Theranostic Probiotic Engineered Living Materials that Dynamically Respond to Inflammation Markers

Abstract

The development of smart, implantable devices localized at the site of inflammation to conditionally and proactively combat active inflammation for inflammatory bowel disease (IBD), have the potential to transform the patient’s quality of life compared to conventional treatment modalities. Engineered probiotic organisms can enable dynamic production of therapeutic compounds in response to inflammatory biomarkers. However, delivery and localization of these engineered organisms to the site of inflammation requires their integration into a material or device that sustains their viability and metabolic activity. To this end, we developed a 3D printed engineered living material (ELM) using an engineered probiotic organism (E. coli Nissle 1917) with genetic circuits to sense biomarkers for inflammation and respond with the production of anti-inflammatory compounds. These organisms were incorporated into poly(ethylene glycol) diacrylate (PEGDA) resins for the light-based 3D printing of 3D constructs. The organisms were physically encapsulated within the PEGDA and were fully viable and metabolically active. The 3D printed ELM devices were able to detect clinically relevant amounts of nitric oxide as an inflammatory biomarker and respond with the production of tryptamine or 1-acetyl-3-carboxyl-β-carboline as representative anti-inflammatory agents. Additionally, the ELM devices were efficacious in treating in vitro models of inflammation including murine macrophages and intestinal epithelial cells. Looking forward, these ELM devices could serve as theranostic modalities for the long-term treatment of inflammatory disorders such as IBD.

Graphical Abstract

3D printed engineered living materials (ELMs) comprising a genetically engineered probiotic organism (E. coli Nissle 1917) were developed to sense biomarkers for inflammation and respond with the production of anti-inflammatory compounds. These living, theranostic devices demonstrated efficacy and biocompatibility in in vitro models, and offer a promising approach to the treatment and management of inflammatory bowel disease.

Introduction

Engineered living materials (ELMs) comprised of polymer networks encapsulating engineered microorganisms are emerging as a powerful platform to treat myriad chronic diseases^1,2,3. For diseases localized to the gut, such as inflammatory bowel disease (IBD)⁴, the extracellular matrix of an ELM provides a protective barrier against the harsh gut environment for engineered microorganisms and localizes treatment at the diseased site. The protective niche of the ELM also enables engineered microorganisms, including probiotics, to reside in the gut for extended periods of time without the need to outcompete native microbiome populations for colonization. Natural biopolymers like alginate, cellulose, and chitosan are prevalent in traditional probiotic-ELM formulations to fabricate enteric coated capsules^5,6,7. However, natural biomaterials widely used for probiotic encapsulation are unstable in the gastrointestinal environment for extended time periods⁸. To address this limitation and create a more stable encapsulation matrix, probiotic Escherichia coli Nissle 1917 (EcN) has recently been engineered to synthesize its own extracellular matrix, providing enhanced protection for the cells against adverse gut conditions^9,10. Despite these advances, both natural polymers and genetically engineered probiotic-derived extracellular matrices remain limited in their processability, which could be addressed using light-based 3D printing techniques to produce medically relevant 3D form factors, such as stents.

Clinical treatment options for IBD typically scale with the severity of disease state and consist of systemically-administered drugs such as amino salicylates, antibiotics, corticosteroids, and immune-modulating agents^11,12,13. These treatments are subject to lack of patient compliance, require frequent dosing, and may have dose limiting side effects^11,12,14. Oral administration is most convenient for patients, however, oral therapies to treat intestinal conditions must (i) avoid premature release in the upper gastrointestinal tract, (ii) incorporate a trigger mechanism for timely drug release, and (iii) ensure that the prescribed dose is administered over the appropriate duration¹⁵. Current strategies that facilitate drug delivery to the intestine, including prodrugs, pH-responsive polymer coatings, selectively biodegradable polymers, timed-release systems, and osmotic-controlled systems^7,8, remain limited by variability in disease location and systemic absorption.

Consequently, implantable drug delivery devices for localized administration, such as stents, patches, and microneedle devices are preferred^16,17. These devices can be positioned near affected areas to provide precise and targeted therapeutic delivery. However, traditional drug-loaded devices, including small molecule-loaded polymeric stents, are still lacking due to their limited drug capacity, shelf-life stability, and uncontrolled rate of drug release^17,18,19. ELMs that are stable for months can potentially address several of these shortcomings and enable long-term treatment and maintenance of IBD.

Advances in the field of synthetic biology have inspired the use of genetically tractable probiotic chassis to create engineered living therapeutics that are capable of sensing disease biomarkers within a patient and responding with the secretion of a therapeutic agent^20,21. E. coli Nissle 1917 (EcN) is highly attractive in this regard, given its amenability to the large set of genetic tools developed for laboratory strains of E. coli and its capacity to serve as a functional probiotic. Recent developments in IBD management have demonstrated EcN systems capable of sensing transient IBD biomarkers in vivo, including nitric oxide (NO) and other reactive oxygen species²², thiosulfate and tetrothionate²³, and calprotectin²⁴. Studies aimed at IBD treatment have engineered EcN capable of secreting catalase and superoxide dismutase⁶, mucosal healing agents¹³, elafin²⁵, and anti-TNF mAbs²⁶.

Additional investigations have coupled sensing with production, in order to create engineered probiotics capable of tuning therapeutic output to presence of a disease biomarker^27,28. On the other hand, free cell probiotic theranostic methods on their own face notable limitations in achieving sustained delivery to diseased tissues, as they encounter viability challenges in the harsh pH of the stomach and the high bile salt content of the small intestine²⁹. Furthermore, EcN does not typically colonize the gut of adult humans without antibiotic pretreatment, thus frequent dosing of free cells is essential to maintain effectiveness³⁰.

