Crohn’s disease proteolytic microbiota enhances inflammation through PAR2 pathway in gnotobiotic mice

ABSTRACT

Emerging evidence implicates microbial proteolytic activity in ulcerative colitis (UC), but whether it also plays a role in Crohn’s disease (CD) remains unclear. We investigated the effects of colonizing adult and neonatal germ-free C57BL/6 mice with CD microbiota, selected based on high (CD-HPA) or low fecal proteolytic activity (CD-LPA), or microbiota from healthy controls with LPA (HC-LPA) or HPA (HC-HPA). We then investigated colitogenic mechanisms in gnotobiotic C57BL/6, and in mice with impaired Nucleotide-binding Oligomerization Domain-2 (NOD2) and Protease-Activated Receptor 2 (PAR2) cleavage resistant mice (Nod2^−/−; R38E-PAR2 respectively). At sacrifice, total fecal proteolytic, elastolytic, and mucolytic activity were analyzed. Microbial community and predicted function were assessed by 16S rRNA gene sequencing and PICRUSt2. Immune function and colonic injury were investigated by inflammatory gene expression (NanoString) and histology. Colonization with HC-LPA or CD-LPA lowered baseline fecal proteolytic activity in germ-free mice, which was paralleled by lower acute inflammatory cell infiltrate. CD-HPA further increased proteolytic activity compared with germ-free mice. CD-HPA mice had lower alpha diversity, distinct microbial profiles and higher fecal proteolytic activity compared with CD-LPA. C57BL/6 and Nod2^−/− mice, but not R38E-PAR2, colonized with CD-HPA had higher colitis severity than those colonized with CD-LPA. Our results indicate that CD proteolytic microbiota is proinflammatory, increasing colitis severity through a PAR2 pathway.

KEYWORDS:

• Crohn’s disease
• proteolytic activity
• inflammation
• DSS-induced colitis
• Proteinase-activated receptor 2 (PAR2)
• gnotobiotic mice

Introduction

=Inflammatory bowel disease (IBD) is an immune mediated condition that includes Crohn’s disease (CD) and ulcerative colitis (UC). The exact cause is unknown but altered host-microbe interactions in individuals with genetic susceptibility are involved in its pathogenesis.^1,2 Global rates of IBD continue to increase annually,^3,4 therefore novel treatments that target the environmental drivers of inflammation are needed.

Over the last decade, there has been growing interest in the role of proteases in IBD.^5–9 The gastrointestinal tract is exposed to different proteases that have key roles in digestion, immunity, visceral perception, and barrier function.^8,10 This activity is tightly controlled by a system of host-derived anti-proteases to prevent dysregulated intestinal inflammation and damage.⁵ An imbalance in proteolytic function, characterized by increased host proteases such as trypsin and elastase and/or decreased host anti-proteases such as the human serine protease inhibitor (serpin) molecule elafin, has been described in IBD.^5,6,9 Increased proteolytic activity can promote inflammation through several mechanisms including matrix remodeling, mucus degradation,⁹ activation of protease activated receptors (PARs) for which proteases have different receptor specificity,¹¹ modification of tight junctions increasing permeability,¹⁰ and/or modulation of cytokines and chemokines.^9,12 Bacteria, including those present in the human gut, are also known sources of proteases;⁸ but their role in inflammation and the precise signaling pathways triggered by this imbalance are only beginning to emerge.

Previously, we have shown increased fecal proteolytic activity in individuals at risk for developing IBD before the clinical onset of UC. Colonization of germ-free mice with fecal slurries induced high fecal proteolytic activity and a spontaneous molecular proinflammatory tone that was associated with increased bacterial protease and peptidase genes, suggesting a microbial contribution to proteolytic imbalance in UC.¹³ It is unknown whether bacterial proteases are involved in CD, and whether this proinflammatory immune tone induced by high bacterial proteolytic activity predisposes to more severe colitis.

Thus, here we colonized germ-free mice with selected fecal communities from CD patients that exhibited either high (CD-HPA) or low (CD-LPA) proteolytic activity, or from healthy control subjects with high (HC-HPA) or low proteolytic activity (HC-LPA). We investigated colitogenic severity and underlying pathways using germ-free wild type, Nod2^-/-, and protease-resistant PAR2 mutant (R38E-PAR2) mice subjected to mild experimental colitis.

