment with similarly engineered DH10b E. coli also suppresses colonic inflammation, we developed T₃EcT, a variant of mT3Ec-Mxi-Spa engineered with the chromosomally encoded PJ23119 VirB gene cassette, and a variant of T₃EcT that secretes SS^OspC2-Nb^2xmTNF, TNF-T₃EcT. After establishing that TNF-T₃EcT constitutively secreted SS^OspC2-Nb^2xmTNF (Figure S6A), TNF-T₃EcT and T₃EcT were administered orally or intrarectally to mice using the strategy outlined above (Figures S6B–C). The levels of each strain shed in stool remained low (between zero and 10³ CFU/g) and, unlike PROT₃EcT, T₃EcT and TNF-T₃EcT were cleared from >50% of the mice (Figures S6D–E). However, while orally administered TNF-T₃EcT provided no protection (Figures S6F), rectally delivered TNF-T₃EcT, but not T₃EcT, suppressed colitis (Figures S6G), likely reflecting repeated transient delivery of SS^OspC2-Nb^2xmTNF into the colon.

To address whether bacterial secreted Nbs were restricted to the gut, we measured Nb levels in the colonic contents, colon tissue homogenates, and serum of mice treated with each strain across all the TNBS experiments. To measure SS^OspC2-Nb^2xmTNF, we used a direct ELISA, which can also detect the anti-TNF mAb. In mice administered the anti-TNF mAb via an intraperitoneal route, we detected mAb in the serum of 50% of the mice (Figure S7A). By contrast, levels of serum SS^OspC2-Nb^2xmTNF were below the level of detection in all mice orally inoculated with TNF-PROT₃EcT and only detectable in 20% of mice treated with TNF-PROT₃EcT or TNF-T₃EcT via enema. We did not detect evidence of SS^OspC2-Nb^2xmTNF in colonic contents or homogenates (Figures S7B–C), likely due to their degradation via gut proteases.

A single oral dose of TNF-PROT₃EcT is associated with TNFα suppression and inhibition of intestinal inflammation

Given that the TNF-PROT₃EcT-treated mice exhibited minimal weight-loss post-administration of TNBS, we tested whether pretreatment with a single dose was therapeutically efficacious. Two days post-TNBS administration, mice pre-treated with a single oral dose of TNF-PROT₃EcT or PROT3EcT (Figure 6A) shed similar levels of each strain in their feces (Figures 6B, S6H), and when tested, colonies of shed TNF-PROT₃EcT retained the ability to secrete anti-TNF Nb (Figure S6I). Strikingly, mice treated TNF-PROT₃EcT exhibited minimal evidence of weight loss, colon shortening, and colitis, whereas PROT₃EcT-4 and the vehicle diluent (PBS) had no effect (Figures 6C–E). Similar protection was observed five days post-TNBS administration when mice were pre-treated with a single oral dose of TNF-PROT₃EcT (Figures S6J–M).

Figure 6. A single prophylactic dose of TNF-PROT₃EcT attenuates TNBS-induced colitis.

(A) Study design schematic. (B) Shed bacteria titers. (C) Body weight change (%). *denotes comparison to PBS, day 1, P=0.0002, day 2, P<0.0001; ^#denotes comparison to PROT₃EcT-4, day 1, P=0.054, day 2, P<0.0001. (D) Colon length. *, P=0.0184; **, P=0.0029. (E) Histologic colitis scores. **, P=0.0045. (F-H) Colon homogenates were analyzed for the levels of TNFα (F) (***, P=0.0005, *, P=0.0433), IL-6 (G) (vs. PROT₃EcT-4 *, P=0.0356; vs. PBS *, P=0.0322) and IL-10 (H) by ELISA. Data reflect 2 independent experiments, each with 4-5 mice per group, and are presented as individual values ± SEM (D-G) or mean ± SEM (B-C). Data were analyzed using a two-way ANOVA with Tukey’s post hoc test (B-C) or a Kruskal-Wallis test with Dunn’s multiple correction test (D-H). TNBS = 2,4,6-Trinitrobenzenesulfonic acid. EtOH = ethanol. See also Figure S6.

For the mice that were sacrificed two days post-administration of TNBS, a time point at which we reproducibly detected elevated pro-inflammatory cytokine levels within colonic tissue in controls, significantly lower levels of TNFα and IL-6 were detectable within the colonic tissue of mice pretreated with TNF-PROT₃EcT but not PROT₃EcT (Figures 6F–G), suggesting that the secreted anti-TNFα Nb sequesters its target and leads to a reduction in IL-6. While others have observed that TNF neutralization or EcN treatment can increase IL-10 production in the gut, we observed equivalent levels of IL-10, regardless of the intervention (Figure 6H).

Discussion

Here we describe the development of PROT₃EcT, a suite of E. coli engineered for the in-situ delivery of high-specificity protein payloads to sites of disease. Using synthetic biology-based approaches, we have engineered both laboratory and non-pathogenic human E. coli isolates with a modified T3SA which can secrete proteins in a regulated or constitutive manner. Furthermore, we have developed PROT₃EcT-3 and PROT₃EcT-4, variants outfitted with a constitutively active secretion system that is maintained in the absence of antibiotic selection, which, when constitutively secreting Nb, exhibits in vitro and in vivo growth rates and colonization profiles equivalent to native EcN.

In terms of its therapeutic potential, orally or rectally administered TNF-PROT₃EcT, a variant that constitutively secretes anti-mTNF Nbs, was as efficacious as systemically administered anti-TNF mAb in limiting the development of TNBS-induced colitis in mice. Although PROT₃EcT, like EcN, appeared to poorly colonize the distal colon, orally administered TNF-PROT₃EcT, but not PROT₃EcT, suppressed inflammation throughout the colon. These observations suggest that low levels of TNF-PROT₃EcT present in the distal colon or descending anti-TNF Nb secreted above are sufficient to block the development of downstream TNFα-mediated colitis. While we were unable to detect Nb within the intestines of mice, presumably due to their degradation via gut proteases, we provide evidence that the T3SA of PROT₃EcT is active within the intestines of mice and TNF-PROT₃EcT, but not PROT₃EcT, significantly ablates its target molecule, TNFα.