Herein, we present 3D printed ELMs containing engineered probiotic organisms with genetic circuits for the in situ, conditional production of anti-inflammatory agents (Figure 1). An engineered probiotic bacterium, E. coli Nissle 1917 (EcN) was incorporated into poly(ethylene glycol) diacrylate (PEGDA) resins for light-based 3D printing. We previously reported mechanically tough PEGDA-glycerol ELMs that were 3D printed and cell compatible³¹. Here, we demonstrate that these matrices can afford stable drug delivery devices that include a sense-and-respond feature for the delivery of anti-inflammatory agents within the gastrointestinal tract. The EcN within these 3D printed PEGDA matrices were engineered with a synthetic circuit to sense the presence of NO as an inflammatory marker and respond with the production of tryptamine or 1-acetyl-3-carboxyl-β-carboline (known for their anti-inflammatory activity in in vitro models^32,33). We validated the production of these therapeutics in media that simulated the intestinal and colonic environments. The efficacy of the ELMs was demonstrated in multiple in vitro models of inflammation, including murine macrophages and intestinal epithelial cells. Our study highlights the potential for developing ELM-based living theranostics for the long-term treatment of IBD.

Figure 1.

a) Components of theranostic ELMs: E. coli Nissle 1917 was metabolically engineered to produce anti-inflammatory agents in response to the inflammation biomarker nitric oxide. These organisms were combined with PEGDA and glycerol (not shown) to afford aqueous resins for 3D printing. b) A light-based 3D printer was used to fabricate ELMs of c) arbitrary geometries including stent form factors. d) Graphical scheme showing the treatment of gut inflammation by a 3D printed ELM device.

Results and Discussion

Engineered EcN can secrete the anti-inflammatory compounds β-carboline and tryptamine

Developing a living theranostic requires both a functional material as well as an engineered organism. For this work, EcN was selected as the host strain, owing to its history of safety in clinical settings³⁴ as well as its compatibility with many genetic tools developed for engineering laboratory strains of E. coli³⁵. We selected two _L-tryptophan-derived bacterial metabolites, tryptamine and 1-acetyl-3-carboxyl-β-carboline (referred hereafter as β-carboline for brevity), as candidates for IBD therapies due to their demonstrated anti-inflammatory activity in in vitro macrophage models^32,33. Both molecules require only a singular heterologous enzyme for production from _L-tryptophan, thus facilitating the creation of EcN bioproduction strains without the need for substantial pathway engineering. Beyond pathway complementation, we also explored boosting intracellular _L-tryptophan substrate pools through traditional aromatic amino acid biosynthetic pathway engineering^36,37,38 as a means to increase production (see Supplementary Information, Supplementary Fig. 1). As this work was a demonstration of a proof-of-concept system for inflammation sensing and response, a comprehensive assessment of production propensity compared to other probiotic hosts, such as lactic acid bacteria and S. boulardii was not carried out. However, given that the target products were derived from _L-tryptophan, a bacteria chassis felt most appropriate given that engineered E. coli strains have been shown to produce over an order of magnitude more _L-tryptophan than engineered yeast strains in laboratory settings³⁹, likely due to less stringent native regulation over the core metabolic pathway⁴⁰. As the genetic toolkits for gram positive lactic acid bacteria are less developed than those applicable for EcN^41,42, EcN made an ideal host for this purpose given speed of strain prototyping.

To create tryptamine-producing and β-carboline-producing EcN, we expressed tryptophan decarboxylase (TDC) from the organism Ruminococcus gnavus⁴³ and marinacarboline biosynthetic enzyme McbB from Marinactinospora thermotolerans SCSIO 00652⁴⁴ (which carries out a pictet spingler reaction on tryptophan to produce the β-carboline product), respectively. To provide a basis for comparison, we crafted two EcN strains for tryptamine overproduction: one containing the full suite of _L-tryptophan accumulation genetic alterations, as well as the TDC and one containing just the TDC. TDC was expressed under an IPTG-inducible promoter so that production was decoupled from the growth phase for this initial strain prototyping stage. Interestingly, the wild-type EcN strain having only the TDC overexpressed performed similarly to, if not better than, the strain with the more extensive engineering (Supplementary Fig. 2). This result could potentially indicate that the presence of a highly active _L-tryptophan-specific enzyme is sufficient to drive native metabolic flux towards higher _L-tryptophan production without a need for additional genetic engineering. In light of this result, we elected to use the EcN WT strain background with only heterologous enzyme overexpression for all additional strains in this work. The final selected tryptamine-producer was capable of producing 147.6 ± 15.3 mg/L tryptamine in 24 h at 37 °C (Supplementary Fig. 2).

To produce β-carboline, we overexpressed McbB in WT EcN. Interestingly, the strain with constitutive McbB expression produced significantly more β-carboline than the strain with inducible expression (Supplementary Fig. 2). This may be due to the fact that McbB catalyzes this reaction in EcN from _L-tryptophan and glycolysis byproduct methylglyoxal⁴⁵, which is toxic to bacteria at high concentration and likely primarily produced during growth stages⁴⁶. It should be noted that the IC₅₀ for this β-carboline with regard to anti-inflammatory activity is approximately 1 μM (0.254 mg/L)³², which is multiple orders of magnitude lower than baseline production by either the inducible or constitutive strains, which were capable of producing 16.0 ± 2.8 mg/L and 91.0 ± 9.6 mg/L β-carboline in 24 h at 37 °C, respectively (Supplementary Fig. 2).

To test the bioproduction performance of these cells in situ for various IBD applications, we sought to evaluate the production stability of these cells in a simulated gut environment. In this test, we find that free EcN cells demonstrated robust production in both fasted state simulated colonic fluid (FaSSCoF) and fed state simulated intestinal fluid (FeSSCoF) (Supplementary Fig. 3). In the case of β-carboline production in FeSSCoF, a dynamic increase in production was observed from day 2 to day 3. The exact mechanism of this boost is not clear but could indicate that the EcN cells went through an initial adjustment phase to the media composition and then began to enter a phase of more robust production. Further study with longer timeframes of testing is needed to confirm this phenomenon. This sustained production was also present in free EcN cells under oscillatory conditions starting from either the fed or fasted state (Supplementary Fig. 4). The production from EcN also remained high in the fasted state simulated intestinal fluid (FaSSIF) however was minimal in the fed state simulated intestinal fluid (FeSSIF) (Supplementary Fig. 5). This phenomenon is likely due to the low pH (~5) and high concentration of bile salts in FeSSIF. When cultured under oscillating FaSSIF and FeSSIF conditions (better simulate an environment of the intestine), the EcN strains regained their production capacity during the fasted state, even after enduring a full day in the fed state (Supplementary Fig. 6). This ability to endure extended periods in FeSSIF highlights the overall robustness of the engineered EcN and potential for deployment in small intestinal conditions, e.g. Crohn’s disease. Taken together, these two strains allow for anti-inflammatory production and serve as a building block toward responsive production described further below.