Results

CD-HPA increased proteolytic activity in adult and neonatally colonized mice

We set out to investigate the role of bacterial proteolytic activity in CD. For this, we first selected donor fecal microbiota from CD patients and controls based on their in vitro proteolytic activity to colonize germ-free mice. Colonization of germ-free mice with human donor microbiota resulted in a microbiota profile that clustered close to the donor sample (Supplementary Figure S1), demonstrating a transfer of microbiota characteristics. Importantly, the proteolytic functional profiles of donor microbiota were also transferred to germ-free mice through colonization. Proteolytic activity was analyzed in fecal samples from adult germ-free mice colonized for 3 weeks with CD-HPA, CD-LPA, HC-HPA, or HC-LPA feces (Figure 1a). Germ-free mice had high baseline proteolytic activities, likely originating from dysregulated host proteases in the absence of a microbiome.¹⁴ Colonization of germ-free mice with microbiota with low PA, whether from HC or CD, decreased proteolytic activities. In contrast, HC-HPA decreased elastolytic activity but not overall or mucolytic activity, while CD-HPA led to higher overall, elastolytic, and mucolytic activities (Figure 1b-d). This suggests CD-HPA lacks the homeostatic modulation of host elastolytic activity in germ-free mice. In parallel, in vitro PAR2 cleavage by germ-free mouse intestinal content was high, which decreased after colonization with HC-LPA, HC-HPA, or CD-LPA, but not CD-HPA (Figure 1e). No differences in PAR1 cleavage were detected between any of the groups (Supplementary Figure S2). Because de novo colonization of adult germ-free mice does not mimic natural colonization, we investigated whether the proteolytic phenotype was also transferred through maternal colonization (Supplementary Figure S3).¹³ At weaning, F1 litters from dams colonized with CD-LPA displayed low intestinal proteolytic activities, compared with litters born from dams colonized with CD-HPA (Supplementary Figure S3b-d). These data confirm a robust and long-term transfer of intestinal proteolytic phenotype through gnotobiotic colonization.

Figure 1. Colonization with CD-HPA lacks a homeostatic modulation of host proteolytic activity in germ-free mice. (a) Fecal samples from a previously characterized human cohort¹³ were selected for mouse colonizations. Feces from a healthy control with low (HC-LPA; n = 1) or high proteolytic activity (HC-HPA; n = 1), or from Crohn’s disease patients with low (CD-LPA, n = 1) or high proteolytic activity (CD-HPA; n = 2 differentially identified by red open or closed dots) were gavaged to germ-free (GF) C57BL/6 mice. (b) Overall proteolytic activity, (c) elastase activity, and (d) mucolytic activity were measured in mouse fecal samples 3 weeks following colonization. Germ-free (GF; n = 32), HC-LPA (n = 14), HC-HPA (n = 5), CD-LPA (n = 7), CD-HPA (n = 9). (e) Luminescent intensity was used to measure PAR2 cleavage in vitro. Germ-free (GF; n = 7), HC-LPA (n = 10), HC-HPA (n = 4), CD-LPA (n = 6), CD-HPA (n = 8). Data are presented as median with interquartile range, with whiskers extending from minimum to maximum. Each dot represents one mouse. Statistical significance was determined by one-way ANOVA with Tukey post-hoc test for multiple comparisons.

CD-HPA increased proinflammatory immune tone

We assessed inflammatory gene expression in adult colonized mice and found 172 genes to be differentially expressed between germ-free, HC-LPA, CD-LPA, and CD-HPA mice (Supplementary Figure S4a, Supplementary Table S1). Several genes related to innate immune function, such as MAP kinases, Rhoa, Myd88, and Tollip were increased in germ-free mice compared with the other groups. Colonization with HC-LPA downregulated genes related to inflammation and barrier function compared with germ-free mice and both CD-HPA and CD-LPA colonized mice (Supplementary Figure S 4a). Thus, colonization with CD microbiota, regardless of the proteolytic phenotype, upregulated several inflammatory and barrier-modulating genes. However, 18 genes were differentially expressed between CD-LPA and CD-HPA mice (Figure 2a). MAP kinases such as Map2k6 and Rps6ka5, chemokines such as Ccl11 or Ccl21, as well as Nuclear Factor Kappa B Subunit 1 (NFƙB1) were upregulated in CD-HPA colonized mice compared to CD-LPA colonized mice. Other genes related to inflammation such as interleukin (IL) receptors Il1rap and Il22ra2 were also increased, while Nox1, Tlr1, and Tnf were decreased in CD-HPA colonized mice compared to CD-LPA colonized mice. The tight junction gene Claudin 1 (Cdln1) was upregulated in CD-HPA while Nox1 was downregulated (Figure 2a). In parallel with these results, colonic polymorphonuclear (PMN) cell counts in CD-HPA mice were higher than in CD-LPA mice (Figure 2b-c), whereas HC-LPA colonization decreased PMN counts compared to germ-free mice (Supplementary Figure S4b and c). These findings confirm previous results that gnotobiotic colonization impacts expression of genes regulating immune and barrier function,^15–17 and identifies novel differential effects based on the proteolytic activity of the CD microbial community.