Future studies will focus on the ability of TNF-PROT₃EcT to suppress inflammation post the induction of colitis, i.e., as a treatment intervention. Nevertheless, our observation that TNF-PROT₃EcT was as efficacious as systematically administered anti-TNFα monoclonal antibodies in suppressing the development of colitis suggests that it has the potential to serve in a similar capacity, i.e., as maintenance therapy in patients with IBD.

Other groups have also engineered microbes to treat gut inflammation, the most closely related being variants of Lactococcus lactis that secrete IL-10, an anti-inflammatory cytokine, or an anti-TNF Nb^11,12. A native secretion system of this Gram-positive bacterium was repurposed for the secretion of these therapeutic payloads. However, the strain of L. lactis used does not colonize the intestines of humans or mice^52,53, perhaps accounting for why its engineered variants only moderately suppressed inflammation when administered daily to mice. In this regard, although our studies focused on different preclinical models of colitis, it is interesting that we observed that pre-treatment with just a single oral dose of colonizing TNF-PROT₃EcT, but not orally administered poorly colonizing TNF-T₃EcT, significantly ameliorated colonic inflammation and injury. These observations suggest an advantage to treating with colonizing bacteria, particularly when using variants that secrete therapeutic agents.

The modular design of PROT₃EcT is such that it can be adapted to secrete different payloads as well as to respond to environmental cues. For example, by altering the conditions that induce expression of VirB, PROT₃EcT’s T3SA could be endowed with an ‘on switch’ triggered by specific signals of the gut’s inflammatory milieu, e.g., reactive nitrogen species^44,54–56. In terms of therapeutic payloads, we demonstrate the versatility of PROT₃EcT to secrete different Nbs, including Nbs that inhibit the activity of bacterial toxins (Nb^stx2), immune checkpoint molecules (Nb^PD-L1 and Nb^CTLA-4), as well as other heterologous proteins. While the Nbs we studied were each derived from immunized alpacas, synthetic yeast- and bacterial-based Nb libraries are also available that can be screened rapidly for Nbs with desired properties^57,58. By altering its route of administration, PROT₃EcT can be expanded for the deposition of therapeutics not only to the gastrointestinal tract but also to solid tumors, as EcN home to and colonize a variety of solid tumors when administered intravenously, at least in mice⁵⁹.

In studies similar to ours, a type I secretion system (T1SS), native to the Gram-negative plant pathogen Erwinia chrysanthemi, was introduced into EcN and engineered to secrete epidermal growth factor into the gut⁶⁰. They also observed evidence of suppression of gut inflammation in a mouse model of colitis. In the case of T1SSs, proteins of interest are engineered with a carboxy-terminal rather than an N-terminal type III secretion signal sequence. In future studies, it will be interesting to compare the behavior of E. coli engineered with Type I vs. modified Type III secretion systems, particularly as it has been previously demonstrated that Nbs remain active when secreted via this platform from laboratory strains of E. coli⁶¹.

Given its inherent anti-inflammatory and anti-microbial properties, EcN is GRAS and has been used for over a century to treat various intestinal diseases and is available over the counter in some countries. However, EcN contains an operon that mediates the synthesis of colibactin, a genotoxin capable of promoting the formation of DNA crosslinks⁶². Other colibactin-producing E. coli promote the development of colorectal cancer (CRC) in mouse models⁶³ and induce mutational signatures found in human CRC⁶⁴. Whether this will turn out to be an issue that limits the use of EcN-based therapeutics in humans remains to be discovered. However, EcN mutants deficient in colibactin biosynthesis are not impaired in their ability to colonize the intestines of at least mice^65,66 or primates⁶⁷. In future studies, we intend to test the ability of colibactin-deficient EcN-based TNF-PROT₃EcT to suppress intestinal inflammation.

Here, we report that two colibactin-negative strains, E. coli HS and DH10ß, like EcN, can also be engineered with a functional modified type III secretion system. While the anti-TNF Nb-secreting, E. coli DH10ß based, TNF-T₃EcT, was unable to colonize the intestines, daily dosing of rectally administered TNF-T₃EcT suppressed TNBS-induced inflammation as effectively as EcN-based TNF-PROT₃EcT. Additionally, while EcN has been shown to colonize, albeit variably, healthy human subjects for up to 2 weeks, the inflammatory milieu of the IBD gut will likely favor colonization of engineered E. coli strains⁶⁸. Furthermore, other commensal E. coli isolates, including HS, which we study herein, or patient-specific isolates^69,70 could serve as more optimal chassis for the PROT₃EcT platform. Regardless, before testing any of these modified strains in humans, they must be equipped with means to control biocontainment. Some strategies being developed include incorporating circuit-activated kill switches, i.e., those turned on at colder environmental temperatures when bacteria are excreted or auxotrophies that limit environmental growth (for review, see⁷¹).

In summary, we describe the development and characterization of PROT₃EcT, programmable E. coli engineered for the delivery of therapeutic payloads to the gastrointestinal tract. While the presented studies support the further development of PROT₃EcT for the treatment of IBD, its modularity permits its rapid adaptation into a therapeutic platform for a broad range of diseases.

Limitations of the study

Although our study provides proof of concept for the use of PROT₃EcT as a strategy for the delivery of therapeutics to the gut, further preclinical studies are required to demonstrate its efficacy as a therapeutic for IBD. While we have demonstrated that TNF-PROT₃EcT is efficacious in limiting TNBS-driven colitis model; it remains to be determined whether this intervention can ameliorate pre-existing colitis or is efficacious in chronic pre-clinical models of colitis.

STAR METHODS

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for other reagents may be directed to and will be fulfilled by the Lead Contact, Cammie F Lesser (clesser@mgh.harvard.edu).

Materials Availability

Engineered strains and plasmids used in this study are available from the lead contact upon request with a completed Materials Transfer Agreement. The anti-hTNF Nb sequence is restricted pending patent application filing and approval.

Data and Code Availability

• Illumina sequences obtained in the present study were deposited in the Sequence Read Archives (SRA) NCBI database under accession number PRJNA913459 and are currently available. Original western blot images have been deposited at Mendeley and are publicly available as of the date of publication and are currently available at DOI: 10.17632/wwzxwcwj62.1.