Production of therapeutic compounds from ELMs under simulated colonic conditions

We selected a light-based 3D printing process (vat photopolymerization) fabricate ELMs with arbitrary 3D form factors based on a computer aided design (CAD) model^31,47,48. Specifically, ELM resins that are based on PEGDA with glycerol were shown to be mechanically tough and resistant to fracture from equilibrium swelling³¹. Thus, these materials were chosen for incorporating the engineered EcN strains for therapeutic delivery.

We envision that the site for the administration of these devices is the small intestine or colon, depending upon the site of inflammation and therefore evaluated the stability and therapeutic performance of these devices in vitro using conditions that simulate intestinal and colonic environments. Specifically, the stability and therapeutic production from ELM devices were characterized in four different biorelevant media that reflect the fasted and fed states of both the intestine⁴⁹ and colon^50,51. To do so, we fabricated the EcN-laden ELMs in a medical stent shape using a stereolithographic apparatus (SLA) 3D printer (Fig. 1). The size of each stent was as follows: x = 15.00 mm, y = 15.00 mm, and z = 30.60 mm. The resin volume used during the 3D printing process was approximately 1 mL, and the total number of layers was 306. The CAD design of the stent is shown in Supplementary Fig. 7.

The ELM devices maintained durable production of both tryptamine and β-carboline in both fed and fasted state colonic fluid for at least 7 days, in which a minimum of ~50 μg/mL β-carboline and ~150 μg/mL tryptamine was consistently produced (Fig. 2b, c).

Figure 2.

Therapeutic performance of 3D printed ELMs in simulated colonic fluids (FaSSCoF and FeSSCoF). a) Composition of simulated intestinal fluids fasted and fed states. Bioproduction of b) of tryptamine and c) β-carboline from ELM constructs in simulated colonic fluids over a 7-d period. ELMs that produced d) tryptamine or e) β-carboline were cycled between FaSSCoF and FeSSCoF conditions and the production of therapeutic was tracked over time. Data presented as mean ± SD; n=3 biological replicates.

Further, we tested robustness by oscillating back and forth between fed and fasted states every 24 h and likewise found that production of each anti-inflammatory remained roughly constant, above 300 mg/L and 60 mg/L tryptamine and β-carboline, respectively (Fig. 2d, e). As with the free cells, these concentrations exceed the required levels of each respective small molecule for treating inflammation, as determined from in vitro macrophage models^52,53 showcasing strong potential for in vivo performance of these ELMs for extended periods in the colon.

Production of therapeutic compounds from ELMs under simulated intestinal conditions

We next tested the production of tryptamine and β-carboline from the engineered EcN strains in simulated intestinal fluid, to investigate the potential additional use of these materials for treatment of inflammation involving the small intestine such as with Crohn’s Disease⁴. For the case of ELMs, tryptamine production became detectable after 3 days of culturing in FaSSIF, while no tryptamine was observed in FeSSIF in these samples (Fig. 3b). β-carboline production started after 1 day of culturing in FaSSIF and continued for 7 days (Fig. 3c). Notably, starting from day 2, a substantial amount of β-carboline was also identified in ELM devices cultured in FeSSIF (Fig. 3c). This inconsistent production in FeSSIF is again likely due to the pH and bile salt conditions associated with the fed state.

Figure 3.

Therapeutic performance of 3D printed ELMs in simulated intestinal fluids (FaSSIF and FeSSIF). a) Composition of simulated intestinal fluids for the fasted and fed states. Bioproduction of b) tryptamine and c) β-carboline from ELM constructs in simulated intestinal fluids over a 7-d period. Mass changes recorded over time for ELM constructs producing d) tryptamine-producing and e) β-carboline in simulated intestinal fluids over a 30-d period. Data presented as mean ± SD; n=3 biological replicates.

Importantly, continuous tryptamine production in FaSSIF and β-carboline production in both FaSSIF and FeSSIF by the ELM device was observed for over 30 days (Supplementary Fig. 8). Moreover, the ELMs maintained their initial mass throughout at least 30 days even in these harsh simulated intestinal conditions (Fig. 3 d, e, Supplementary Fig. 9). The milder colonic environment is thus expected to have an even lesser effect on the mechanical integrity of the ELMs. Thus, while the small intestine was not the primary target for this device, this ELM is certainly capable of surviving in small intestinal conditions and could take advantage of fasting periods as a time for anti-inflammatory production by EcN.

Confocal imaging highlights the protective feature of extracellular polymer networks in the harsh intestinal environment

The poor survival rate of most probiotic organisms and overall colonization resistance in the gut limit their deployment and use as therapeutic agents^5,6,29,30,54. Probiotic-ELM devices that can withstand harsh environments and not have to colonize against the colonic microbiome thus have an advantage as a platform for continuous probiotic release and therapeutic delivery⁵⁵. Thus, we evaluated the cell release and resulting viability of the released cells from ELM devices cultured in FaSSIF and FeSSIF, as these were the harshest conditions identified in the results above. These results indicated that EcN cells were released from ELM device into FaSSIF media and maintained viability as seen by their ability to colonize when seeded on agar plates (Supplementary Fig. 10). On the other hand, no cell growth was observed in FeSSIF samples (Supplementary Fig. 11). The FeSSIF represents the condition in the intestine after a meal, where there is an elevation in surfactants, specifically bile salts. While probiotic organisms exhibit bile salt hydrolase activity⁵⁶, they often possess low overall tolerance to elevated bile salt concentrations^29,30,57. On the other hand, the lower bile salt concentrations in FaSSCoF and FeSSCoF provide a milder environment for the released cells, which grow on agar plates (Supplementary Fig. 12), and more viable cells were observed under the microscope (Supplementary Fig. 13).