Figure 2. Colonization of GF mice with CD-HPA induces an inflammatory immune tone. (a) Heatmap of gene expression, determined by NanoString nCounter codesets, in colonic tissue of CD-LPA (n = 4) and CD-HPA1 (n = 4) colonized mice. Data presented as Z-scores of differential gene expression and analyzed in R package. Only significantly altered genes are shown. (b) Polymorphonuclear (PMN) cell counts in the colonic mucosa of CD-LPA (n = 7) or CD-HPA (n = 9) feces. Each dot represents one mouse. Open circles refer to mice colonized with CD-HPA donor 2. Data are presented as median with interquartile range with whiskers extending from minimum to maximum. Statistical significance was determined by one-way ANOVA with Tukey post-hoc test for multiple comparisons. (c) Representative images of the different groups of colonic mucosa sections stained with hematoxylin and eosin and examined at 40× magnification. Arrowheads show examples of PMN cells.

Microbial composition and function differ between CD-LPA and CD-HPA

We then set to further investigate fecal microbial profile differences between CD-LPA or CD-HPA colonized mice. Reduced alpha diversity was observed in CD-HPA using both Chao1 and Shannon indexes (Figure 3a). Beta diversity revealed CD-LPA and CD-HPA formed two distinct clusters (Figure 3b). We also observed a natural transfer of the microbial community through maternal colonization. Both dams and offspring colonized with CD-LPA microbiota had higher alpha diversity compared with CD-HPA (Supplementary Figure S3e). Beta diversity revealed that the stool microbiome profile of the offspring clustered with that of their dams, showing two separate clusters between CD-LPA and CD-HPA (Supplementary Figure S3f). We then investigated differential abundance of genera (Supplementary Figure S5) and species (Figure 3c) between CD-LPA and CD-HPA mice. Of note, some species considered beneficial, such as Akkermansia muciniphila, Alistipes putredinis, and Ruminococcus bromii were decreased, while Hungatella (formerly Clostridioides) hathewayi and Romboutsia ilealis were increased in CD-HPA mice (Figure 3c). We also found that the changes in bacterial composition were accompanied by distinct profiles of predicted peptidases and proteases genes between CD-HPA and CD-LPA mice (Supplementary Figure S6a), supporting the observed differences in fecal proteolytic activity (Figure 1). Since serine proteases of host^18,19 and microbial origin^13,20 have been shown elevated in active IBD, as well as before diagnosis of UC,¹³ we used the metagenomic contribution data generated by PICRUSt2 to identify the bacteria responsible for the 6 serine proteases that were found at different levels in the predicted metagenome between CD-LPA and CD-HPA mice (Supplementary Figure S6b and Figure 4a-b). Three of the serine proteases were found at lower levels in CD-HPA mice compared with CD-LPA and were not further investigated. Furthermore, the serine protease K08372 found at higher levels in the predicted metagenome of the CD-HPA group compared with CD-LPA was mainly due to the metagenomic contribution of the probiotic genus Bifidobacterium (Supplementary Figure S6b) and therefore hypothesized to not be colitogenic. Two other serine proteases, K04772 and K13275, were increased in CD-HPA mice compared with CD-LPA mice. The species we identified predicted to contribute to K04772 was Hungatella hathewayi, and Romboutsia ilealis for K13275 (Figure 4). Since PICRUSt functional predictions are based on DNA amplicons and may not reflect the active gene expression, we quantified the mRNA transcripts for the H. hathewayi K04772 and R. ilealis K13275 (Figure 4c-4e). H. hathewayi K04772 transcripts were found at higher levels in CD-HPA compared with CD-LPA (Figure 4c) and had a positive correlation with the PICRUSt-predictions (Figure 4d), suggesting it is actively being expressed. Conversely, K13275 transcripts from R. ilealis had the opposite profile, suggesting they are being actively repressed (Figure 4e-f). The data suggest that K04772 from H. hathewayi may be one of the proteases present in CD-HPA microbiota contributing to its proteolytic profile.

