The Generation of an ocular anti-inflammatory biotherapeutic to enhance wound healing.

Abstract Microbes exist at and colonize mucosal surfaces striking a balance with the host immune system, so that these microbes can thrive on host tissues without causing pathology. Because of this, mucosal barrier-colonizing bacteria can be leveraged to act as long-term delivery vehicles for naturally derived therapeutics. Here, we use a mouse model of corneal wound healing to show that the eye-colonizing bacterium, Corynebacterium mastitidis (C. mast) can be engineered to produce and secrete bioactive murine anti-inflammatory interleukin (mIL)-10. Specifically, we used transposon mutagenesis to identify a native C. mast-specific secretion signal that was used to direct C. mast to secrete mIL-10. Mini-transposons were generated to deliver secretion capable mIL-10 to the bacterial genome. After screening, two isolates were identified that can: 1) colonize the eye, 2) produce and secrete mIL-10, and 3) enhance wound healing in an IL-10-dependent manner. This proof of concept illustrates that eye-colonizing bacteria can be engineered to deliver therapeutics to the ocular surface for the alleviation of ocular surface disease(s).

Title: The Generation of an ocular anti-inflammatory biotherapeutic to enhance wound healing.

Introduction
Corneal wounds remain a serious concern for ocular surface health as over 1 million Americans are afflicted with corneal wounds annually and they account for 3% of all visits to the emergency room 1 .During wound healing, the eye is susceptible to infection and inflammatory damage, which can lead to other more serious issues like nerve damage 2,3 , opacification, and loss of visual acuity 4,5 .Therefore, increasing the rate of wound healing is paramount in preventing long-term disease.
Corneal wound healing relies on an appropriately mounted immune response that encourages healing and limits inflammation.After corneal wound healing, the upregulation of inflammatory cytokines, interleukin (IL)-1b, IL-6, and TNFa, suppresses STAT3 activity, which can impair the proliferation of corneal epithelium and disrupts normal healing of the cornea 6,7 .
Alternatively, IL-17 from cornea-infiltrating ɣδ T cells can activate STAT3, which enhances wound healing 8,9 .Similarly, IL-10 can also activate STAT3 to encourage proliferation and migration of epithelial cells both in the cornea 7 and other barrier surfaces throughout the body 7,10,11 .IL-10 also directly inhibit inflammation by suppressing cytokine production and proliferation of inflammatory myeloid cells.Furthermore, mIL-10 directly interfere with the activation of T cells that require CD28 co-stimulation [12][13][14] and suppresses T cells that exacerbate inflammation in many ocular surface diseases like infectious keratitis and Dry Eye Disease 2,15,16 Virtually all barrier sites have a resident microbiome that directly interacts with the host and fosters tissue barrier function 17 .Microbial members of the microbiome have adapted mechanisms that allow them to thrive at the sites of colonization.These unique adaptations make them attractive candidates to continually deliver naturally derived therapeutics to sites of interest.Indeed, scientists have begun engineering components of the intestinal microbiome to modulate metabolism 18 as well as cytokine production to regulate immunopathology associated with mouse models of colitis [19][20][21] and clinical trials as a therapeutic for Crohn's disease 22 .When IL-10 was expressed by Lactococcus lactis, a gut bacterium, CD4+ T cell-mediated colitis was reduced 21 ; however, the direct effects of mIL-10 signaling in this model are still unclear.On the ocular surface, therapeutics are continually washed away or diluted by the flow of tears 4,23 , so engineering eye-colonizing bacteria that can deliver therapeutics to the ocular surface is a viable avenue towards reducing the burden of applying therapies like eye drops multiple times per day.

Bacterial culture and genetic constructs
C. mastitidis AS1was isolated from ocular swabs and cultured with LB media containing 0.5% Tween-80 grown aerobically at 37°C.For T-cell culture experiments cell culture media was used consisting of DMEM with 10% FBS; 1% NEAA, glutamax, Hepes, and sodium pyruvate.Bacteria added to T-cell culture used the same cell culture media with the addition 0.5% corn oil and 0.05% w/v casamino acids.The phosphatase Z gene was isolated via PCR from DNA extracted from Enterococcus faecalis using primers indicated in Table 1.The AphA3 kanamycin resistance gene construct was isolated froma previously created plasmid, pTony3 25 , and the erythromycin resistance gene cassette was created by using the same promoter attached to a codon harmonized erythromycin resistance gene of C. striatum.Codon harmonized mIL-10 was ordered as a gBlock Gene Fragment (IDT, USA).Codon harmonization was performed using the online Galaxy software, Codon Harmonizer 26 .All newly created sequences can be found at accession # PP471979-PP471980 and all primers used are detailed in Table 1.