Interestingly, for the β-carboline producing ELMs, we still observed continuous production from the ELMs for over 7 days in FeSSIF (Fig. 3c), which highlights the protective role of the polymer matrix for the EcN cells that remain embedded against adverse FeSSIF conditions. To confirm the presence of viable cells within the ELM scaffold under these conditions, β-carboline producing EcN cells were imaged in an ELM matrix after 2-day incubation in FeSSIF (Fig. 4a, b, Supplementary Fig. 14 c, d). Confocal imaging demonstrates that EcN cells form clusters in the ELM polymer matrix with the inner most clusters showing the presence of live cells(Fig. 4b). This protective effect was more pronounced for the β-carboline producing EcN, compared to the tryptamine-producing EcN. As noted earlier, β-carboline production entails consumption of the toxic metabolic byproduct methylglyoxal⁴⁵, which could lower the strain’s baseline metabolic stress and thus make it more prepared to withstand the harsher intestinal conditions. The lower baseline metabolic stress combined with the protective nature of the polymer matrix could explain why the β-carboline producing ELMs where able to survive extended periods in FeSSIF (Fig. 3c) while the tryptamine producing ELMs could not (Fig. 3b).

Figure 4.

Live/dead cell staining of EcN cells in an ELM construct which was cultured in FeSSIF for 3-c. a) Schematic representation of z-stack image direction of ELM construct that was in the shape of a disc (3 mm in depth). b) 3D fluorescence visualization of z-stack images of EcN cells ELM construct. Live and dead EcN cells were imaged under Nikon A1R HD25 laser scanning confocal microscopy. Red cells represent dead cells whereas green cells represent live cells.

ELM Devices that secrete anti-inflammatory agents in response to nitric oxide as a biomarker for inflammation

IBD typically presents as a chronic condition that requires repeated treatments over time of inflammatory episodes or flares. Thus, we next sought to develop a second generation EcN that can be incorporated into the matrix to develop a sense-and-respond ELM device capable of producing tryptamine or β-carboline only when inflammation was present (Fig. 5 a, c). To do so, nitric oxide (NO), a well characterized transient biomarker of IBD²², was selected as the inflammation marker to take advantage of functional sensors that have been built in bacteria^58,59. Specifically, the norRVW operon, native to bacteria, consists of the pNorV promoter as well as its regulatory transcription factor NorR, which binds pNorV in the presence of NO to activate transcription⁴⁹. The NorR promoter is known to be highly specific for NO^60,61 and has previously been demonstrated to not be influenced by other reactive nitrogen species (NO_x) as well as to be functional in inflamed mouse ileum explants⁶² with a sensing range of 30-100uM. Prior work has demonstrated an enhanced induction amplitude in the presence of physiologically relevant concentrations of NO when these synthetic circuits are designed based on a positive feedback loop (i.e. with NorR transcription immediately downstream of target genes and thus linked to pNorV activation, as opposed to constitutive NorR expression)⁵⁹. Additionally, the removal of one integration host factor (IHF) binding site of the native NorR promoter to yield a pNorVβ has been shown to significantly improve maximum signal output, leading to a dynamic range of roughly 6.7 folds change over the operational range of 1.5uM-500uM NO in a rich media⁵⁹. Taking these observations as our base design, we validated performance of the circuit with GFP as an output using both defined and gut-relevant media. An appreciable GFP fold change of 2.5-4x was observed over the medically relevant range of 10 – 100 μM NO in all media except FeSSIF (Supplemental Fig. 15), as consistent with observations above for this media.

Figure 5.

Therapeutic production performance of NO-responsive ELMs in simulated intestinal and colonic fluids. a) Bioproduction modules of metabolically engineered EcN strain, EcN pNorVβ-TDC, which produced tryptamine in response to NO. b) Bioproduction of tryptamine by the ELM constructs in simulated colonic fluids with different nitric oxide concentrations. c) Bioproduction modules of metabolically engineered EcN strain, EcN pNorVβ-McbB, which produced β-carboline in response to nitric oxide. d) Bioproduction of 1-acetyl-3-carboxyl β-carboline by ELM constructs in simulated simulated colonic fluids. Data presented as mean ± SD; n=3 biological replicates.

Following the successful validation of the NO-sensing circuit using a simple fluorescent output, we created circuits capable of sensing NO and producing anti-inflammatory molecules in response. The resulting strains with integrated circuits for tryptamine and β-carboline will hereafter be denoted as EcN pNorVβ-TDC and EcN pNorVβ-McbB, respectively (Fig. 5 a, c). The medically relevant range for NO thresholds for IBD includes >17.4 μM for active ulcerative colitis and >14.0 μM for active Crohn’s Disease⁶³. Thus, we tested production in the presence of 15-100 μM NO to test the ability of engineered strains to sense and respond differentially to low and high biomarker levels for active IBD, which can even exceed 100 μM in extreme cases⁶⁴. Both the ELM device and free engineered EcN cells were able to differentially sense and respond to NO levels ranging from 15-100 μM (Fig 5. b, d; Supplementary Fig. 16). The production of anti-inflammatory compounds correlated with increasing NO levels in all media except FeSSIF, matching the GFP results above.

In vitro models demonstrate biocompatibility and efficacy of ELM Devices

We next tested the in vitro biocompatibility of engineered EcN using Caco-2 cells, a widely used model cell line for intestinal epithelium due to their ability to differentiate into enterocyte-like cells with tight junctions and brush border enzyme expression⁶⁵. They are commonly used to assess barrier integrity, cytokine secretion, and drug transport in GI disease models. Cytotoxicity of engineered EcN was assessed via the alamarBlue assay⁶⁶. Engineered EcN showed minimal cytoxicity effects on Caco-2 cells, matching previous accounts suggesting an excellent safety profile for EcN (Supplementary Fig. 17). This result also demonstrated that the resulting titers for the anti-inflammatory agents were within a range safe for the intestinal epithelium.