Figure 3. Mice colonized with CD-LPA and CD-HPA have a distinct microbial composition. Fecal microbiota composition in germ-free mice colonized with CD-LPA (n = 7) or CD-HPA (n = 9) was determined by 16S rRNA gene sequencing. (a) Alpha diversity of fecal microbiota profiles calculated using the Chao 1 and Shannon indexes. Data are presented as median with interquartile range with whiskers extending from minimum to maximum. Statistical significance determined by t-test. (b) Beta diversity of fecal microbiota profiles of CD-LPA and CD-HPA colonized mice at genus level illustrated with PCoA using Bray-Curtis and weighted UniFrac distance metrics. Each dot represents one mouse. Open circles refer to mice colonized with CD-HPA donor 2. (c) Heatmap of differentially expressed bacteria at the species level between CD-LPA and CD-HPA colonized mice. Data presented as Z-scores with the average relative abundance of that species shown in the rightmost column. Extreme outliers are marked with a slash and are not included for the statistical analysis.

Figure 4. Serine proteases are differentially expressed in CD-LPA and CD-HPA colonized mice. PICRUSt2 predicted metagenomic contribution was used to identify bacterial species responsible for the increase in (a) K04772 and (b) K13275. (c) Targeted transcriptional analysis using RT-qPCR of the H. hathewayi serine protease K04772. (d) Correlation between quantified transcripts and PICRUSt2-predicted genes for H. hathewayi serine protease K04772. (e) Targeted transcriptional analysis using RT-qPCR of the R. ilealis serine protease K13275. (f) Correlation between quantified transcripts and PICRUSt2-predicted genes for R. ilealis serine protease K13275. Each dot represents one mouse. Open circles refer to mice colonized with CD-HPA donor 2. Data are presented as median with interquartile range with whiskers extending from minimum to maximum. Statistical significance was determined by t-test for the expression of transcripts and Spearman index for correlations.

C57BL/6 and Nod2^−/−, but not R38E-PAR2, germ-free mice colonized with CD-HPA displayed higher colitis severity

To investigate colitogenic potential of CD-HPA, we induced acute intestinal injury in germ-free C57BL/6 adult mice colonized with HC-LPA, HC-HPA, CD-LPA, or CD-HPA (Figure 5a). Compared with C57BL/6 mice colonized with CD-LPA, CD-HPA colonized mice had higher microscopic scores (Figure 5b-c). Microscopic scores were similar between HC-LPA and HC-HPA colonized mice and no differences in colonic expression of Il1rap, Il22ra, and Nfkb were detected (Supplementary Figure S7). To explore underlying pathways, we investigated colitis severity in Nod2^−/− and R38E-PAR2 mice colonized with CD-LPA or CD-HPA (Figure 6a). Germ-free mice developed colitis as demonstrated before,²² indicating the microbiota is not required for DSS mediated injury (Supplementary Figure S8). Nod2^-/- mice colonized with CD-HPA had higher microscopic scores compared with CD-LPA colonized mice (Figure 6b). In contrast, R38E-PAR2 mice had milder colitis severity when colonized with CD-HPA compared with CD-LPA (Figure 6c), along with a blunted Pi3k and Mapk3k response, as measured by colonic gene expression (Supplementary Figure S9). This indicates the PAR 2 pathway mediates the colitogenic effect of the microbiota.

Figure 5. CD-HPA microbiota has a colitogenic effect in C57BL/6 germ-free mice. (a) Germ-free (GF) C57BL/6 mice were colonized with feces from HC-LPA (n = 5), HC-HPA (n = 5), CD-LPA (n = 5), or CD-HPA (n = 4). Three weeks later, acute colitis was induced, where mice received 2% DSS in drinking water for 5 days followed by 2 days of water. (b) Histological scores of colonic intestinal inflammation were determined using a modified pathological score.²¹ Statistical significance was determined by one-way ANOVA with Tukey post-hoc test for multiple comparisons. Each dot represents one mouse. (c) Representative images of colonic mucosa sections stained with hematoxylin and eosin and examined at 20× magnification.

Figure 6. PAR2 cleavage-resistant mice colonized with CD-HPA are protected from colitis. (a) Germ-free (GF) Nod2^−/− and R38E-PAR2 were colonized with feces from CD-LPA (n = 4 Nod2^−/−; n = 3 R38E-PAR2) or CD-HPA (n = 3 Nod2^−/−; n = 4 R38E-PAR2). Three weeks later, acute colitis was induced where mice received 2% DSS in drinking water for 5 days followed by 2 days of water. Colonic histological scores were determined using a modified pathological score,²¹ and representative images of colonic mucosa sections stained with hematoxylin and eosin and examined at 20× magnification are shown for (b) Nod2^−/- and (c) R38E-PAR2 mice. Statistical significance was determined by t-test between CD-LPA and CD-HPA within each strain of mice. Each dot represents one mouse.