• No original code was generated for these studies.

• Any additional information required to reanalyze the data reported in this paper is available from the Lead Contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Bacterial strains

Unless otherwise noted in the experimental method details, E. coli strains were grown in luria broth (LB: 10g/L tryptone, 5g/L yeast extract, 10g/L NaCl) and Shigella flexneri in tryptic soy broth at 37°C with aeration on a roller or on solid media (15% agar). When appropriate, antibiotics (100 μg/ml spectinomycin, 100 μg/ml ampicillin, 50 μg/ml kanamycin, 12.5 μg/ml tetracycline ), Congo Red (10 μM), D-alanine (50 μg/mL), and/or IPTG (1 mM) were added. Bacterial strains are summarized in Table S4. Bacterial strains were grown from a glycerol stock (10% w/v glycerol) by streaking onto solid media. After 18 h of growth, single colonies were picked and inoculated into liquid cultures. Construction strategies and transformation protocols are described in method details.

Cell lines

All cell lines were cultured in a 5% CO₂ incubator at 37°C. HeLa cells were maintained in Dulbeco’s Modified Eagle Medium (c) (Thermo Fisher Scientific #11965), and HCT8 and L929 cells in GlutaMAX-supplemented RPMI 1640 (Thermo Fisher Scientific #61870127). In each case, the media was supplemented with 10% heat-inactivated fetal bovine serum (FBS, R&D systems #S11150), 100 IU/mL penicillin, and 100 μg/mL streptomycin) (Life Technologies #15140). Antibiotics were removed prior to conducting infections.

Mice

All mice were purchased at 6 weeks of age from The Jackson Laboratory and upon arrival allowed to acclimate for at least one week prior to the start of an experiment. Mice were housed in microisolator cages under specific pathogen-free conditions in the barrier facility of Harvard T.H. Chan School of Public Health and had access to food (irradiated PicoLab^® Mouse Diet 20, 5058) and water (autoclaved reverse osmosis) ad libitum. A 12:12 hour light/dark cycle was used. Female C57BL/6 mice were used for colonization studies (Figure 3C and 3D) and female BALB/c mice were used for the TNBS colitis models (Figures 5 and 6) and female C57BL/6 mice were used for the DSS studies (Figure S5). For these studies, the mice were randomly grouped in cages at a maximum of five animals per cage. For the in vivo luciferase imaging assays, kanamycin (1 g/L) and spectinomycin (2 g/L) (Figures 3G and 3H) was administered to mice via their drinking water. For the microbiome studies (Figures 3 and 4), female and male mice were used, and mice were housed n=2 per cage to minimize cage effects. No blinding was performed, except for histopathology scoring. Animal experiments were approved and carried out in accordance with Harvard Medical School’s Standing Committee on Animals and the National Institutes of Health guidelines for animal use and care.

METHOD DETAILS

Plasmids

Plasmids are summarized in Table S4. Sequences of DNA inserts and oligos are summarized in Table S5A–B. Strains transformed with temperature-sensitive plasmids pCP20, pKD46 and pTKRED were maintained at 30°C and cured by incubation 42°C. All gateway entry clone inserts were sequence verified. Synthetic DNA fragments were purchased from Integrated DNA Technologies (IDT), Twist Bioscience, or Genewiz.

Plasmid construction

APC expression plasmids:

DNA fragments composed of APC (amino acids 1342 to 1887) or (1342 to 1982) fused to an OspC2ss were generated using SOEing PCR with oligomers (P1/P2 and P3/P4 or P3/P5) to generate attB1- SS^OspC2-APC60-attB2 and SS^OspC2-APC70-attB2, respectively. The resulting fragment was introduced in pDONR221 followed by pDSW206-ccdB-FLAG via BP and LR reaction to generate pDSW206- SS^OspC2-APC60and pDSW206-SS^OspC2-APC70.

Nb^PD-L1, Nb^CTLA-4, Nb^ASC and Nb^NP1 Gateway compatible destination vectors:

Gateway compatible destination vectors that enable the in-frame introduction of sequences upstream of Nb^PD-L1, Nb^CTLA-4, Nb^ASC and Nb^NP1 were generated by using Gibson cloning to join (1) PCR amplified Nb-HA fragments from pHEN6 Nb52 [anti-IAV NP]/Nb KV 022 [anti-IAV NP]/Nb PD-L1 B3/Nb CTLA-4 H11/Nb ASC with oligomers (P6/P7) and (2) pDSW206-ccdB-MyoD NS with oligomers (P8/P9). Each Nb-containing fragment was introduced into the pDSW206-based vectors via Gibson cloning. The resulting clones were sequence verified. The resulting plasmids are referred to as pDSW206-ccdB-Nb* (* = name of Nb present in construct).

Nb^pD-L1, Nb^CTLA-4, Nb^ASC and Nb^NP1 expression plasmids.

A variety of type III secretion signal sequences were introduced into pDSW206-ccdB-Nb^ASC via LR reaction using pENTR-secretion sequence entry clones. An OspC2 secretion signal (OspC2ss) was introduced into pDSW206-ccdB-Nb^PD-L1, pDSW206-ccdB-Nb^CTLA4 or pDSW206-ccdB-Nb^NP1 via a LR reaction with pENTR-OspC2ss.

Alr plasmids.

A DNA fragment containing alr and its native promoter with flanking attB1 and attB2 sites PCR-amplified from EcN using oligomers (P10/P11) was introduced into pDONR221 via a BP reaction followed by pCMD136-ccdB-FLAG via an LR reaction. The resulting plasmid, pCMD136-alr, was used as a template for nested PCR with oligomers (P12/P13, P12/P14 and P15/P16) to generate a DNA fragment (BspHI-Palr-alr-T7t-AseI) that was introduced via traditional cloning into pDSW206 digested with NcoI/NdeI to create pCGP-alr.

Nb^STX2 expression plasmids.