Next, we demonstrated the capacity of these responsive EcN strains to both prevent and treat inflammation using an in vitro murine macrophage model (Lipopolysaccharide (LPS)-challenged RAW264.7 cells; Fig. 6a). The RAW 264.7 cell line is frequently used to model macrophage-mediated inflammation in IBD, as these cells respond to pro-inflammatory stimuli such as LPS by producing nitric oxide and cytokines, enabling quantification of immune activation⁶⁷. In the presence of LPS, inflammation is induced in RAW264.7 cells. These cells produce sufficient quantities of the inflammatory biomarker NO to activate the NO-responsive circuits used in our strains and to enable readout using the Griess Assay⁶⁸.

Figure 6.

Engineered EcN exhibits anti-inflammatory effects in LPS-challenged murine macrophages. a, b) Schematic workflow for evaluating NO produced by RAW264.7 cells a) without or b) with treatment. c) Evaluation of NO production by RAW264.7 cells 48 h post various treatments, administered either simultaneously with 1 μg/ml LPS challenge or 24 h prior to treatment. Statistical analysis: one-way ANOVA followed by Welch test. (*p > 0.05). ns: no significant difference. Data presented as mean ± SD; n=3 biological replicates.

To evaluate the potential for our engineered EcN to prevent inflammation, RAW264.7 cells were simultaneously treated with 1 μg/ml LPS and appropriate treatments (Fig. 6, Supplementary Fig. 18). As a basis of comparison, we also exogenously treated RAW264.7 cells with mesalazine, an FDA-approved small molecule drug for first-line treatment of IBD⁶⁹. For prevention of inflammation, both engineered EcN strains outperformed mesalazine (p = 0.00017) with respect to reducing NO-production by the RAW2 cells (Fig. 6c). In addition, there was no significant difference (p = 0.28316) in NO reduction levels between the tryptamine or β-carboline producing EcN treatments.

To evaluate the potential for treating active inflammation, RAW264.7 cells were dosed with LPS for 24 h before the start of any treatment. For the case of treating active inflammation, the engineered EcN brought NO levels down to that of the negative control (p = 0.7310) and significantly outperformed mesalazine (p = 0.00007) (Fig. 6c). These results highlight not only the prophylactic and therapeutic potential of both living therapeutics and anti-inflammatory agents. The robustness of our engineered EcN to combat inflammation in macrophages across varying degrees of active inflammation (controlled by length of LPS induction time prior to treatments), highlights their potential for deployment during both active and remissive phases of IBD.

Finally, we tested the capacity of EcN ELMs to combat inflammation in intestinal epithelial cells. Over the past decades, the Caco-2 cell line derived from human colon adenocarcinoma has served as a valuable model for studying toxicity, absorption, and metabolism in the context of the intestinal barrier. Through spontaneous morphological and biochemical differentiation during routine in vitro culture, Caco-2 cells develop features akin to intestinal cells, including a microvilli structure, brush border, tight junctions, and the ability to secrete hydrolases and synthesize carrier transport systems for various substances like sugar, amino acids, and drugs^70,71,72. As Caco-2 cells do not produce sufficient NO to fall in the medically relevant range for IBD⁷³, EcN or EcN-laden ELMs were pre-cultured in FeSSCoF or FaSSCoF with or without the addition of 15 μM NO (24 h) before exposure to the Caco-2 cells. Supernatants derived from these cultures were added to the apical compartment of a 96 trans well plate containing Caco-2 cells prior to induction of inflammation on the apical side with 0.1 μg/ml LPS and on the basolateral side with a pro-inflammatory cocktail (0.025 ug/ml IL1β, 0.05 μg/ml TNFα, and 0.05 μg/ml IFNγ)⁷⁴. In this setup, the apical compartment is representative of the intestinal lumen whereas the basolateral side represents cells facing underlying tissue. After 24 h incubation, samples were collected from both apical and basolateral sides and IL-6 and IL-8 levels were measured via ELISA assays (Fig. 7a). The ELMs were tested for efficacy as a preventative treatment for inflammation (similar to the setup in the 0 h treatment group in Fig. 6b), representing a solution to maintain remission for patients with IBD. Thus, a potential future implementation of an ELM device could include an ELM implanted during a time of remission to secrete therapeutic compounds in response to any sudden onset of inflammation. Both tryptamine and β-carboline produced by the EcN ELMs (Fig. 7 b-e) or free EcN cells (Supplementary Fig. 19) under FeSSCoF and FaSSCoF conditions in the presence of NO led to a reduction in the secretion of pro-inflammatory cytokines IL-6 and IL-8. In terms of treatment efficacy, there was no significant difference in IL-6 reduction between tryptamine and β-carboline which were produced under FaSSCoF conditions with 15 μM NO, on either the apical (p = 0.2258) or basal (p = 0.4655) side (Fig. 7b). A similar trend was observed under FeSSCoF conditions, with no significant difference in IL-6 reduction on the apical side (p = 0.1679) (Fig. 7c). For IL-8 reduction, β-carboline showed better performance compared to tryptamine, with significantly greater reduction observed on the apical side when these compounds were produced under both FaSSCoF (p = 0.0004) (Fig. 7d) and FeSSCoF (p = 0.0031) (Fig. 7e) conditions in the presence of 15 μM NO.

Figure 7.

Reduction in secretion of pro-inflammatory cytokines, IL-6, and IL-8 by Caco-2 cells upon inflammation challenge with therapeutics produced by NO-responsive ELM constructs. a) Schematic representation of the experimental workflow. b, c) Human IL-6 production levels by Caco-2 cells post-inflammation induction treated with tryptamine or β-carboline, which were produced by NO-sensing ELMs (with 15 μM NO) in b) FaSSCoF or c) FeSSCoF. d, e) Human IL-8 production levels by Caco-2 cells post-inflammation induction treated with tryptamine or β-carboline, which were produced by NO-sensing ELMs (with 15 μM NO) in d) FaSSCoF and e) FeSSCoF. Statistical analysis: one-way ANOVA followed by Welch test. (*p > 0.05). ns: no significant difference. Data presented as mean ± SD; n=3 biological replicates.

