Neutrophil extracellular traps orchestrate formation of peritoneal adhesions


 Peritoneal adhesions are a poorly understood but highly prevalent condition that can lead to intestinal obstruction, pelvic pain, and infertility. While there is consensus that stress-induced inflammation triggers peritoneal adhesions, the process of their formation remained elusive to date. Herein, we show that neutrophil extracellular traps (NETs) serve as essential scaffold for adhesion formation and that DNases interfere with this process. Thus, peritoneal adhesions in murine models and in humans showed that these lesions are largely based on extracellular DNA derived from neutrophils. Furthermore, treatment with DNASE1 or a DNASE1L3 analog significantly reduced or even prevented peritoneal adhesions in experimental models. These data not only suggest that NET formation plays an essential role in peritoneal adhesions but also show that therapeutic application of DNases can prevent the formation of peritoneal adhesions.


Introduction
Peritoneal adhesions, a common consequence of serosal repair after abdominal interventions, represent a major burden for clinicians and surgeons a like. In fact, the formation of adhesions has been shown to occur as often as 93-100% of patients following abdominal surgery and can cause serious complications such as intestinal obstruction, pelvic pain, and infertility 1 . As a result, the quality of life of millions of patients throughout the world is affected by peritoneal adhesions. These lesions are also associated with considerable costs of roughly two billion US dollar per year in the US. Their actual burden in the medical setting is highlighted by a Scottish survey performed over a 10-year period. They suggested that 5.5% of all hospital re-admissions can directly be attributed to the formation of adhesions 2 . Lastly, the formation of adhesions has not only been linked to a reduced quality of life and signi cant health care costs, but also mortality rates of 6-15% 3 .
To date, no molecular therapies exist that interfere with the formation of peritoneal adhesions. Instead, therapeutic interventions to impede the formation of adhesions are limited to bioabsorbable lms, placed on surgical wounds to prevent adhesions of the peritoneum and the viscera. A number of barrier agents with various features are also available. However, these measures are complicated and time consuming. Furthermore, a recently published large Cochrane review, including 19 randomized controlled trials on the prevention of adhesion after gynecological surgery, did not reveal conclusive evidence of the effectiveness of such approaches 4 . Taken together, a treatment preventing peritoneal adhesions represents an unmet medical need. and anti-in ammatory activities. These remain activated for several days 6 and create a delicate balance which, if disrupted, can tip over from protecting the host from microbes to mediating hyperin ammation, inhibiting healing and increasing mortality 6 .
Neutrophils, the predominant leukocytes of acute in ammatory reactions, are immediately recruited into injured areas and remain there for about two days 7 . They are (1) the most abundant immune cells in the human circulation, (2) regarded as the rst line of defense of the innate immune system, and (3) the main leukocyte subset involved in the early phases of wound healing 8,9 . In response to infection and/or injury, neutrophils form NETs, which consist of high molecular weight double-stranded DNA laments that build robust scaffolds. These are decorated with histones and cytotoxic proteins, such as myeloperoxidase (MPO) and neutrophil elastase (NE), accounting for 70% and 20% of all proteins of NETs, respectively 10,11 . Neutrophils release NETs by multiple mechanisms: (1) NETosis, a programmed cell death pathway, (2) non-lytic discharge of parts or their entire nucleus, and (3) mitochondrial DNA release, providing an additional DNA source for NET formation 12 .
NETs are "double-edged swords" as they regulate homeostatic and pathological in ammation. During infection NETs exhibit antimicrobial functions, trap and kill extracellular pathogens in blood and tissue 12,13 . However, NETs also form during sterile in ammation. NETs stimulate platelet adhesion and coagulation 14 and the proteolytic activity of aggregated NETs traps histones 15 and contributes to the resolution of in ammation 16 .
While the function of NETs is essential in combating infection and in ammatory responses, a spatial and temporal inappropriate production of NETs can have detrimental effects. In fact, it has been shown that NETs contribute to the pathology of several in ammatory conditions, such as autoimmune diseases, wound healing, sepsis, and ischemia reperfusion injury 13,[17][18][19] . A common denominator of these disorders is the involvement of NETs as mediators of thrombosis and hyperin ammation and of the occlusion of vessels and ducts [20][21][22][23][24][25][26][27][28][29] .
Based on response-to-stress and in ammation as potential triggers for peritoneal adhesions we hypothesized that NETs play an important role in the formation of such adhesions Furthermore, we considered that therapeutic application of DNases that cleave NETs may represent a new treatment option for peritoneal adhesions.