A synthetic DNA fragment [attB1-OspC2ss-three Nb^Stx2 (JFG-H6, JFD-A5 and JGH-G1)-E-tag-attB2] was introduced into pDONR221 via a BP reaction followed by pDSW206-ccdB-3xFLAG via an LR reaction to create pDSW206- SS^OspC2-Nb^3xStx2. Heterodimeric (JFG-H6, JFD-A5) and monomeric (JFG-H6) DNA fragments PCR amplified from pDSW206- SS^OspC2-Nb^3xStx2using oligomers (P17/P18 and P17/P19) were introduced into pDONR221 via a BP reaction followed by pDSW206-ccdB-3xFLAG via LR reactions to create pDSW206- SS^OspC2-Nb^2xStx2and pDSW206- SS^OspC2-Nb^3xStx2.

To replace the IPTG-inducible Ptrc promoter and the lacIq repressor in pDSW206- SS^OspC2-Nb^2xStx2 with a constitutive promoter, two complementary oligos (P20/P21) were annealed to create a BBa_J23115 promoter (Anderson Collection) with cohesive SphI and SacI ends. The dsDNA fragment was cloned into SphI/SacI digested pDSW206- SS^OspC2-Nb^2xStx2to create pDSW206-J23115- SS^OspC2-Nb^2xStx2. The (PJ23115- SS^OspC2-Nb^2xStx2) fragment in this plasmid was PCR amplified (P22/P23) and introduced into KpnI/XbaI-digested pCGP-alr to create pCGP-alr-PJ23115- SS^OspC2-Nb^2xStx2. Two complementary oligos (P19/P20) were annealed to create a BBa_J23018 (Anderson Collection) promoter with cohesive SacI ends. The dsDNA fragment was cloned into SacI-digested pCGP-alr-PJ23115- SS^OspC2-Nb^2xStx2 replacing PJ23115 to create pCGP-alr-PJ23108- SS^OspC2-Nb^2xStx2.

Nb^TNF expression plasmids.

Synthetic DNA fragments [attB1–SS^OspC2-Nb^mTNF-attB2] and [attB1- SS^OspC2-Nb^2xhTNF-attB2] were introduced into pDONR221 via BPs reaction followed by pDSW206-ccdB-FLAG via LR reactions to create pDSW206- SS^OspC2-Nb^mTNF and pDSW206- SS^OspC2-Nb^2xhTNF. A DNA fragment composed of homodimer Nb^mTNF fused to an OspC2ss was generated using SOEing PCR with oligomers (P17/P27 and P26/P23) to generate attB1- SS^OspC2-Nb^2xmTNF-attB2. The resulting fragment was introduced in pDONR221 followed by pDSW206-ccdB-FLAG via BP and LR reaction to generate pDSW206- SS^OspC2-Nb^2xmTNF and pDSW206-alr-PJ23115- SS^OspC2-Nb^2xhTNF were constructed as previously described for Nb^Stx2.

VirB expression plasmids.

Entry clones that contain virB under the control of various promoters were obtained via SOEing PCR using two synthetic DNA fragments and oligomers (P28/P29). One synthetic DNA fragment contained a promoter flanked by an upstream attB site and downstream by 40 bp of homology to virB. The second DNA fragment contained the open reading frame of virB codon-optimized for expression in E. coli with an upstream RBS and a downstream stop codon followed by an attB site. The RBS Calculator tool version 1.1⁸², with organism option as E. coli str. K-12 substr. MG1655 was used to choose the RBS. The resulting DNA fragments were introduced into pDONR221 via a BP reaction followed by pCMD136-ccdB-FLAG via an LR reaction. pTKIP-PJ23119-virB was generated by PCR amplifying PJ23119-virB from pCGP-PJ23119-virB-Nb with oligomers (P30/P31) that add a 5′ KpnI site and a 3′ rrnB-homology region. Using pCMD136 as a template, the rrnB terminator was amplified with a 5′ virB-homology region and a 3′ HindIII site using oligomers (P32/P33). The two fragments were fused together by SOEing PCR using oligomers (P30/P33). The product was digested with KpnI and HindIII and ligated into the polylinker of pTKIP-hph.

MxiE-Luciferase expression plasmid.

A DNA fragment that contains a MxiE-promoter was PCR amplified from pTSAR1Ud2.4s² using oligomers (P34/P35) that add flanking 5′ XhoI and 3′ KpnI restriction sites and an RBS. The digested PCR product was ligated into XhoI/KpnI pMM534 to generate pMxiE-lux.

Strain construction

PROT₃EcT-1 and PROT₃EcT-2.

A synthetic 1.3 kb landing pad insertion site was introduced into the atg/gid loci of EcN and E. coli HS to generate EcN-LP^atp/gid and EcHS-LP^atp/gid using the lambda red recombination system and the pTKRED helper plasmid. The landing pad fragment was PCR amplified from pTKIP-tefA with oligos (P36/P37) to introduce homology to the atg/gid locus and integration was confirmed by PCR with oligo pairs (P38/P39 and P40/P41). The pmT3SA plasmid which contains the 20 kb mxi-spa operons flanked by LP and SceI sequences was introduced into EcN-LP^atp/gid and EcHS-LP^atp/gid via triparental mating: donor (DH10ß/pT3SA), helper HB101 (pRK2073) and recipient (EcN- or EcHS-LP^atp/gid/pKD46). pKD46-cured EcN- and HS-LP^atp/gid containing pT3SA were transformed with pTKRED and the landing pad recombination system was used to generate PROT₃EcT-1 and PROT₃EcT-2. KAN resistant/TET susceptible transformants were screened for proper integration junctions by PCR with oligo pairs (P38/P42 and P40/P43).

PROT₃EcT-3.

PROT₃EcT-1 was modified with a landing pad at its yieN/trkB locus to create PROT₃EcT-1-LP^yie/trk. By PCR, the landing pad with appropriate homology regions was amplified with oligos (P44/P45) and its integration was confirmed with P46/P39 and P47/P41. PROT₃EcT-1-LP^yie/trk was transformed with pTKred and pTKIP-PJ23119-virB and the landing pad platform was used to introduce the VirB expression construct into the chromosome. Integration was confirmed by PCR with oligos P48/P33 and P49/P32.

PROT₃EcT-4.

