Microbial host interactions and impaired wound healing in mice and humans: defining a role for BD14 and NOD2

Chronic wounds cause significant patient morbidity and mortality. A key factor in their 2 etiology is microbial infection, yet skin host-microbiota interactions during wound repair 3 remain poorly understood. We investigated microbiome profiles of non-infected human 4 chronic wounds and showed that reduced diversity was associated with subsequent healing 5 outcome. Furthermore, poor clinical healing outcome was associated with increased local 6 expression of the pattern recognition receptor NOD2 . To investigate NOD2 function in the 7 context of cutaneous healing, we treated mice with the NOD2 ligand muramyl dipeptide (MDP) 8 and analyzed wound repair parameters and expression of anti-microbial peptides. MDP 9 treatment of littermate controls significantly delayed wound repair associated with reduced re- 10 epithelialization, heightened inflammation and upregulation of murine β -Defensins ( mBD ) 1, 11 3 and particularly 14. We postulated that although BD14 might impact on local skin microbial 12 communities it may further impact other healing parameters. Indeed, exogenously administered 13 mBD14 directly delayed mouse primary keratinocyte scratch wound closure in vitro . To further 14 explore the role of mBD14 in wound repair, we employed Defb14 -/- mice, and showed they had 15 a global delay in healing in vivo , associated with alterations in wound microbiota. Taken 16 together these studies suggest a key role for NOD2-mediated regulation of local skin 17 microbiota which in turn impacts on chronic wound etiology. 18 19


INTRODUCTION
Chronic wounds, which include pressure sores, venous and diabetic foot ulcers (DFUs), are a 2 global problem leading to substantial morbidity and mortality (Gottrup, 2004). Following 3 injury, skin-resident microbiota and pathogenic species may colonise the wound and proliferate 4 (Eming et al., 2014). Hence understanding the role of bacteria, both pathogenic and 5 commensal, in the context of skin wounding is important yet comparatively little research 6 attention has focused on this area (Loesche et al., 2017, Misic et al., 2014. Corynebacterium and a variety of other organisms (Attinger and Wolcott, 2012, James et al., 11 2008, Mancl et al., 2013, Rhoads et al., 2012. The innate immune system detects infection and 12 injury via pattern recognition receptors (PRRs), such as the Nod-like receptors. PRRs respond 13 to highly conserved microbial structures-pathogen-associated molecular patterns that can 14 trigger inflammatory and defense responses such as keratinocyte-mediated production of anti-15 microbial peptides (AMPs). AMPs provide rapid and efficient anti-microbial activity against a 16 wide range of pathogens (Dutta andDas, 2016, Harder et al., 2013). The skin has many AMPs 17 including Cathelicidins, β-defensins, S100A15, RNase-7 and Histones (Buchau et al., 2007, 18 Dorschner et al., 2001, Gallo and Hooper, 2012, Halverson et al., 2015, Simanski et al., 2010 Sorensen et al., 2006, Yang et al., 2017   demonstrate an association between the bacterial profile of non-infected human DFUs and 2 healing outcome, correlating with upregulated expression of the PRR NOD2. Using both 3 NOD2 stimulated and Defb14 null murine models we reveal new insights into the role of the 4 innate defense response in controlling the skin microbiota during wound repair. at clinical presentation (week 0). Patients were then separated into two groups according to 12 their time to heal over a period of 12 weeks; DFU healed ≤7 weeks (n = 10) versus non-healed 13 ≥12 weeks (n = 9). Eubacterial DNA profiles (UPGMA dendrogram) at presentation (week 0) 14 showed clear segregation between wounds that would heal versus those that would not ( Figure   15 1a; wound closure at ≤7 weeks (green) versus ≥12 weeks (purple), n = 19). 16S rRNA Illumina 16 high-throughput sequencing of a further set of DFU samples (n = 25) and non-metric multi-17 dimensional analysis (NMDS) showed no clear separation between the microbial profiles of 18 the healed compared to the non-healed wounds ( Figure 1b); however, non-healing wounds 19 were associated with significantly reduced overall phylum diversity ( Figure 1c). Phylum level 20 relative abundance was consistent between healed and non-healed wounds (Fig 1d); however, 21 interestingly genus level taxonomic classification of the wound microbiome revealed a 22 significantly altered microbial community in healed versus non-healed wounds, including 23 relative abundance variation within common skin-associated taxa such as Staphylococcus (23% 24 in healed wounds versus 19% in non-healing wounds), Anaerococcus (3% in healed wounds 25 5 versus 10% in non-healing wounds) and Coprococcus (classified in other genera category, 1 Figure 1e (P=<0.05)). The taxonomic information for all mapped reads at the genus level can 2 be found in the supplementary material (Table S2). Finally, the overall presence of bacteria in 3 wounds was assessed by direct Gram stain of DFU biopsy tissue which revealed no significant 4 difference in bacterial numbers between the groups (Figure 1f-g). Collectively this data 5 suggests that bacterial community diversity rather than overall bacterial burden correlates with 6 DFU healing outcome. 7 8 NOD2 is upregulated in human chronic wounds that fail to heal 9 We next assessed whether PRR expression was altered as PRRs have been implicated in the 10 skin microbiome regulation (Campbelle t al., 2013, Dasu et al., 2010, Lai et al., 2009, Lin et 11 al., 2012. Several TLRs trended towards increased expression in non-healing wounds ( Figure   12 2a-e) but only the intracellular PRR NOD2 was significantly increased (P<0.05, Figure 2f).