Results
Laparotomy-induced adhesions in Dnase1and Dnase1l3-KO mice First, we examined the role of two endogenous DNases; DNASE1 and DNASE1L3 in abdominal adhesions. It is known that de ciency of the former contributes to lupus, whereas the latter is associated with scleroderma and autoimmunity 30-33. Mice de cient in DNASE1 or DNASE1L3 insu ciently metabolize extracellular DNA and nuclear remnants. We studied these knockout mice with the same protocol for the formation of adhesions which included laparotomy, coagulation and two sutures. We observed that the knockout of Dnase1 or Dnase1l3 augmented formation of peritoneal adhesions ( Figure  1A,B,E). This nding underscored the importance of DNases in the regulation of abdominal adhesions.
The most distinctive phenotype with respect to the formation of adhesions were Dnase1l3-KO mice, indicating that DNASE1L3 is of paramount importance to prevent the formation of adhesions. The peritoneal thickness was also increased in these DNase mutant mice with more pronounced results in Dnase1l3-KO than Dnase1-KO mice (Figure1C,D).
Upon closer examination, local and systemic DNase treatment not only affected the formation of adhesions, but also in uenced various aspects of peritoneal wound healing, as seen by an altered histology of the affected area (Figure 2A-H). Compared to controls, the Dnase1l3-KO mice displayed a signi cantly higher collagen type I to type III ratio ( Figure 2F When measured NE activity in the surgical sites, we observed a robust NE activity (Supplement 3A). Importantly, this activity was resistant to both the pharmacological (Sivelestat) as well as the endogenous inhibitor a1-antitrpysin (a1-AT) (Supplement 3B,C). It appears that neutrophil recruitment was similar in WT and KO mice in the context of adhesion formation.

Laparotomy-induced adhesions in WT mice-treatment with DNases
We next aimed to identify the procedure that most effectively reduces and/or prevents the formation of peritoneal adhesions. Recombinant DNASE1 and/or a DNASE1L3 analog were applied topically during the surgery and also systemically 24h and 48h afterwards. The DNASE1L3 analog signi cantly reduced adhesions at day 21 post-surgery; evaluated with the Leach and Nair adhesion scores ( Figure 3A,B)).
Severity of adhesion was re ected by peritoneal thickness, which was signi cantly reduced by treatment with DNASE1 or DNASE1L3 analog compared to controls ( Figure 3C-E).
Anastomosis, deserositation, and thermal injury-driven adhesion formation; treatment with Dornase alfa Next, we evaluated the effects of DNASE1 on wound healing in three typical clinical settings: (1) intestinal anastomosis (2), deserosation as a consequence of abdominal surgery, and (3) thermal injury. Assessing the effects of DNase on "bene cial" wound healing is of importance to surgeons, as DNases could potentially facilitate the spread of bacteria or delay wound healing. However, our results showed that mortality rates amongst animals treated with DNASE1 undergoing deserosation, intestinal anastomosis, and thermal injury were not elevated in comparison to controls; i.e. animals treated with inactivated DNASE1 (failure of intestinal anastomosis: DNASE1 0.0% vs. controls 10.0%, p>0.05; deserosation: Dornase alfa 22.2% vs. controls 29.3%, p>0.05; heat exposure: DNASE1 0.0% vs. controls 0.0%, p>0.05).
No case of wound infection or incisional hernia was found. In contrast, peritoneal adhesions were again signi cantly reduced in mice treated with DNASE1 ( Figure 4A We conclude that the immune system plays a key role in the formation of the adhesions.