Mice
Female C57BL/6 mice purchased from Jackson Laboratories (Bar Harbor, ME, USA) were housed in the Animal Resource Facility at the University of Pittsburgh Medical Center (Pittsburgh, PA, USA) and colonized with C. mast at 6 weeks of age in all experiments according to our previous inoculation schemes.Briefly, mice were ocularly inoculated with 1x10 6 CFU of C. mast every other day for a total of 3 inoculations.Two weeks after the final inoculation, the corneas were wounded as described below.Experimental procedures were reviewed and approved by the University of Pittsburgh Institutional Animal Care and Use Committee and the use of animals was in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research.

Generation of the secretion cassette.
Secretion signals for C. mast were identified by creating a phosphatase (PhoZ) transposon library.Using Overlap extension PCR (OEP), promoterless phoZ lacking its secretion signal from Enterococcus faecalis was fused with an AphA3 kanamycin resistance gene from a previously created plasmid, pTony3 25 ,to create a transposome that was inserted randomly into the wild-type C.mast genome using EZ-Tn5™(Lucigen).Positive mutants secreting PhoZ were selected for on kanamycin plates (100 µg/ml) supplemented with BCIP (5bromo-4-chloro-3'-indolyphosphate p-toluidine salt) (25µg/ml) that turns colonies blue when phosphatase is secreted.Blue colonies were grown in broth to log phase and supernatant was collected and analyzed using the 1-Step™ PNPP Substrate Solution from ThermoFisher and repeated in triplicate.The genomes from the best secreting isolates were sequenced via Illumina whole genome sequencing,150 Mbp: 66 times genome coverage, to find the insertion site of the transposome cassette.The insertion site sequence was analyzed using signalP 5.0 27 to find the strongest predicted secretion signal.

mIL-10 production and functionality in-vitro
The DNA sequence of the strongest signal was added to a codon harmonized mouse-IL-10 gene preceded by a strong constitutive promoter in a synthesized double stranded DNA fragment (gBlock™ Gene Fragment, IDT, USA).This was fused with a harmonized erythromycin resistance gene from C. striatum using overlap extension PCR to create the new transposome cassette and inserted into the C.mast genome.Positive isolates were screened using antibiotic selection (100 µg/ml) with over 200 positive mutants collected.These were grown to log phase and screened for mil-10 production in their supernatant using IL-10 Mouse Uncoated ELISA kit (ThermoFisher) and normalized to OD600.In three separate experiments, the best cytokine secreting isolates were verified by growing each to exactly 0.5 OD600 and quantifying supernatant mIL-10.The functionality of the cytokine was tested in vitro by analyzing the effect of mutant supernatant on t-cell proliferation.Naïve (CD44 -CD62L + ) or memory (CD44 hi CD62L low ) CD8 + , CD4 + and ɣδ T cells were isolated by fluorescence activated cell sorting (FACS) on a Sony Biotechnology MA900 cell sorter.After sorting, cells were labeled with a proliferation dye (Cell Trace™ Invitrogen) and stimulated for 3 days with plate-bound aCD3 (1mg/ml) and soluble aCD28 (5mg/ml)I.Six hours after the initial stimulation, 10 µl of supernatant from log-phase bacterial cell culture broth of AS1 or mIL-10 secreting mutants.
After stimulation, proliferation was measured via flow cytometry using a Beckman Coulter Cytoflex LX.T-cell culture supernatants from these assays were then assessed for IFN-γ and IL-17 production using ELISA (BioLegend).

Fitness of mil-10 transposon mutants
To assess the fitness of the mIL-10 secreting mutants they competed with wild-type (WT) C. mast in the same broth culture through three separate experiments.Transposon mutants and WT AS1 were grown to log phase in liquid media, pelleted, washed and resuspended at 0.5 OD600 in media without antibiotic.1mL from each transposon mutant resuspension was mixed with 1 ml of WT and allowed to grow together for 6 hours.The cultures were then ten-fold serially diluted to 10 -8 and 100ml was plated on agar both with and without antibiotic to select for the mutant.Comparing the CFU on each of the plates allows for the quantification of the ratio of mutant bacteria contained in the culture.These isolates were also evaluated in-vivo for their ability to colonize the mouse eye.
Transposon mutants and WT AS1 were grown to log phase in liquid media Then pelleted and resuspended at 1x10 6 CFU/ml in PBS.Five mice per sample were inoculated with 5 µl of each mutant, control WT and PBS dropped onto mouse eyes that had been brushed with a sterile swab to disrupt the tear film.This was repeated every other day for a total of 3 inoculations.Two weeks after the last inoculation mouse eyes were swabbed and plated on selective media.
Colonies were counted and compared to the colonization rate of WT AS1 to determine in vivo fitness in three separate experiments.