After first resolving the KAN^R marker previously used to introduce the mxi-spa operons into PROT₃EcT-1 using the FLP recombinase, the lambda red recombination system⁷⁴ was used to sequentially delete aIr and dadX from PROT₃EcT-3 using oligomers (P50/P51 and P52/P53). The KAN^R was removed from the aIr locus, before proceeding to delete dadX. Deletions were confirmed by PCR with oligomers (P54/P55 and P56/P57, respectively).

Gentamicin protection assays

HCT8 cells were seeded in 96-well plates (4×10⁴ cells per well) for 18 h prior to exposure to bacteria. Bacteria grown overnight with aeration at 37°C were back-diluted and subcultured for one hour before the addition of IPTG (1 mM). One hour later, the HCT8 cells were washed and then infected at an MOI of 100. After 30 min, cells were washed 3 times and gentamicin (50 μg/mL) was added, and 30 minutes later, cells were washed 6 times then lysed with 1% triton X-100 in PBS. Bacteria were plated and enumerated. Percentage of internalized bacteria was determined by calculating the ratio of gentamicin-resistant bacteria to the initial inoculum.

Translocation assays

Translocation assays were performed as previously described²² with some modifications. HeLa cells were seeded in 6-well plates (3x10⁵ cells per well) for 18 h prior to exposure to bacteria. Bacteria grown overnight with aeration at 37°C were back-diluted and subcultured for one hour before the addition of IPTG (1 mM). One hour later, the cells were washed 3 times then infected at an MOI of 100. After 1 h, gentamicin (50 μg/mL) was added, and 2 h later, cells were washed 6 times then lysed with RIPA buffer (25 mM Tris, pH 8, 150 mM NaCl, 0.1% SDS, 1% NP-40 plus cOmpleteTM cocktail protease inhibitors. The soluble and insoluble (bacterial-containing) fractions were separated by centrifugation before the addition of protein loading dye (final concentration: 10% glycerol, 50 mM Tris/HCl pH 6.8, 2% SDS, 0.02% bromophenol blue, 1% beta-mercaptoethanol). 60 μL of each fraction was loaded onto 12% SDS-PAGE gel for analysis. Proteins were transferred to nitrocellulose membranes and immunoblotted with mouse anti-FLAG (1:10,000) and anti-Beta actin (1:10,000). Unprocessed immunoblots can be found at DOI: 10.17632/wwzxwcwj62.1.

Liquid secretion assays

Liquid Secretion assays were performed as previously described³ with some modifications. Overnight cultures of E. coli grown in LB (Luria broth) were back diluted 1:50. Once cultures reached OD₆₀₀ of 1.2-1.5, the bacteria were pelleted and resuspended in 2 mL of fresh media or PBS and incubated for the times indicated. IPTG (1 mM) was added when studying Ptrc regulated virB and/or nb. When indicated, bacteria were exposed to 10 μM Congo red. After designated periods of time, total cell and supernatant fractions were separated by centrifugation at 20,000g for 2 min. The cell pellet was taken as the whole cell lysate fraction. The supernatant fraction was subjected to a second centrifugation step to remove any remaining bacteria. To account for differences in bacterial titers, for each set of experiments, the volume of protein loading dye used to resuspend each sample was normalized by the OD₆₀₀ reading of the slowest growing culture. For the time course assays, samples were not normalized. The pellet was resuspended in 100 μL or more protein loading dye, depending on the OD₆₀₀. Proteins in the supernatant were precipitated with trichloroacetic acid (TCA) (10% v/v) and resuspended in 50 μL or more protein loading dye, depending on the OD₆₀₀. Proteins resuspended in loading dye were incubated at 100°C for 15 min. Ten microliters of TCA-precipitated supernatant samples (20%) and five microliters of the pellet (5%) were loaded onto a 12% SDS-PAGE gel for analysis (i.e., the ratio of supernatant to pellet samples loaded was 3:2). In a typical experiment for an OD₆₀₀ of 1.5, this equates to loading an OD₆₀₀ of approximately 0.075 or 7.5x10⁶ bacteria cell lysates (OD₆₀₀/CFU conversion of 1*10^8), for pellet samples. Proteins were transferred to nitrocellulose membranes and immunoblotted with mouse anti-FLAG (1:10,000), mouse anti-HA (1:1000) or rabbit anti-GroEL (1:100,000). Unprocessed immunoblots can be found at DOI: 10.17632/wwzxwcwj62.1. When stated, SDS-PAGE gels were stained with GelCode^™ Blue Stain Reagent, per the manufacturer’s instructions.

Plasmid retention assay

Bacterial cultures were back diluted daily for 7 days into LB media containing no antibiotics. Each day cultures were sampled and plated on LB media to quantify total bacteria and LB/ampicillin plates to quantify the percentage of bacteria that retained the plasmid Nb-expressing plasmid.

Stx2 binding assay

Bacterial supernatants were assessed for their ability to bind to Stx2 (0157) using an ELISA, as previously described ⁸³. Briefly, 1 μg of Stx2 was coated onto high bind plates. After overnight incubation, the plates were washed in PBS-Tween 20 (0.05%) and blocked for 1 h using 4% milk/PBS-Tween 20 (0.05%) then serial dilutions of bacterial supernatants were added in a total volume of 100 μL. Following a 1 h incubation, the plates were washed, and HRP/anti-FLAG antibody added. After a 1 h incubation the plates were washed and developed using TMB substrate solution. After a 20 min incubation, the reaction was stopped using 1N HCL, and the plates read at 450 nm.

Vero cell cytotoxicity assay

Bacterial supernatants were assessed for their ability to neutralize Stx2 (0157) using an Stx2-induced cytotoxicity assay in Vero cells, as previously described⁸⁴. Briefly, Vero cells were seeded on 96-well plates (5 × 10⁴ cells/well). After overnight incubation, the culture medium was replaced with serial dilutions of bacterial supernatants prepared in RPMI media containing a final concentration of Stx2 (2 ng/mL), a dose known to kill more than 90% of the cells under the conditions used. After 48h, plates were developed by crystal violet staining and the absorbance at 590 nm was read.