Role of DNASE1L3 in healing of thermal wounds in mice
To further elucidate the role of DNASE1L3 in wound healing, a second repair model based on thermal injury was employed. Animals with the Dnase1l3-KO had similar wound closure time ( Figure 5B) but signi cantly worse scar appearance as measured by the Yeong scale ( Figure 5A, 4C). Treatment with a DNASE1L3 analog signi cantly improved scarring and resulted in a signi cantly faster wound closure ( Figure 5A-C). Similar to the adhesion model, the Dnase1l3-KO animals had a signi cantly higher collagen type I to type III ratio ( Figure 5D) and a more parallel alignment of collagen ( Figure 5E,F). Particularly the last nding is indicative of immature scars 35 . The absence of DNASE1L3 did neither signi cantly affect neutrophil concentrations nor NETs formation. However, there was a trend towards higher Ly6G positive neutrophils and higher expression of NE and citH3 ( Figure 5G-J). Treatment with a DNASE1L3 analog signi cantly reduced neutrophil activation and NETs formation ( Figure 5H,I).
Human adhesions contain neutrophil-borne proteins and extracellular, NET-like DNA.
To nd out whether NETs are also components of human peritoneal adhesions, we analyzed human surgery material. As shown in representative images ( Figure 6 and Supplement 5), the NET-associated proteins NE and MPO were expressed in the human peritoneal adhesions ( Figure 6A). Additionally, the expression of DNASE1L3, especially at the margins of the adhesion, was detected ( Figure 6D). The isotype controls ( Figure 6B, 6E) were negative for the uorescence signal, con rming the speci city of the signal obtained for NE, MPO and DNASE1L3. Figure 6C shows a Hematoxylin & Eosin (HE) staining of the same biopsy. Enlarged details of the surgical biopsy material are depicted in Supplement 5 with additional positive immuno uorescence signals for the expression of the hallmark NET marker citH3. In a longitudinal section of a human adhesion the canonical NET markers MPO, citrullinated histone H3 and DNA were detected in addition to brin (Supplement 6).

Discussion
Our ndings suggest that NETs play a pivotal role in the formation of peritoneal adhesions in mice and humans. We support previous research, which suggests that the pathogenesis of the formation of adhesions is based on a combination of in ammation, coagulation, and brinolysis 36 .
It is established that the in ammatory phase precedes wound healing 37 . However, there are also controversial reports: (1) Wounds in areas with inherently lower levels of macrophages, neutrophils, and T cell in ltration, such as oral wounds, heal instantly with marginal in ammation and scar formation 38 , (2) neutrophil depletion in mice accelerates the re-epithelialization rate of uninfected diabetic wounds 39 , (3) in the wounds of diabetic mice elevated levels of citrullinated histone H3 (citH3) were found and healing was delayed; (4) wound healing was accelerated in peptidylarginine deiminase 4 (PadI4)-KO mice which have very limited NETs formation when compared with WT mice, and (5) DNASE1 accelerated healing of incisional wounds in diabetic mice 9 .
In peritoneal as well as thermal wound healing, NETs appear to have similar disadvantageous effects 18 . This is supported by our ndings that at sites of peritoneal injury NETs were found and DNases were expressed in both humans and mice. This nding suggests that NETs and DNases regulate the formation of adhesions. A loss-of-function single nucleotide polymorphism in Dnase1l3 is associated with systemic sclerosis, a rare condition that is characterized by hardening of the skin and tissue due to excessive accumulation of collagen 40 . We show that Dnase1l3-de cient mice formed massive deposits of collagen in the skin and peritoneum upon injury. Further, topical application of a recombinant DNASE1L3 analog reduced scarring and facilitated healing. Collectively, the data suggest that DNASE1L3 is critical for tissue regeneration 41,42 .
Assessment of peritoneal adhesions revealed widespread abundance of NETs characterized by extracellular DNA and neutrophil elastase activity. NETs may likely represent scaffolds for peritoneal adhesions. Thus, injury leads to recruitment of neutrophils which increase steadily at in ammatory foci, where they form NETs. When NETs reach high local densities, they aggregate and form structures that span over several centimeters, that can easily bridge two intestinal loops. Since DNA builds the backbone of NETs, they are robust, sticky, highly exible and elastic. If they are not cleaved by DNases in time, they go through a certain degree of maturation, activate platelets and brinogen and are decorated with high molecular weight brin. The latter physically strengthen these structures by providing attachment points for broblasts, endothelial cells and, smooth muscle cells and immune cells, which eventually organize and develop into mature adhesions.
The study's results are promising and DNase therapy to prevent the formation of adhesions is ready to be evaluated in a clinical study. As our data shows, DNases do not appear to negatively affect wound healing. Indeed, our results suggest an enhanced physiological wound healing process after treatment with DNases, particularly with the DNASE1L3 analog. The main effect appears to be the reduction of NETs-associated in ammation as con rmed by the RNA seq data. We observed that DNases application or KO-Models reduced and increased the formation of adhesions, respectively; in both models wound healing was preserved. It did not enhance the incidence of hernia formation or anastomotic insu ciency.
DNases are an elegant option to prevent the formation of adhesions; they are cost effective and can metabolize extracellular DNA in vitro 10