Discussion

Microbial proteolytic imbalance has been implicated in UC,¹³ but whether similar functional microbial differences exist in CD is unclear. Using “model” human microbiota donors, we show that both adult and neonatal colonization of germ-free mice with high proteolytic activity feces from CD patients develop this functional phenotype in the intestine. Colonization with CD-HPA led to a spontaneous proinflammatory immune tone and worsened experimental colitis in wild type and Nod2^−/- mice, but not in mice with a PAR2 mutation that makes it resistant to cleavage by proteases, indicating the proinflammatory pathway requires intact PAR2 cleavage signaling. The approach to use “model” microbiota with well-known and distinct functional profiles has been previously employed successfully.^23,24 Here, it allowed us to select microbiota with different proteolytic activities. While average proteolytic activity has been shown to be higher in patients with active IBD versus remission and healthy controls,⁶ not all IBD patients will exhibit this activity. Indeed, the range distribution of fecal proteolytic activity observed in our cohort is similar to that reported in previous studies.⁶ However, we have previously shown that patients at risk for IBD that later develop UC have higher proteolytic activity than those that remain healthy,¹³ suggesting imbalanced proteolytic activity could contribute to inflammation in UC and be a functional microbial biomarker of disease progression.

Colonization of mice with LPA microbiota, whether derived from HC or CD, decreased host proteolytic activity in germ-free mice, but CD-HPA lacked this homeostatic capacity. The results are supported by previous findings in which germ-free mice colonized with irritable bowel syndrome HPA microbiota failed to suppress host proteolytic activity in germ-free mice.¹⁴ Previous reports indicate germ-free mice exhibit innate immune imbalance and increased PMN cells, such as eosinophils, that are downregulated after conventionalization.²⁵ In line with this, we found lower PMN cell counts in mice colonized with low proteolytic microbiota compared with germ-free and CD-HPA mice. Colonization of germ-free mice impacted 172 genes related to immune and barrier function, which is in agreement with previous work using mono-colonization with Bacteroides thetaiotaomicron.¹⁵ Decreased expression of key genes involved in innate immune responses such as Mapk, Rhoa, Myd88, or Tollip were observed post colonization. However, genes related to inflammation and barrier function were differentially modulated by colonization based on disease state (HC vs CD) and by the proteolytic activity of the donor. In line with a previous study where colonization of germ-free mice with commensal microbiota was needed for development of a physiologic colonic barrier,¹⁷ we found that colonization with HC-LPA downregulated proinflammatory genes in germ-free mice. Thus, even though colonization with CD microbiota, independently of proteolytic activity, was associated with some proinflammatory gene expression, differences were further observed between CD-LPA and CD-HPA. We have previously shown that UC patients have high proteolytic activity in feces before development of clinical disease and that colonization of germ-free mice with these communities induce a proinflammatory immune tone.¹³ Here, we observed that most of the genes upregulated in CD-HPA were involved in p38-MAPK and NFκB pathways known to be related to the release of key proinflammatory cytokines.^26–28 Interestingly, induction of some of these cytokines via MAPK signaling cascade can be driven through the activation of PAR2.^29,30 Other important mediators of inflammation such as interleukin receptors (Il1rap and Il22ra2), previously implicated in different inflammatory disease models including IBD,^31–34 were also upregulated in CD-HPA mice compared with CD-LPA. Additionally, downregulation in transcription of genes such as Nox1 and Tlr1 in CD-HPA mice suggest a deficient barrier defense against pathogens, as previously reported by other authors.^35,36 Altogether our results identify new differential effects on immune modulation and the proinflammatory gene expression profile based on the proteolytic function of the CD microbiota.