13
NOD2 is implicated in barrier function, epithelial turnover and repair (Cruickshank et al., 2008) 14 therefore we investigated NOD2 function in keratinocytes. Keratinocyte scratch wound closure 15 was significantly reduced following treatment with the NOD2 ligand, MDP (P<0.05, Figure   16 2g-h). Scratch closure was also inhibited by a range of TLR ligands (Figure S1a); however, 17 TLR2 ligands did not affect closure. The addition of mitomycin C to inhibit proliferation  We next investigated the impact of NOD2 activation using C57BL/6 mice subcutaneously 1 injected with MDP or vehicle control, prior to incisional wounding. MDP treatment 2 upregulated Nod2 mRNA in the wound ( Figure S1b) and showed a trend for upregulation of 3 the Nod2 associated downstream signalling molecules Rip2 but not Tak1, (Figure S1c Figure   11 3i-j). Collectively, these results demonstrate that MDP-mediated activation of NOD2 12 significantly delays repair. NOD2 has a known role in gut and lung epithelial AMP production specifically defensins 16 (Rohrl et al., 2008, Tan et al., 2015. MDP treated wounds had significantly upregulated levels 17 of mBD3 (P<0.05) and mBD14 (P<0.05) mRNA compared to control wounds ( Figure 4a).

18
Similarly, in vitro, MDP stimulated NHEKs significantly induced hBD1, hBD2 (the human 19 orthologue to mBD3) and particularly hBD3 (the human orthologue to mBD14; P<0.05, Figure   20 4b). We further explored the effect of mBD14 on wound healing, focusing on the keratinocyte 21 response. We used a mBD14 peptide (Reynolds et al., 2010), which we confirmed as 22 biologically active as it inhibited P. aeruginosa growth ( Figure S2a) and scratch-wounded 23 primary mouse keratinocyte monolayers were treated with 1, 10 or 25 μg/ml of mBD14 24 peptide. Keratinocyte migration was significantly decreased in a dose-dependent manner 25 7 (P<0.01, Figure 4c-d). Importantly, cell viability was unaffected by the peptide as determined 1 by examination of morphological features, suggesting that mBD14 directly influences 2 epidermal migration. The sequence homology between mBD14 and hBD3 is approximately 3 69% (Hinrichsen et al., 2008, Rohrlet al., 2008, therefore we tested mBD14 peptide on human 4 keratinocytes with similar results ( Figure S2b). We also investigated the impact of hBD3 on 5 keratinocyte function using hBD3 transfected cells; however, we saw no effect on keratinocyte 6 scratch closure ( Figure S2c). Chronic wounds had altered communities of bacteria compared with wounds that healed well 3 and we had shown that mBD14 peptide inhibited the growth of P. aeruginosa ( Figure S2a) 4 therefore, we assessed bacterial abundance in Defb14 -/mice. Total eubacterial abundance was 5 significantly increased in Defb14 -/mice compared to controls as revealed by Gram-staining 6 (P<0.01, Figure 6a-b) and 16S qPCR (P<0.05, Figure 6c). qPCR analysis of common skin 7 bacterial species revealed increased levels of P. aeruginosa (P<0.01) as well as P. acnes 8 (P<0.05, Figure 6d-g) implicating BD14 in a bacterial dysbiosis that is detrimental to healing.  In DFU patients, rather than the more "common" wound pathogens, we observed changes in 24 genera abundance such as Corynebacterium, Enterococcaceae, and Helcococcus associated 25 9 with non-healing. We assessed the DFU microbiome at time of clinical presentation before the 1 outcome of healing was known. Previous and complimentary longitudinal analysis of DFU-2 associated bacteria have linked poor healing to a more stable microbiome, whereas wounds 3 that healed well had a more dynamic microbiome that transitioned between community types 4 (Loesche et al., 2017). Similarly, our findings implicate a less diverse microbiome at the 5 initiation of healing, which may in turn impact upon the subsequent dynamics of the 6 microbiome during healing. It remains unclear whether such observations will be broadly 7 applicable to other wound types such as venous leg ulcers, decubitus ulcers and wounds that  Several previous studies have shown that TLRs are differentially regulated when comparing 13 acute wounds to chronic wounds, while a number of PRRs, such as TLR3, are important for 14 wound chronicity (Campbell et al., 2013, Dasu et al., 2010, Lai et al., 2009, Lin et al., 2012.

15
By contrast, our study tested PRR levels in longitudinally evaluated healing versus non-healing 16 chronic wounds. In this context, the only PRR to show statistically significant alteration was 17 NOD2. As the expression of NOD2 can be upregulated in response to bacterial ligation, it is 18 plausible that the observed differential NOD2 levels in non-healing wounds may reflect a 19 response to the differential bacterial composition of the wound environment.

24
Human chronic wounds 25 1 obtained in accordance with the Declaration of Helsinki. 25 wound biopsy patient samples 2 (mixed sex, aged ≥40 years) with chronic DFUs (defined as distal to the medial and lateral 3 malleoli, with a known duration ≥4 weeks, grade A1/B1, University of Texas ulcer 4 classification, no infection or ischaemia) were obtained at the time of presentation (week 0).

5
All patients received standard-of-care treatment, including regular debridement, non-6 antimicrobial dressing, and offloading. No local anaesthetic was used at any time during 7 treatment. At week 0 wound biopsy samples were collected from the margin of DFUs prior to 8 debridement using aseptic technique. Photographs of patient's wounds were taken weekly over 9 12 weeks to determine longitudinal healing outcome. DFUs were then separated into two 10 groups, those who healed (full wound closure at ≤7 weeks; 10 patients) and those who failed 11 to heal (wound not closed at 12 weeks; 9 patients) following current best practice treatment.    Table S1. EDTA, 1.2% triton X-100) and lysozyme (20 mg/ml) for 30 min at 37°C. DNA was extracted 2 using a Qiagen DNeasy™ blood and tissue kit (Qiagen, West Sussex, UK).