Bacterial growth assays

Overnight cultures of E. coli strains grown in LB were back diluted 1:100, normalizing to the slowest growth culture. At the time points indicated, OD600 readings were recorded. For the in vitro competition assay, overnight cultures of E. coli strains grown in LB were diluted 1:100 and added at 1:1 ratio into fresh media. At the time points indicated, serial dilutions were plated on LB agar containing antibiotic selective for each of the strain and after 18 h at 37^°C, CFUs were counted.

Solid plate secretion assay

Solid plate secretion assays were performed using a BMC3-BC pinning robot (S&P Robotics) as previously described. Briefly, single colonies grown overnight in 96 well plates were quad spotted onto a solid agar²¹. After overnight growth, a 384-pin tool is used to transfer equivalent amounts of bacteria to a media-containing plate over which a nitrocellulose membrane was immediately laid. After 6 h, the membrane was removed, washed to remove adherent bacteria, and immunoblotted for protein of interest. All incubations were carried out at 37°C

Fecal shedding assay.

Fecal pellets were collected and weighed. A 10x volume of PBS was added and the samples homogenized by pipetting and mashing using wide mouth pipette tips before being serially diluted and plated on MacConkey agar plates with antibiotics. After overnight incubation at 37°C colonies were counted and the total number of CFU estimated. On days where bacterial inoculations were performed, fecal pellets were sampled prior to inoculation.

In vitro luciferase monitoring

In a 96-well white plate, 1:100 dilutions of overnight bacterial cultures were incubated at 37°C with shaking for 5 h. Readings were performed on a SpectraMax i3x Multi-Mode Microplate Detection Platform (Molecular devices).

In vivo luminescence assays

To image luciferase-expressing bacteria in the GI tract, mice pre-treated for 1 day with kanamycin (1 g/L) and spectinomycin (2 g/L) in their drinking water were orally gavaged with 10⁸ CFU of PROT3EcT-3 or EcN harboring the constitutive luciferase or MxiE reporter plasmids. After sacrificing the mice, the cecum, colon, and small intestine were harvested, the contents gently removed, and the tissues placed on a black mat for imaging using an IVIS Spectrum CT. Tissues were imaged using a luminescence filter, with field of view (FOV) = D (22.2 cm), fstop = 1 and large binning. Data was analyzed using Living Image Software 4.3.1.

Discovery and characterization of Nbs targeting murine and human TNFα.

Two alpacas were each immunized once with 200 μg murine (m)TNF-α in CpG/alum adjuvant, followed by four boosters with 100 μg mTNFα in alum adjuvant only. Two different alpacas were similarly immunized with human TNFα. Construction of the Nb display phage libraries from both alpaca pairs and panning and screening of the libraries were done as previously described⁸³. Initial panning for anti-mTNFα Nb was performed with the target immobilized by coating to MaxiSorp Nunc-lmmuno tubes. Because the yield was low, panning was repeated with mTNFα bound to JTT-B10 Nb, a Nb obtained in the initial screen. This second panning yielded 20 unique mTNFα-binding Nb families. The anti-hTNF Nb was isolated by panning a 6His-tagged hTNFα immobilized using an anti-His-tag mAb. The coding sequences of representative members of each anti-TNFα family were introduced into a pET32b expression vector and expressed as thioredoxin, 6His, E-tag fusion proteins in E. coli Rosetta-gami 2 (DE3) pLacI as fusions to thioredoxin and to 6His to facilitate purification using standard Ni-IMAC chromatography methods, and a carboxyl-terminal E-tag for detection. Based on ELISA⁸³, 11 unique anti-mouse and 1 unique anti-human TNF Nbs with 10 nM or better apparent affinities were selected for further analyses (Table S2).

A competition study was conducted to determine whether any of the other unique Nb bound to epitopes not recognized by JTT-B10 by performing replicate dilution ELISAs in which the only variation was that one set of ELISAs contained 20 μg/ml of the JTT-B10 Nb protein as a competitor in which the E-tag detection tag was replaced with a myc tag. The study identified three Nb families that bind to epitopes not competed by JTT-B10 (Table S2).

Quantification of PROT3EcT secreted Nb^mTNF

500 μL of bacterial culture supernatants were concentrated using Amicon^® Ultra 0.5 mL Centrifugal Filters per the manufacturer’s protocol with modifications. Filters containing bacterial supernatants were centrifuged for 20 min at 2,500 g. Loading dye was added to the supernatant and 15 uL of this mixture was incubated at 100°C for 15 min before being run on a 12% SDS-PAGE gel. A purified Nb^2xTNF at a known concentration was serially diluted in loading dye, boiled, and run alongside the concentrated bacterial supernatant samples. Gels were probed with GelCodeBlue Reagent and densitometry was performed using ImageJ to generate a standard curve of the Nb^2xTNF standards from which the concentration of SS^OspC2-Nb^2xmTNF in bacterial supernatants was interpolated.

L929 cell cytotoxicity assay

Murine and human Nbs^TNF and bacterial supernatants were assessed for their ability to neutralize mTNFα or hTNFα using a TNFα-induced cytotoxicity assay in L929 cells, as previously described ⁸⁵. Briefly, 100 μl/well of murine fibroblast L929 cells seeded in 96-well plates (5 × 10⁴ cells/well). After overnight incubation, the culture medium was replaced with serial dilutions of bacterial supernatants or purified Nb prepared in RPMI media containing a final concentration of 1.0 μg/ml actinomycin D and 4 ng/mL murine or human TNF-α. Plates were then incubated at 37 °C for 24 h after which an MTT assay was performed as per the manufacturer’s instructions.