There is increasing interest in the functional contribution of the microbiota in IBD.^37–40 Proteolytic imbalance has been described as one of the possible functional mechanisms through which the microbiota could induce inflammation in IBD,^8,9 but few studies have shown an association between the proteolytic phenotype and dysbiosis.^13,41 In this study, we demonstrate that CD-HPA microbiota from colonized mice was characterized by a decrease in known beneficial bacteria such as Akkermansia muciniphila, Alistipes putredinis, and Ruminococcus bromii, previously shown to be reduced in IBD patients.^42,43 We also found that opportunistic pathogens were increased in CD-HPA colonized mice, such as Romboutsia ilealis, that has been reported to be increased in an experimental colitis model.⁴⁴ Recently, Lloyd-Price et al.⁴² found Hungatella hathewayi to be increased compositionally and transcriptionally in IBD patients, suggesting high metabolic activity of the species during inflammation. These findings are in accordance with our results where the protease K04772 of H. hathewayi is actively transcribed in samples from CD-HPA colonized mice compared with CD-LPA colonized mice. NCBI classifies the quantified H. hathewayi transcript as belonging to the TIGR02037 family, which includes both K04771 (also known as degP and htraA and classified in the protease database MEROPS as S01.273) and K04772 (also known as degQ and hhoA and classified in MEROPS as S01.274). Proteases from these families are found in pathogens, such as Enteropathogenic E. coli (EPEC) and Helicobacter pylori, where they have been described as virulence factors,⁴⁵ shown to degrade E-cadherin,⁴⁶ and hypothesized to cleave PAR2.⁴⁷ This suggests that K04772 from H. hathewayi may be one potential protease in CD-HPA contributing to the proinflammatory phenotype in our model. However, it is common for similar functions to be conserved between environments even when the bacterial species vary, therefore we cannot rule out the possible involvement of other bacteria and proteases that could contribute to a higher colitogenic potential, especially in other CD patients with high proteolytic activity. Nevertheless, our data suggests K04772 of H. hathewayi is a target of interest for future work that may include the isolation of the taxon, genomic deletion of K04772 and characterization of the mutant bacteria, as well as metaproteomics analysis to demonstrate translation of the protein in mouse models.

To further investigate the underlying colitogenic pathways of CD-HPA, we used a model of intestinal injury employing a low dose of dextran sodium sulfate in drinking water in several mouse strains. We found that C57BL/6 mice colonized with CD-HPA had higher colitis severity compared with CD-LPA mice. On the other hand, mice colonized with HC-HPA or HC-LPA developed similar colitis suggesting a higher colitogenic drive from CD-HPA microbiota compared to HC-HPA. The colitogenic drive of CD-HPA was demonstrated also in Nod2^-/- mice, a model with deficient Nucleotide-binding and oligomerization domain NOD-like receptor signaling and one of the main polymorphisms associated with CD.^48–50 The results indicate that CD-HPA does not require the NOD2 signaling pathway exclusively to increase colitis severity. However, PAR2 cleavage-resistant mice colonized with CD-HPA were protected from colitis. While protease-independent down-regulation of proinflammatory pathways could contribute to the lower colitis,⁵¹ the results support the requirement of PAR2 cleavage signaling pathway for CD-HPA colitogenic effect. Indeed, PAR2 is known to be activated by proteolytic cleavage of the N terminal extracellular domain leading to a proinflammatory cascade,^9,52 and it has been previously described as a possible proinflammatory bacterial-driven mechanism in the small intestine of mice mono-colonized with Pseudomonas aeruginosa-producing elastase.⁵³ By selecting “model” donors with specific proteolytic characteristics, we show that a high protease activity, with demonstrated PAR2 cleavage in vitro, is transferred to colonized mice leading to higher colitis severity. The reduced colitis severity observed in PAR2 cleavage resistant mice supports that high fecal proteolytic activity of donor CD microbiota is colitogenic through the PAR2 pathway.

In conclusion, we provide new evidence that high microbial proteolytic activity in CD is colitogenic and identify PAR2 signaling as a required underlying pathway. In this study, H. hathewayi is a taxon that may contribute to the proinflammatory phenotype through the serine protease K04772. Although our results do not preclude the involvement of host proteolytic activity, they support the investigation of specific microbial proteolytic signatures in inflammatory bowel diseases, as their characterization could identify patients who will preferentially benefit from anti-proteolytic therapies.

Material and methods

Human donor selection for wild type and knock-out mouse colonization

We selected fecal samples from three CD patients in active flare or remission as determined by Harvey Bradshaw Index (HBI)^54,55 and/or biopsy by colonoscopy, and from two healthy volunteers who participated in a previous study¹³ (Table 1). Fecal samples from CD patients were selected based on their in vitro proteolytic activity (Figure 1a); as low proteolytic activity (CD-LPA, n = 1), and high proteolytic activity (CD-HPA, n = 2). The selected fecal samples from the healthy control subjects were confirmed to have low proteolytic activity (HC-LPA, n = 1) or high proteolytic activity (HC-HPA, n = 1). Collection and use of these samples were approved by the Hamilton Integrated Research Ethics Board (# 12–599 T and # 2820) and all study participants read and signed the written informed consent.

Table 1. Patient demographics.

Age	Sex	Diagnosis	HBI	Anatomopathology	Location	Treatment	Proteolytic Activity
58	F	HC	NA	Normal	NA	Perindopril	Low
25	M	HC	NA	Normal	NA	None	High
56	F	CD	1	Chronic Inflammation	Ileocolonic	Mesalazine	Low
32	F	CD	8	Chronic Inflammation	Ileocolonic	None	High
24	F	CD	14	Chronic Inflammation	Ileocolonic	None	High

HC: healthy control.