Furthermore, we observed an NO-dependent decrease in IL-8 levels in response to β-carboline produced by the EcN ELM under FaSSCoF conditions on the apical side. This trend was quantified as a 2-, 3-, and 5-fold reduction in IL-8 levels with β-carboline production in the presence of 0, 15, and 100 μM NO, respectively (Supplementary Fig. 20). Thus, use of the theranostic ELM device as an in situ anti-inflammatory production platform enables a degree of control over anti-inflammatory response and proper dosing in a manner that tracks inflammation biomarkers that are present. Interestingly, this dose-dependent response was only evident in the apical compartment, representative of the intestinal lumen. This could potentially indicate that minimal secretion of our target anti-inflammatory compounds is sufficient to yield robust decrease in inflammation in the basolateral side, which faces underlying tissue. The apical side could then require higher levels of the target compounds, secreted at higher NO concentration levels, to yield similar robust decreases in inflammation levels. Additionally, permeability studies (Supplementary Fig. 21) carried out on both tryptamine and β-carboline secreted from free cells and ELMs demonstrated high permeability (Papp > 10x10⁻⁶ cm/s) (Supplementary Table 4) and Efflux Ratio < 2 (Supplemental Fig. 22), indicating likelihood for in vivo absorption⁷⁵. Taken together, the results from multiple in vitro inflammation models highlight the potential for ELMs to be used for long-term inflammation treatment in which a responsive, just-in-time amount of anti-inflammatory is delivered during the onset of an inflammatory event, enabling patients to remain in remission.

Conclusion

Establishing ELMs with engineered probiotics into medically relevant form factors can enable the in situ production of target therapeutics at the site of disease. In this study, we designed and fabricated 3D-printed ELM theranostic devices capable of sensing inflammatory biomarkers and responding by secreting two different anti-inflammatory small molecules, tryptamine and β-carboline. The ELMs were able to robustly produce target compounds in a range of media simulating intestinal and colonic environments, with production levels titrated to medically relevant concentrations of NO for active vs inactive IBD. Furthermore, our ELM device demonstrated safety and efficacy of treating inflammation with respect to multiple well-established in vitro models for IBD utilizing immune and intestinal epithelial cells, demonstrating the potential for safety and efficacy in vivo.

Future studies incorporating co-culture systems of intestinal epithelial cells with mucus-producing goblet cells, immune cells, or primary intestinal organoids for further safety and efficacy testing can provide more accurate modeling of the intestinal environment during active inflammation^76,77. Additional future work should certainly validate efficacy of this ELM platform in murine IBD models, as well as propensity for long-term delivery of gastrointestinal implanted stent-shaped ELMs in pig models. Looking forward, we envision this ELM device as a closed-loop therapeutic modality to help establish and maintain remission in IBD patients over extended time periods, with minimal systemic exposure, enabling a paradigm shift in IBD management.

Materials and Methods

Bacterial strains and plasmids

All genetic parts, plasmids, and strains used in this study are listed in Supplementary Tables 1-3. Routine cloning and plasmid propagations were carried out in E. coli DH10β. E. coli Nissle 1917 was obtained from Ardeypharm (Germany) and used for de novo production and all in vitro experiments, in which strains were transformed with genetic circuits and/or biosynthetic pathways built on plasmids. For genetic integration, strains were constructed via λ-RED recombination following standard protocols⁷⁸. All plasmids were constructed via combining PCR fragments which were generated via Q5 High Fidelity DNA Polymerase (IDT) using Gibson Assembly⁷⁹. Oligonucleotide primers for PCR amplification were purchased from Integrated DNA Technologies (Coralville, IA), with gBlocks manufactured by IDT or Twist Bioscience. Bacterial transformations were carried out via electroporation (2 mm Electroporation Cuvettes, BioExpress) with a BioRad Genepulser Xcell at 2.5 kV. Transformants were selected on Lysogeny broth agar plates which contained appropriate antibiotics. All plasmid and integration sequences in transformed strains were verified via Sanger Sequencing.

Culture Media and Conditions for Bacterial Cells

Bacterial cells were routinely cultured in Lysogeny broth (LB) medium containing yeast extract (5 g L⁻¹), tryptone (10 g L⁻¹), and NaCl (10 g L⁻¹) for preparation of inoculants and cell propagation with supplementation of appropriate antibiotics as needed (50 μg/ml Kanamycin and/or 30 μg/ml Chloramphenicol and/or 100 μg/ml Ampicillin). For Nitric Oxide sensor kinetic testing and specified production experiments, a defined M9G+CAA media was made according to the following formula: 1X M9 salts, 1 g/L casamino acids, 5 g/L glucose, 20mM MgSO₄, 0.1mM CaCl₂ with supplementation of appropriate antibiotics were added as needed. Biorelevant fed and fasted state simulated intestine and colon kits were used to prepare intestinal and colonic fluids^70,71,72, according to the procedure provided from manufacturer. Simulated intestinal media and simulated colonic media was prepared fresh for each experiment and used within 48h. For testing of baseline anti-inflammatory production, engineered EcN were grown in culture tubes overnight at 37°C in LB supplemented with appropriate antibiotics. The following morning, cells were diluted to an OD₆₀₀ of 0.1 and resuspended in appropriate testing media in 96 deep well plates using 1ml working volume per well (pure M9G CAA, pure LB, 50/50 mixture of LB and FeSSIF, 50/50 mixture of LB and FaSSIF, 50/50 mixture of LB and FeSSCoF, or 50/50 mixture of LB and FaSSCoF). 50/50 mixtures were used for gut simulatory conditions as the Biorelevant media does not contain all essential nutrients for microbes to grow. For testing in gut simulation media, media was refreshed daily. For oscillatory condition testing, fed and fasted variations of SSIF or SSCoF were oscillated on a daily basis. Supernatant samples were taken daily for analysis of metabolites via HPLC.