TNBS and DSS mouse models of colitis and treatment protocols

Time points and doses for all treatments and administrations are indicated in the Figures and text. TNBS was diluted to 20 mg/mL in ethanol (50% v/v) and 100 μl administered via enema by inserting a 3.5 French catheter 4 cm into the colon. DSS (3% w/v) was provided ad libitum in the drinking water for 5 days at the time points indicated in the Figures. Bacterial strains were prepared as described and administered via oral gavage or enema. Anti-TNF mAb was administered intraperitoneally (i.p.). Mice were euthanized by CO₂ overdose. Upon sacrifice, blood was harvested by cardiac bleed, the GI tracts excised, and colon lengths measured. Blood was collected into serum separator tubes, spun for 5 min at 5000 rpm and serum stored at −20°C. The colon was cut longitudinally, the contents were removed, and the tissue dissected. Half of the tissue was fixed using 4% paraformaldehyde (PFA) overnight at 4°C for histology. The other half was homogenized in 1 mL of PBS containing 1x HALT protease inhibitor cocktail before being centrifuged for 10 min at 20,000 g and the supernatant stored at −20°C for later analysis by ELISA.

16S rRNA gene amplicon sequencing of stool

16S rRNA gene amplicon sequencing was performed on stool isolated from mice using protocols adapted from the Earth Microbiome Project, as previously described⁸⁶. From extracted fecal DNA, the 16S rRNA V4 region was amplified by PCR with a single-indexing approach and indexing barcodes as listed in Table S5B. Sequencing was performed on a MiSeq instrument (Illumina, San Diego, CA) using 1x150bp paired-end protocol (Mid Output: 10M clusters) by the Molecular Biology Core Facilities (MBCF) at Dana-Farber Cancer Institute (DFCI). The quality of the reads was checked using FastQC (v0.11.5). Read pairs were imported to the QIIME2 environment (version 2021.2), where they were joined, denoised, and checked for chimeras using the DADA2 plug-in in prior to taxonomic assignment. Taxonomic assignment of each amplicon sequence variant (ASV) was performed using a pre-trained Naive Bayes classifier Tables S1A and S3A with the SILVA database (version 138.1). Differential abundance analysis of all ASVs was performed using MaAslin2 ASVs found in less than 10% of samples were not included in the testing. Each ASV was modeled as a function of treatment and sex (categorical variables), with caging as a random effect. ASVs with a corrected q-value of less than 0.2 are considered significant. Raw sequences were deposited in NCBI SRA databank, Bioproject PRJNA913459.

Cytokine and Nb ELISAs

The concentrations of mouse TNFα, IL-10, IL-6 were quantified by ELISA per the manufacturer’s instructions. The anti-Nb ELISA was performed as previously described⁸⁵.

Histology

PFA-fixed colon tissue was transferred to 70% ethanol before processing by routine paraffin embedding, sectioning and H&E staining by the DF/HCC Rodent Histopathology Core. A pathologist (J.N.G.), blinded to experimental parameters, determined colitis scores. Each of the following four histologic parameters were scored as absent (0), mild (1), moderate (2), or severe (3): mononuclear cell infiltration, polymorphonuclear cell infiltration, epithelial hyperplasia, and epithelial injury. The scores for the parameters were summed to generate the cumulative histologic colitis score⁸⁷. The cumulative histologic colitis score was then multiplied by an extent score, indicating the proportion (%) of colon involved by colitis: (1) < 10%; (2) 10%–25%; (3) 25%–50%; (4) > 50%. Images were captured at 10× or 40× magnification with a Nikon Eclipse NI-U and NSI-Element Basic Research software (Nikon).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were performed using GraphPad Prism v.8.3.0 or MaAslin2. Details of statistical analysis and sample size are provided in the figure legends. No samples were excluded from any experiments performed in this study unless in the case of technical failure. Experimenters, with the exception of the pathologist, were not blinded to experimental conditions.

Supplementary Material

Supplemental Material

Table S2 mTNF Nb affinities (related to Figure 3)

Table S1

Table S1 16S rRNA gene amplicon fecal survey data and analysis from C57BL/6 mice (related to Figure 3)

Table S3

Table S3 16S rRNA gene amplicon fecal survey data and analysis from BALB/c mice (related to Figure 4)

Table S4

Table S4 Plasmids and strains (related to STAR Methods)

Table S5

Table S5 DNA and oligo sequences (related to STAR Methods)