Methods
Study design: The study was approved by the Hamburg State Administration for animal research (73/15, 63/16). A total of 143 six-week-old mice were utilized for the experimental model and all environmental parameters within the animal facility complied with the German guide for the care and use of laboratory animals (Animal Welfare Act). All animals including the genetic knockouts (Dnase1-KO and Dnase1l3-KO) used to examine the role of DNases in the process of the formation of adhesions and wound healing had the same genetic background (C57BL/6). The Dnase1-KO and the Dnase1l3-KO mice were generated as described earlier 13,44,45 . We obtained the WT mice from Jackson Laboratory and employed littermates for all treatment groups.
Additionally, peritoneal samples of nine children who underwent a second laparotomy within two weeks after their primary laparotomy were included in the study (Department of Pediatric Surgery of the University Medical Center Hamburg-Eppendorf from 2017 to 2019) and analyzed histologically. Samples were obtained only from cases with non-infectious conditions that lead to re-operative surgery.
Anonymized tissue collection was in accordance with the guidelines of the medical research ethics committee of Hamburg (Ethik-Kommission der Ärztekammer Hamburg, PV5489) and with the 1964 Helsinki declaration and its later amendments. We obtained written informed consent from the legal representatives.

Human sample collection
Peritoneal tissue samples from the maximum of the adhesion were collected at the time of secondary relaparotomy surgery. Patients with infectious conditions were excluded from this current study. We stained and analyzed the samples as described above for murine samples.

Tissue sampling
After blood collection, morphologic analysis was performed and captured using a 4K/12-megapixel camera. Next, the scar was dissected and evenly distributed into test tubes containing Bouin solution.
Generation of Dnase1 knockout mice Dnase1 mutant mice were generated by CRISPR/Cas9 mediated mutagenesis in JM8A3 embryonic stem (ES) cells from C57BL/6N origin 44. In brief, ES cells were transfected with pX458 (obtained from Addgene, Watertown, MA, in which the Dnase1 speci c gRNA-sequence 5´ TGACATCGCTGTTATCCAAG 3ẃ as inserted 45. GFP-expressing ES cells were sorted and mutations in individual ES cell clones were analyzed by sequencing the amplicon generated by primers anking the target sequence in exon 3. One clone showing a 65 bp deletion from intron 2 into exon 3 was selected for blastocyst injections, generation of chimeric mice and further breeding with C57BL/6N mice. The deletion does not allow splicing into exon 3, and potential alternative splicing into exons 4, 5 or 6 containing the active sites for DNASE1 enzymatic activity lead to frameshift mutations and premature stop codons.

Animal Procedures
We rst determined the time course of extracellular DNA formation in the course of NET formation after injury (laparotomy) in wild-type mice (see methods section). We visualized extracellular DNA using SYTOX orange and observed a peak 72 hours after the induction of adhesions (not shown). We observed the cumulative maximum of adhesions at 21 days using the Leach and Nair adhesion score (not shown). Therefore, we used both day 3 as well as day 21 post injury to assess the effects of DNases on NET formation and abdominal adhesions (Supplement 1A).
Mice were randomized into groups of equal size. For better standardization, a single surgeon performed all operations. Anesthesia was induced with 5% iso urane (Baxter, Unterschleiß heim, Germany) and maintained with 2.5% iso urane gas delivered through a facemask. Pre operative antisepsis was performed with iodopovidone and all mice received 0.02 mg/kg bodyweight (BW) buprenorphine (Reckitt Benckiser, Mannheim, Germany) subcutaneously, 30 min preoperatively for analgesia.
The study involved two models: (1) adhesion model to assess the effect and implications of NETs and DNase treatment on the formation peritoneal adhesion, (2) secondary intention wound healing model to evaluate DNases in wound healing after a thermal injury.