*Harvey Bradshaw Index: assessment for Crohn’s disease (CD) activity.

< 5 remission.

5–7 mild activity

8–16 moderate activity

> 16 severe activity.

Mouse models

C57BL/6N, Nod2^−/−, and protease-resistant PAR2 mutant (R38E-PAR2) germ-free mice were generated by 2-stage embryo transfer⁵⁶ at McMaster’s Axenic Gnotobiotic Unit (AGU) and were bred under germ-free conditions. Original breeding pairs of specific pathogen free (SPF) Nod2^−/− (on a C57BL/6 background) mice were kindly donated by Prof. Jonathan Schertzer (McMaster University). SPF R38E-PAR2 breeding pairs (on a C57BL/6 background) were originally provided by Dr. Wolfram Ruf from Johannes Gutenberg University Mainz prior to germ-free re-derivation. R38E-PAR2 mice have an intact PAR2 pathway but the external cleavage domain is mutated and therefore resistant to proteolytic activation.⁵⁷ Germ-free mice were kept in flexible film gnotobiotic isolators at McMaster’s AGU. Male and female mice were used in all experiments. All mouse experiments were approved by the McMaster University Animal Care Committee and McMaster Animal Research Ethics Board in an amendment to the Animal Utilization Protocol (# 210930).

Mouse colonization procedures

Donor stool samples were collected using anaerobic atmosphere generation bags (Sigma), aliquoted anaerobically, and stored at −80°C until use. Adult germ-free C57BL/6, Nod2^-/-, or R38E-PAR2 mice (8–12 weeks) were colonized with either CD-LPA, CD-HPA1 or 2, HC-LPA, or HC-HPA microbiota. Thawed aliquots from human donor fecal samples were diluted 1:10 in sterile phosphate-buffered saline (PBS) in anaerobic conditions, and mice received 200 µL of fecal suspension by oral gavage. CD-HPA and HC-HPA donor samples had over 1 standard deviation total proteolytic activity compared with the average observed in healthy controls. After colonization, microbiota-humanized mice were housed in ISO negative cages (Techniplast). For the neonatal model, CD-LPA or CD-HPA1 or 2 was used to colonize four C57BL/6 females that were then paired with stud males and housed in ISO negative cages. F1 litters were studied at 5–6 weeks of age.

Pregnant dams colonized with the microbiota from CD-HPA donor 2 did not result in thriving litters. Due to limited amount of sample, we therefore used litters born from dams colonized with CD-HPA donor 1 or adult mice colonized with CD-HPA donor 1 for most experiments. Donors used are marked in the figures throughout the manuscript.

All mice were kept on a 12-hour light/12-hour dark cycle with free access to autoclaved food (Teklad 7004, Envigo) and autoclaved water within the AGU, according to specific standard operating procedures of the facility.

Colitis induction

Colitis severity was investigated in 9 to 12-week-old germ-free C57BL/6 colonized with either HC-LPA, HC-HPA, CD-LPA, or CD-HPA1 and germ-free Nod2^−/− and R38E-PAR2 mice colonized with either CD-HPA1 or CD-LPA. Three weeks after colonization, acute colitis was induced by 2% dextran sodium sulfate (DSS, MW 36–50 kDa; MP Biomedicals) in drinking water followed by two days on water only. DSS intake and general health condition were monitored daily. Microscopic inflammatory scores in colonic sections were determined, as described below.

Measurement of fecal proteolytic activity

Total proteolytic, elastolytic and mucolytic activities were measured using fecal supernatants from mice. Non-specific proteolytic activity was determined using the substrate azocasein (Sigma-Aldrich) as previously described.¹³ Fecal supernatants were incubated in 50 mmol/L Tris-HCl buffer pH 8.2 supplemented with 1 mmol/L CaCl2, 50 mmol/L NaCl, and Triton 0.25% (weight/volume (w/v)), and 0.5% (w/v) azocasein at 37ºC. The reaction was stopped by adding 10% trichloroacetic acid, and absorbance was measured at 366 nm. Elastolytic activity was analyzed using the substrate Suc-Ala3-pNa (Sigma-Aldrich). Fecal supernatants were incubated in the same buffer at 37ºC with the substrate. Absorbance at 410 nm was measured at several timepoints. Enzyme activity was determined by using a standard curve of pancreatic porcine elastase (Sigma-Aldrich). Mucolytic activity was determined by a bioassay using Brain Heart Infusion (BHI) media enriched with 0.05% (w/v) of mucin from porcine stomach type III (Sigma-Aldrich). Fecal supernatant was spotted into agar wells and incubated for 24 h. Zones of clearance were measured after staining agar plates with 10% amido black for 24 h. A standard curve was used to determine mucin-degrading activity of the samples using collagenase type I (0.25–1.0 FALGPA units/mg) from Clostridium histolyticum (Sigma-Aldrich).