In-vivo corneal wounds
To evaluate the effect of the mIL-10 producing mutants on the ocular surface we inoculated the eyes as previously described and bacteria were quantified to ensure colonization and no contamination.Two weeks after the last inoculation a 2.5mm epithelial corneal wound was made on the right eye of each mouse using an Alger brush without disrupting the basement membrane.A 0.01% Fluorescein solution was then dropped onto the eye and pictures were taken at T=0. Fluorescein-stained cornea pictures were again taken at T=10 hours and every three hours after until the first wounds showed full closure at 22 hours.The size of the wound at each time point without full closure was measured by increasing the saturation of only the green pigment in the image and using the analyze particles function of Fuji (ImageJ).The size at each timepoint was then compared to the size at T=0 to find the amount of healing at each time point for each individual mouse.These data were replicated with three total experiments.Finally, the mIL-10R was blocked in vivo using CD210 antibody administration (Bio X Cell).
Two weeks after the final bacterial inoculation, 200 mg of CD210 in 1X PBS was injected intraperitoneally (IP).Controls were given IP injections of IgG.Twenty-four hours after injection corneal wounds were performed and analyzed at t=0 and t=19 hours as previously described.
The amount of healing for each isolate was compared between CD210 treated mice and IgG injected mice as well as each isolate compared to the WT-control.

Statistics
All ANOVA and Student's t-tests were performed using Graphpad PRISM software.
Comparisons across more than 2 samples used an ANOVA with Dunnett's multiple comparisons test to the WT AS1 sample.When only 2 samples were compared a student's ttest was performed.For the in vitro competition assay a one-tailed single sample t-test was performed with an expected mean of 0.5.When comparing epithelial wounds with and without IL-10R blocking an unpaired student's t-test was used.Only wounds that healed at least 15% after 19 hours were included in the analysis.All error bars represent standard error of the mean.
After identifying the active C. mast secretion signal, we generated a new cassette that consisted of the adhesion secretion signal upstream of a codon harmonized mIL-10 and an antibiotic resistance gene.To incorporate this cassette into the C. mast genome, we generated another set of transposomes according to our previous study 25 , which was used to generate another transposon mutant library.After screening supernatants of the transposon mutant library, we found over 50 isolates that secreted ELISA detectable mIL-10; these were collected, and the three best candidates are shown in Fig. 2a.These data suggested that C.mast properly produced and secreted mIL-10 that is recognized by commercially available antibodies.The transposons for these isolates were inserted into the genes for trehalose 6 phosphate, DnaK, and 23s rRNA for mil10.4,13,and 14 respectively.
Knowing that mIL-10 is glycosylated 29 , we wanted to ensure that mIL-10 was bioactive in cells that would respond to IL-10 signaling.To address this, we isolated memory (CD44 hi CD62L low ) CD4 + , CD8 + , and ɣδ T cells from C57BL/6 mice.We non-specifically activated these cells by stimulating αβ T cells with plate-bound aCD3 and soluble aCD28 for 3 days.In the presence of supernatants from mIL-10-C.mast, T-cells proliferated to a lesser extent as compared to T cells stimulated in supernatants from WT C. mast (AS1) (Fig. 2b).Similarly, the production of the proinflammatory cytokine, IFNg, was inhibited by supernatants from IL-10-C.mast, which supports the regulatory nature of mIL-10 during in vitro stimulations demonstrating functionality (Fig. 2c).Notably, the cytokine, IL-17, which is required for optimal protection of the ocular surface from spontaneous infection 30 , was not inhibited by IL-10-C.mast supernatants (Fig. 2d).Because IL-17 is not reduced in the presence of supernatants from IL-10-C.mast, we posit that IL-10-C.mast does not affect the host-microbe interactions concerning the ocular microbiome.