SLA 3D Printing of ELM Devices and Culturing Conditions

To prepare the resin, 40 wt% PEGDA and 5 wt% glycerol was dissolved in an appropriate cell culture medium. Then, 1 mL of a cell suspension containing 1×10⁹ cells/mL was introduced into 20 grams of the resin formulation. 1 wt% of lithium phenyl-2,4,6-trimethylbenzoylphosphinate was used as a photo initiator. The constructs were 3D printed in the Open Mode using a Formlabs Form 2 printer with a layer height of 100 μm. CAD models were designed in Autodesk Fusion or downloaded from Thingiverse. 3D-pinted probiotic-ELMs placed in culture media supplements simulated fluids (1:1) and their bioproduction performance was evaluated. Simulated fluids were refreshed every day during 7 d of culturing. Supernatant samples were taken daily for analysis of metabolites via LCMS.

Quantification of Anti-Inflammatory Metabolites

Metabolite quantification from supernatant samples was performed using Bruker Esquire LC-Ion Trap Mass Spectrometer with Agilent HPLC. In each day samples were collected from cultured media, centrifuged at 4400 rpm for 10 min, filtered with 0.2-μm nylon syringe filters (Wheaton Science), and mixed with EtOH 1:1 v/v ratio, then analyzed. The mass spectrometer was operated in positive mode, the scan range was between 50.00 m/z to 1100.00 m/z, skim1 was 15.0 volts and capillary exit was 55.0 volts. In HPLC analysis Zorbax Eclipse Plus C18 column (3.0 × 150 mm, 3.5 μm; Agilent) was used. The mobile phase consisted of 1% (v/v) acetic acid in water or acetonitrile. Analyte detection was performed at a wavelength of 280 nm using a flow rate of 0.3 mL min⁻¹ with a column temperature of 25°C.

Tryptamine was quantified through reference with an analytical standard. As no analytical reference standard is available for 1-acetyl-3-carboxyl-β-carboline, we produced 1-acetyl-3-carboxyl-β-carboline for use as an HPLC reference via in vitro enzymatic production. E. coli BL21(DE3) transformed with a plasmid for McbB production (Supplementary Table 2) were grown overnight in LB supplemented with appropriate antibiotics at 37°C. Cells were diluted 2000X into a 500 mL baffled flask and grown at 37°C to an OD_600nm of 0.8. Protein expression was induced via addition of 0.2 mM IPTG, with cells subsequently grown for 24 h at 20°C. After completion of protein expression, cells were harvested via centrifugation at 4,000 g for 20 minutes. The cell pellets were then resuspended in 25 mL of PBS, pH 7.0, containing 10 mM imidazole, 1 g L⁻¹ lysozyme, and 5 μL of Pierce^™ Universal Nuclease (ThermoFisher, Waltham, MA). The mixture was incubated on a rocker for 30 min at 4°C. Cells were then lysed via sonication, with resulting lysate centrifuged at 14,000 g for 20 min. The resulting supernatant then contained soluble proteins. The His tagged protein was purified from the supernatant via a HisPur^™ Ni-NTA Resin (Thermo Fisher Scientific, Waltham, MA), which was used according to the manufacturer’s instruction. The resulting eluate was dialyzed with 3C protease, which was added to the dialysis cassette, into the appropriate buffer which was followed by size exclusion chromatography. Purified protein was stored in 20 mM Tris (pH 7.5), 100 mM NaCl and 10 mM β-mercaptoethanol, with protein concentration determined via a Coomassie Plus Bradford Assay kit (Thermo Fisher Scientific), which was used according to the manufacturer’s instructions. Absorbance measurements for the assay were taken on a microplate reader (Tecan Infinite Pro). The melting temperature of McbB was analyzed via DSC and found to be 73°C. In vitro production of 1-acetyl-3-carboxyl-β-carboline was carried out as follows: Reactions were set up in a deep well plate with working volumes of 1 mL per well. 10 μg of purified enzyme was mixed with 0-500 μM _L-Tryptophan in a reaction buffer (50 mM Potassium Phosphate Buffer, pH 7.4) containing 10 mM methylglyoxal³⁸. Reactions were incubated at 50°C for 24 h, followed by analysis of supernatant via the HPLC method described above. Approximation of conversion value of 1-acetyl-3-carboxyl-β-carboline peak area to μM product formed was enabled by calculation of μM _L-Tryptophan consumed, as well as verification that no other visible peaks were formed on the HPLC during the reaction.

Live and dead cell analysis

Live and dead EcN cells were imaged under fluorescence microscopy (EVOS Fluorescent Imaging Microscopy) after staining with a Biotium Viability/Cytotoxicity staining kit for bacteria live and dead cells according to the procedure provided by the manufacturer.

Confocal Imagining

Nikon A1R HD25 laser scanning confocal microscope was performed to take cell images. 400 μL of probiotic-ELM resin was placed in a silicon mold with 3 mm depth and cured under 405 nm for 5 min. Then it was placed in 6 well plates and cultured in FeSSIF which was supplemented with LB media 1:1 ratio for 2 d. Before confocal imaging, culture media was removed from photocured samples, and samples were washed with PBS 3 times. Samples were stained with Biotium Viability/Cytotoxicity staining kit for bacteria live and dead cells according to the procedure provided by the manufacturer and incubated at room temperature in the dark for 30 minutes. The stained gels were then transferred to Ibidi μ-dish 35 mm high cell culture imaging dishes and imaged.

Nitric Oxide induction kinetic experiments

EcN cultures were grown overnight at 37°C in LB supplemented with appropriate antibiotics. Cultured were diluted into a 96 well plate with a starting OD₆₀₀ of 0.1 for each well and added to a total culture volume of 200 μL per well. Media formulation per well was any of the following: pure M9G CAA, 50/50 M9G CAA/FeSSIF, 50/50 M9G CAA/FaSSIF, 50/50 M9G CAA/FeSSCoF, 50/50 M9G CAA/FaSSCoF. A fresh inducer stock was added to each well at the desired concentration (0-1 mM DETA-NO). The kinetic experiments were carried out in a microplate reader (Tecan Infinite 200 Pro), in which the plate was maintained at 37°C. Measurement of GFP Mut2 levels (fixed gain = 75, exciting/emitting wavelength 475/510 nm) as well as Abs_600nm values for cell density were taken every 15 minutes over the course of 24h using a built-in kinetic function. Results of the experiments were analyzed and plotted using GraphPad Prism 10 software.