KEY RESOURCES TABLE

Reagent or resource	Source	Identifier
Antibodies
Anti-GroEL antibody, rabbit	Sigma	Cat#G6532
Anti-FLAG, mouse (clone M2)	Sigma	Cat#F1804
Anti-HA.11, mouse (clone 16B12)	Biolegend	Cat#901501
Anti-Beta actin, mouse (clone AC-15)	Abcam	Cat#ab6276
Anti-rabbit antibody	Jackson ImmunoResearch Laboratories	Cat#111-035-144
Anti-mouse antibody	Jackson ImmunoResearch Laboratories	Cat#115-035-003
Anti-TNFα, mouse (clone TN3-19.12)	BioxCell	Cat#BE0244
Anti-his-tag mAb	Genescript	Cat#A00186
Chemicals, Peptides and Recombinant Proteins
Congo red	Sigma Aldrich	Cat#C6277
Recombinant Mouse TNFα	Biolegend	Cat#575206
Recombinant Human TNFα	Biolegend	Cat#570106
PicoLab® Mouse Diet 20	Constant Nutrition	Cat#5058
CpG/alum adjuvant	InvivoGen	Cat#2395
Stx2 (0157)	Tufts University	N/A
His-tagged hTNFα	Genescript	Cat#TNA-H5228
TNBS	Sigma	Cat#92822
DSS	MP Biomedicals	Cat#160110
Ambion UltraPure Buffer Saturated Phenol	Thermo Scientific	Cat#15513039
Commercial assays
Mouse TNFα ELISA	Biolegend	Cat#430915
Mouse IL-10 ELISA	Biolegend	Cat#431414
Mouse IL-6 ELISA	Biolegend	Cat#431316
Gibson Cloning Kit	New England Biolabs	Cat#E5510S
MTT assay	Trevigen	Cat#4890-25-K
TMB substrate solution	Biolegend	Cat#421501
Amicon® Ultra 0.5 mL Centrifugal Filters	MerckMillipore	Cat# UFC5010BK
Deposited data
16S rRNA gene amplicon sequencing. ASVs with a corrected q-value of < 0.2 were considered significant and deposited.	This paper	SRA BioProject: PRJNA913459
Experimental Models:
Cell Lines
L929	Gift of Marcia Goldberg	N/A
HCT8	Gift of Marcia Goldberg	N/A
HeLa CCL2	American Type Culture Collection (ATCC)	N/A
Mice
C57BL/6	Jackson Labs	IMSR_JAX:000664
BALB/c	Jackson Labs	IMSR_JAX:000651
Bacterial Strains (see Table S4 for additional details)
Shigella flexneri	Gift of Marcia Goldberg
E. coli DH10ß	Life Technologies	N/A
EcN	Mutaflor, Canada	Cat#18290015
EcHS	Gift of Mark Goulian	N/A
E. coli HB101	Gift of Stephen Lory	N/A
mT3Ec_Ipa-Mxi-Spa	Reeves et al.23	N/A
mT3Ec_Mxi-Spa	Ernst et al.21	N/A
mT3Ec_Mxi-Spa-VirBINT	This study	N/A
Stx2-mT3Ec_Mxi-Spa-VirBINT	This study	N/A
TNF-mT3Ec_Mxi-Spa-VirBINT	This study	N/A
EcN-LPatp/gid	This study	N/A
EcHS- LPatp/gid	This study	N/A
PROT3EcT-1	This study	N/A
PROT3EcT-2	This study	N/A
PROT3EcT-1- LPyie/trk	This study	N/A
PROT3EcT-3	This study	N/A
PROT3EcT-4	This study	N/A
Stx2-PROT3EcT-4	This study	N/A
TNF-PROT3EcT-4	This study	N/A
Oligonucleotides (see Table S5B)
Recombinant DNA (see Table S4 for additional details)
pDSW206-ccdB-MyoD NS	Reeves et al.23	N/A
pDSW206-OspC2-FLAG	Schmitz et al.72	N/A
pRK2073	Leong et al.73	N/A
pKD46	Datsenko and Wanner74	N/A
pCMD136-ccdB-FLAG	Mou et al.22	N/A
pDSW206	Weiss et al.75	N/A
pDSW206-ccdB-FLAG	Schmitz et al.72	N/A
pTKIP-hph	Kuhlman and Cox76	N/A
pTKred	Kuhlman and Cox76	Addgene #41066
pTKLP-tetA	Tas et al.77	Addgene #41062
pMM534	Mimee et al.44	Addgene #71325
pTSAR1Ud2.4s	Campbell-Valois et al.78	Addgene #112530
pCP20	Datsenko and Wanner 74	N/A
pKD4	Datsenko and Wanner 74	N/A
pMixE-lux	This study	N/A
pNG162-IpgC	This study	N/A
pDONR221	Invitrogen	Cat#1253601
pCGP-alr	This study	N/A
pmT3SA	Ernst et al.21	N/A
pmT3SS	Reeves et al.23	N/A
pNG162-virB	Reeves et al.23	N/A
pCMD136-PompC-virB	This study	N/A
pCMD136-PvirF-virB	This study	N/A
pCMD136-PJ23119-virB	This study	N/A
pCMD136-PJ23115-virB	This study	N/A
pCGP-PJ23119-virB-Nb	This study	N/A
pTKIP-PJ23119-virB	This study	N/A
pENTR-OspC2ss-APC60	This study	N/A
pENTR-OspC2ss-APC70	This study	N/A
pDSW206-OspC2ss-APC60	This study	N/A
pDSW206-OspC2ss-APC70	This study	N/A
pDSW206-OspGss-MyoD	Reeves et al.23	N/A
pDSW206-OspGss-Mef2C	Reeves et al.23	N/A
pDSW206-OspCss-TALE	Reeves et al.23	N/A
pENTR-OspC2ss	Reeves et al.23	N/A
pENTR-IpaH4.5ss	Reeves et al.23	N/A
pENTR-IpaH7.8ss	Reeves et al.23	N/A
pENTR-IpaH9.8ss	Reeves et al.23	N/A
pENTR-OspEss	Reeves et al.23	N/A
pENTR-OspFss	Reeves et al.23	N/A
pENTR-OspD3ss	Reeves et al.23	N/A
pENTR-VirAss	Reeves et al.23	N/A
pENTR-OspGss	Reeves et al.23	N/A
pDSW206-ccdB-NbASC	Reeves et al.23	N/A
pDSW206-ccdB-NbPD-L1	This study	N/A
pDSW206-ccdB-NbCTLA4	This study	N/A
pDSW206-ccdB-NbNP1	This study	N/A
pDSW206-ccdB-NbASC	This study	N/A
pDSW206-OspC2ss-3xNbStx2	This study	N/A
pDSW206-OspC2ss-2xNbStx2	This study	N/A
pDSW206-OspC2ss-NbStx2	This study	N/A
pDSW206-PJ23115-OspC2ss-2xNbStx2	This study	N/A
pCGP-alr-PJ23115-OspC2ss-2xNbStx2	This study	N/A
pCGP-alr-PJ23108-OspC2ss-2xNbStx2	This study	N/A
pENTR-OspC2ss-NbmTNF	This study	N/A
pDSW206-OspC2ss-NbmTNF	This study	N/A
pENTR-OspC2ss-NbmTNF	This study	N/A
pDSW206-OspC2ss-2xNbmTNF	This study	N/A
pDSW206-PJ23115-OspC2-2xNbmTNF	This study	N/A
pCGP-alr-PJ23115-OspC2-2xNbmTNF	This study	N/A
pENTR-OspC2ss-2xNbhTNF	This study	N/A
pDSW206-OspC2ss-2xNbhTNF	This study	N/A
Software and Algorithms
GraphPad Prism 9	GraphPad Software	https://www.graphpad.com/
BioRender	BioRender Company	https://biorender.com/
ImageJ	NIH	https://imagej.nih.gov/
QIIME2	Bolyen et al.79	https://qiime2.org/
DADA2	Callahan et al.80	https://benjjneb.github.io/dada2/
MaAsLin 2	Mallick et al.81	https://github.com/biobakery/Maaslin2
R	version 4.2.1	N/A

Highlights.

Engineering of PROT₃EcT, E. coli outfitted with a programmable protein secretion system

PROT₃EcT secrete functional nanobodies into their surroundings

Pretreating with Anti-TNFα secreting PROT₃EcT limits colitis in a preclinical IBD model