IL-10-C. mast has a similar in vivo fitness compared to wild-type C. mast
To assess whether IL-10-C.mast would survive a physiological environment, we performed in vitro and in vivo experiments to quantify fitness.First, mutant and WT strains of C. mast were grown to log phase and mixed at a 1:1 ratio.This allowed a direct comparison of growth between each mutant and WT C. mast (Fig. 3a) After 6 hours in liquid broth, isolates 4 and 14 did not have a significantly lower ratio of mutant bacteria as the initial time point.
Conversely, isolate 13 trended towards a reduced ability to compete with WT C. mast, showing a significantly reduced ratio of mutant vs WT bacteria as in the initial culture.We next analyzed the mutants for their ability to colonize the mouse conjunctiva.Five million CFU of log-phase mutant or WT C.mast were applied to the eye every other day for a total of 3 inoculations.After 1 week, eyes were swabbed, and colonizing bacteria were quantified.Both isolate 4 and 14 were able to colonize the mouse eye with no significant change in CFU compared to WT while isolate 13 was unable to colonize the mouse eye at all (Fig 4b).These data indicate that neither isolate 4 nor 14 were significantly less fit than WT C. mast; however; isolate 13 was significantly impaired in its ability to colonize the eye.Given that isolate 13 could not colonize the eye, and that the transposon cassette was incorporated into the DnaK gene, we conclude that the transposon impaired the function of the critical DnaK chaperon protein and reduced the fitness of isolate 13 to the point that colonization was not possible.

IL-10 produced by C. mast can speed wound healing in an IL-10 dependent manner.
To assess the therapeutic potential of IL-10-C.mast, we asked whether the secreted mIL-10 could affect physiology at the ocular surface.To do this, we used a mouse model of corneal wound healing due to the advantageous nature that IL-10 has on the closing of corneal wounds.One week after mice were ocularly colonized with either WT C. mast, IL-10-C.mast, or PBS, a 2.5 mm abrasion of the corneal epithelium was made using an Alger brush.Fluorescein imaging was used to assess the size of the wounded area at the time of injury and until 19 hours after injury (Fig. 4a&b) The IL-10 secreting bacteria significantly increased the rate of wound healing compared to controls.To ensure that the observed effects were directly related to IL-10 signaling, we administered IL-10R blocking antibody (CD210) intraperitoneally 24 hours prior to wounding and we observed fluorescein staining at t=0 and t=19 hours (Fig. 4c).Our results showed that wound healing was significantly slowed after CD210 administration in mice colonized with IL-10-C.mast.CD210 treated mice that were colonized with IL-10-C.mast had the same rate of wound healing as PBS and WT AS1 controls supporting the notion that engineered IL-10 was biologically functional and was responsible for the increased rate of wound healing after corneal injury.Together, this outlines the therapeutic potential of engineering an ocular resident microbe to deliver bioactive therapeutics directly to the ocular surface to treat disease.