Model 1: Adhesion formation
Adhesions were induced using a bipolar electrocoagulation method 52: Standardized lesions were in icted on an area measuring 0.5 cm × 1.5 cm by sweeping the bipolar electro coagu lation forceps over the abdominal peritoneum for 2 seconds. The current was delivered using the following settings: Bipolar Soft, Effect 4, 40 Watts. The defects were subsequently closed using two interrupted sutures (6/0 Vicryl, Ethicon, Norderstedt, Germany) to induce an ischemic eld around the traumatized area. The sutures were placed equidistantly (5 mm) along the defect and 1 mm from the wound's edge.
To establish the most effective therapy, several treatment combinations using recombinant human DNASE1 (Dornase alfa, Roche, Mannheim, Germany) with a dosage of 10 mg/kg BW, as well as NTR-10, a recombinant human DNASE1L3 analog provided by Neutrolis, Cambridge, MA, USA, with a dosage of 1 mg/kg BW were tested. The control groups received a vehicle. Ultimately, a sham group without (1) the intervention, other than the laparotomy, and without (2) treatment was included.
To determine the effects of DNases on wound healing, three typical clinical scenarios were reproduced: (1) Deserositation -induced by rubbing a mini-prep on the wall of the small intestine, (2) Intestinal anastomosis -performed with a 8x0 Vicryl continuous suture after dissection of a small segment of the small intestine, and (3) Thermal injury -induced by heat exposure on the intestine using a red lamp with a distance of 1 meter for 10 minutes.
Model 2: thermal injury Thermal injuries, serving as a model for secondary intention wound healing, were induced as described previously 18. In short, a 1.5cmx1.5cm large burn injury was induced on the neck of the animals. At the two timepoints (72 hours or 21 days), animals were euthanized after anesthesia using iso urane as described above. Re-laparotomy was conduct ed and assessment of the formation of adhesions or dermal scaring was performed. Ultimately, resection of the lesion took place prior to all animals being euthanized via decapitation.

Assessment of the adhesions
All adhesions were evaluated immediately after re-laparotomy. Macroscopic grading of the formation of adhesions was assessed by two independent surgeons, blinded to the animal groups and blinded to each other, using the Leach grade, as well as the Nair grade. The Leach grade was originally designed to score adhesions of the uterine horn and was thus modi ed for this study to evaluate peritoneal adhesions. The Leach score consists of three factors 53: (1) severity of adhesions (0=no adhesion, 1= lmy avascular, 2= vascular or opaque, 3=cohesive attachment), (2) degree of adhesions (0=no adhesion, 1=adhesion separable with gentle traction, 2=adhesion separable with moderate traction, 3=requiring sharp dissection), and (3) extent of adhesions (0=no adhesion, 1=1-25%, 2=26-50%, 3=51-75%, 4=76-100%).

Assessment of the scars
Burn scars were evaluated before euthanasia using the modi ed Yeong scale, by two surgeons, blinded for the treatment groups 51 . This 3-item wound evaluation scale was speci cally developed for thermal injuries assessing the scar surface appearance, height and color mismatch from 1 (best) to 4 (worst) for each item, resulting in a total score ranging from 1 (best scar) to 12 (worst scar).

Microscopic grading
All specimens were evaluated histologically. In our burn model the scars were marked with blue dye for better microscopic evaluation and standardization. All specimen were then washed in phosphate buffered saline (PBS) and xed in 10% buffered formalin before being embedded in para n and cut into 3µm thick sections, slides were then stained using hematoxylin and eosin (HE) and examined by two researchers who were blinded to the groups in light microscopy, using a magni cation of ×4 and x10. Assessment of wound healing (epithelialization) was carried out in a standardized manner and expressed as a percentage of the whole wounded area. The unhealed wound was measured as the distance between both edges of the wound and the total wound diameter as the distance between the wound edges.
Immunohistochemistry (HE, Ly6g, Collagen I/III, SMA, Fibrin) Hematoxylin and Eosin (HE) and Lymphocyte Antigen 6 Complex Locus G6D (1A8-Ly6G) staining was performed with a standardized staining procedure. Collagen bers were stained using Pico Sirius red (ab150681, Abcam, Cambridge, UK), using polarized light microscopy was used to differentiate collagen I from III. An antibody for smooth muscle actin (SMA, ab5694, Abcam, Cambridge, UK) was applied to the samples, serving as a marker for myo broblast, which induce wound contraction. Fibrin deposition was determined using a brinogen antibody (ab58207, Abcam, Cambridge, UK). Subsequently, the stained samples were incubated according to manufacturer's instructions. In accordance with each antibody examined, an appropriate isotype control antibody was used as a negative control. All samples were