Mammalian Cell Lines and Culture Conditions

Human colon epithelial cells (Caco-2, ATCC HTB-37; human colorectal adenocarcinoma cells originally derived from a 72-year-old European male) were used as a model intestinal epithelial barrier function. These cells were cultured in Eagle's Minimum Essential Medium (EMEM) (EMEM, ATCC) supplemented with 10% fetal bovine serum (FBS, Biowest, France) and 1% penicillin/streptomycin (Sigma-Aldrich) for LPS inflammatory assays. Polarized Caco-2 cells were cultured in Dulbecco’s Modified Eagle’s Medium - low glucose (1 g/L) (Sigma) supplemented with (final concentrations): 0 % v/v Fetal Bovine Serum (Biowest), 2 mM L-glutamine (Lonza) and 100 μ/mL; 0.1 mg/mL Penicillin-Streptomycin for permeability and ELISA assays. Murine macrophages (RAW 264.7, ATCC TIB-71; derived from a male BALB/c mouse with Abelson leukemia) were used to model inflammatory immune responses relevant to IBD. RAW264.7 cells were cultured in DMEM 10% fetal bovine serum (FBS, Biowest, France) and 1% penicillin/streptomycin (Sigma-Aldrich).

For viability studies, Caco-2 cells were seeded in a 96-well microplate at a density of 3x10⁴ cells per well and grown for 24 h at 37°C. The media was replaced, and Caco-2 cells were cultured with or without engineered EcN at a concentration of 1x10⁸ cells/mL for 24h. For anti-inflammatory assays, RAW264.7 cells were seeded in a 96-well microplate at a density of 3x10⁴ cells per well and grown for 24h at 37°C. Media was replaced with 140 μl DMEM which lacked phenol red (to avoid interference with the colorimetric Griess Assay). For the case of simultaneous inflammation induction and treatment, the 140 μl of media with 1 μg/ml LPS (Sigma Chemical Co., St. Louis, MO, U.S.A) and the presence or absence of 1x10⁶ cells/ml engineered EcN or other treatments (mesalazine, tryptamine). Cells were cultured 48h at 37°C before performing the Griess Assay. For the case of inflammation induction before treatments, cells were cultured with 1 μg/ml LPS for 24 h, followed by the addition of treatments (1x10⁶ cells/ml engineered EcN, mesalazine, or tryptamine). Cells were cultured 48h at 37°C before performing the Griess Assay.

Cytotoxicity Studies

Cytotoxicity of engineered EcN strains against Caco-2 cells was assessed by measuring the metabolic activity of Caco-2 cells using an alamarBlue Assay (Thermo Fisher Scientific). Caco-2 cells were seeded in a 96 well microplate at a density of 3x10⁴ cells per well and grown for 24 h at 37°C. The media was replaced, and Caco-2 cells were cultured with or without engineered EcN at a concentration of 1x10⁸ cells/mL for 24 h. To carry out metabolic studies, the Caco-2 cells were rinsed with PBS and incubated at 37°C for 3 h with 10 μL alamarBlue reagent dissolved in 80 μL PBS. After incubation, 75 μL per well were analyzed on a plate reader (Tecan Infinite 200 Pro) by reading the absorbance at 570 nm. % Metabolic Activity was calculated by dividing absorbance values of each experimental condition by the control.

Anti-inflammatory In Vitro Assays

In vitro murine macrophage model:

Nitric oxide production by RAW264.7 cells was quantified using the Griess Assay to assess inflammation following LPS-induced inflammation. Briefly, 50 μL of supernatant was mixed with 50 μL Griess Reagent (Millipore Sigma G4410) and allowed to react for 15 min. Absorbance was read at 550 nm in a microplate reader (Tecan Infinite 200 Pro), with nitrite concentration calculated using a standard curve of sodium nitrite over the working range of the assay (1-100 μM).

In vitro Caco-2 cell model:

Anti-inflammatory cytokine production by Caco-2 cells was quantified by cytokine production, including IL-6 and IL-8. Polarized Caco-2 cells placed in 96 well plate was purchased from ReadyCells. Cells were differentiated for 21 days in 96 Transwell^® inserts with semiporous (0.4 μm) polystyrene membrane, resulting in an apical compartment and a basal compartment that mimic the intestinal lumen and blood circulation, respectively. After receiving the cell, shipment medium was replaced with Dulbecco’s Modified Eagle’s Medium - low glucose (1 g/L) (Sigma) supplemented with (final concentrations): 0 % v/v Fetal Bovine Serum (Biowest), 2 mM L-glutamine (Lonza) and 100 μ/mL; 0.1 mg/mL Penicillin-Streptomycin. After 3 days culturing, TEER values were measured prior to inflammation and permeability assays to ensure tight junction integrity and monolayer maturation (TEER > 500 Ω_x cm²). ELM constructs were pre-cultivated in colonic fluid from a fed or fasted state, with or without the presence of NO, for 24 hours. Supernatants, derived from these cultures, were applied to the apical compartment of a 96-transwell plate containing Caco-2 cells before inducing inflammation on the apical side with 0.1 μg/ml LPS and on the basolateral side with a pro-inflammatory cocktail (0.025 μg/ml IL1β, 0.05 μg/ml TNFα, and 0.05 μg/ml IFNγ). After 24h incubation, samples were collected from apical sides and IL-8 and IL-6 levels were measured via ELISA assays (Sigma).

Statistical analysis

Statistical differences were determined using one-way ANOVA and the Welch test (p < 0.05) in Igor Pro (Version 8.04). Means and standard deviations (± S.D.) of replicates were calculated using Excel.

Supplementary Material

Supporting Information is available from the Wiley Online Library or from the author.

Data availability statement

The data reported in this manuscript are available from the corresponding authors upon reasonable request.