Discussion
The ocular surface continues to be a difficult tissue for the application of therapeutics given the sensitive nature of the cornea, aversion of patients to touching the eye, and the constant flow of tears across the ocular surface.Current standards of care for local delivery of therapeutics include invasive and/or laborious methods that include subconjunctival injections and topical application up to once every 2 hours.To remedy this, we devised a strategy where we would engineer the known eye-colonizer, C. mast, to produce and secrete an antiinflammatory and pro-wound healing cytokine.In this study, given the anti-inflammatory and pro-healing qualities of IL-10, we decided to engineer C. mast to secrete IL-10 at the ocular surface.
Despite Corynebacterium spp.having been engineered in the past, most studies have focused on engineering C. glutamicum-a soil-habiting bacterium-to produce amino acids in an industrial production setting 31 , our study; however, focused on genetically manipulating C. mast for the purposes of generating a therapeutic delivery vehicle or a live biotherapeutic product (LBP).The benefits LBPs are that colonization with the LBP negates the need for repeated treatment regimens that reduce compliance.Similarly, LBP therapies could theoretically reduce production costs as the therapeutic is derived from the bacterium rather than synthetic methods.Furthermore, LBPs have proven effective in pre-clinical and some clinical trials involving Crohn's disease and irritable bowel syndrome (IBS) 20,22 .More recently, IL-10 produced in Mycoplasma pneumoniae, was shown to be effective at limiting the inflammatory response related to Pseudomonas aeruginosa lung infection 32 .With that, we hypothesized that IL-10 delivered by an eye-colonizing microbe may be beneficial for ocular surface health.Indeed, the production and release of IL-10 from C. mast resulted in more efficient tissue repair at the ocular surface after corneal injury.This suggests that this may be an effective method of treatment for traumatic ocular surface injuries, which affect over 1 million Americans per year 1 .Conversely, how this system would affect infectious ocular surface disease has yet to be determined.This strain of C. mast was originally isolated from ocular surface and skin of mice, but unlike laboratory strains of bacteria, genetic tools and protocols to manipulate C. mast have not been well-developed.Recently, our group used transposon mutagenesis to incorporate the gene of mCherry into the C. mast genome 25 .While we used this same system to incorporate IL-10 into C. mast, in this study, we needed to identify a signal sequence, which we could use to direct C. mast to secrete IL-10 into the extracellular space.Through the PhoZ transposon library we discovered a strong Sec/SPI signal peptide, a standard signal that is trafficked outside the cell by the Sec translocon and cleaved by Signal Peptidase I, that would serve this purpose.
A difficulty in modifying host immune responses is the balance of regulating inflammation, wound repair, and preventing infection.Given that we induced expression of IL-10, an immune regulatory cytokine, we were concerned that the continual production and release of IL-10 may affect the growth and control of C. mast resulting in C. mast becoming a pathobiont.Notably, we observed that IL-10 from C. mast reduced the inflammatory cytokine, IFNg, but it did not affect the production of IL-17 from ɣδ T cells, which is produced in response to C. mast colonization.
More importantly, we monitored mice colonized with WT C. mast and IL-10-C.mast for multiple months (data not shown) and found that neither caused clinical signs of pathology during colonization.Similarly, there were no differences in the host response-IL-17 from ɣδ T cells or neutrophil recruitment-to colonization with either WT C. mast or IL-10-C.mast.Therefore, we concluded that even though IL-10-C.mast produced an immune regulating cytokine, it is unlikely that this genetically engineered microbe would become pathogenic in an immune competent host.
Because of our engineered C. mast's ability to remain long-term, induce IL-17 immunity similar to WT, and lack of pathology, we demonstrate that use of engineered C. mast as an LBP has the potential to be well-tolerated and beneficial in a clinical setting.
LBPs such as this still require an incredible amount of tuning.Here, we simplified the system to ensure that we reduced as many variables as possible.Indeed, we were able to focus our efforts on identifying a secretion signal and ensuring that functional IL-10 was produced during colonization.Future iterations of this technology may rely on other cytokines and/or factors that may better tune the ocular surface microenvironment to encourage homeostasis.Specifically, there may be proteins or cytokines that help reduce inflammation related to Sjogren's Syndrome and severe dry eye disease.Alternatively, it may be possible to use this technology to produce growth factors that may assist the longevity of adoptively transferred stem cells or reduce the possibility of corneal graft rejection.On the microbe side, since this bacterium is susceptible to a number of antibiotics used for the eye, it can be removed from the ocular surface if desired, and work to determine optimal approach is underway.
Most LBP research has focused on influencing the host-gut microbiome axis; however, more attention is being directed at low biomass sites.Recently, LBPs have proven useful in the lungs 32,33 ; however here, we have demonstrated the first proof-of-concept of an LBP that can colonize the ocular surface.Specifically, the purpose of this LBP is to treat ocular surface wounds, but this technology can likely extend to other diseases at the ocular surface.In mice, this LBP was well-tolerated and did not result in uncontrolled outgrowth or gross pathology.
Additionally, this LBP secreted physiological levels of functional IL-10 in culture and in vivo at the ocular surface, which demonstrates that this technology may negate the need for burdensome treatment regimens that need to be applied to the eye multiple times a day.In the future, using engineered ocular resident microbes as LBPs may provide to be an alternative approach to treat ocular surface diseases.

Figure 3 .Figure 4 .
Figure 3. IL-10-C.mast has a similar in vivo fitness compared to wild-type C. mast .(a)Log phase mutant and WT bacteria were equally mixed in antibiotic free LB at 0.5 OD600.They were grown for 6 hours before being plated on agar with and without antibiotic.Comparing the CFU on each plate the ratio of mutant vs WT AS1 was determined.Bars represent the mean ratio of mutant CFU to total CFU ± SEM.Data were pooled from three independent experiments(n=3) and p values calculated using one-tailed single sample t-test with an expected mean of 0.5(n=3 df=2; 4(t=-0.1552);13(t=-3.4368);14(t=3.7634)).(b) 5x106 CFU of bacteria were applied to the ocular surface of C57BL/6 mice once every other day for a total of three inoculations.One week after the final inoculation, the eyes of mice were swabbed and CFU per swab were quantified using agar plates.Bars represent the mean CFU/eye ± SEM.Individual n values were from individual mice.Data were pooled from 3 experiments and statistical p values were determined using ANOVA (n=14, F=5.468, df=3,52)).