Histological analysis

Colonic tissues were fixed in 10% formalin solution and embedded in paraffin. Colonic sections (5 µm) were stained with hematoxylin and eosin (H&E) and examined under light microscopy. Polymorphonuclear (PMN) cell count in the epithelium was performed at 40× magnification and expressed as number of PMN cells in an area of 9 × 10⁴ µm². The degree of intestinal inflammation was blindly evaluated using a modified pathology score²¹ based on the following parameters: a) Presence of PMN and erythrocytes in luminal exudate (absent, score = 0, scant = 1, moderate = 2, dense = 3); b) Epithelial damage and desquamation (no pathological change, score = 0, mild regenerative change = 1, moderate with patchy desquamation = 2, severe with diffuse desquamation = 3); c) Acute inflammatory infiltrate of the mucosa (PMNs) (absent, score = 0, scant = 1, moderate = 2, dense = 3); d) Mononuclear cell infiltrate of the mucosa (absent, score = 0, one small aggregate = 1, more than one aggregate = 2, several aggregates = 3); e) Goblet cell depletion (abundant number of goblet cells, score = 0, mild depletion = 1, moderate = 2, severe = 3); f) Cryptitis and crypt abscesses (absent, score = 0, isolated cryptitis = 1, diffuse cryptitis = 2, crypt abscesses = 3); g) Architectural damage (no pathological change, score = 0, mild and isolated architectural change = 1, moderate change = 2, severe and diffuse changes = 3); h) Edema (absent, score = 0, mild edema = 1, moderate = 2, severe = 3). An average of all parameters was used to determine the microscopic inflammatory score (0–3). Images were acquired using ImagePro Plus (Media Cybernetics, Rockville, USA).

NanoString gene expression

Total RNA extraction from colonic tissues was performed with the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Deoxyribonuclease (DNase) digestion during purification was carried out by using RNase-free DNase set (Qiagen). Mouse genes involved in immune and barrier pathways were determined using two different NanoString nCounter Gene Expression CodeSets (mouse inflammation panel, 254 genes and customized mouse neuroimmune/barrier panel, 71 genes) and first analyzed with nSolver 4.0 software. The normalized data was further analyzed in R using the same statistical approach as used for the microbiota data.

16S ribosomal RNA gene sequencing

Total DNA was extracted from mouse fecal samples using a custom protocol as previously described.⁵⁸ The hyper-variable regions (V3-V4) of the bacterial 16S ribosomal RNA gene were amplified by polymerase chain reaction (PCR) using a slightly adapted protocol described by Bartram et al.⁵⁹ Briefly, for the PCR, forward barcoded primers targeting V3 region (v3f_341f-CCTACGGGNGGCWGCAG) and reverse primer targeting V4 region (v4r_806r-GGACTACNVGGGTWTCTAAT) were used. Forward primers included six-base pair barcodes to allow multiplexing samples. Purified PCR products were sequenced in the Illumina MiSeq platform by the McMaster Genomics facility. Sequences obtained were trimmed with Cutadapt software⁶⁰ and processed with Divisive Amplicon Denoising Algorithm 2 (DADA2)⁶¹ using the SILVA reference database version 138.1. A phylogenetic tree of the sequences was calculated using FastTree 2⁶² and data was explored using the phyloseq⁶³ package in R (version 4.1.1) and custom scripts. A total of 2,402,009 reads were obtained with an average of 30,794.99 per sample and ranged from 5253 to 97,929 per sample. Functional predictions of the microbial communities based on 16S rRNA gene sequences were made with PICRUSt2^64,65 using the KEGG database.⁶⁶

Gene expression by RT-qPCR

Total RNA extraction from feces was performed using Stool Total RNA Purification Kit according to manufacturer’s instructions (Norgen Biotek, Thorold, Canada). Differential ratio amplicons (Ramp) method was used to determine the quality of the bacterial RNA as previously described.⁶⁷ Nine out of 14 samples had high quality, defined as having a Ramp of >0.6. Although there was a small subset of samples that had a lower Ramp value, these were not significantly different from the rest of the samples in the same group. Bacterial cDNA was prepared using random hexamers with SuperScript IV Reverse Transcriptase (Invitrogen, Waltham, USA). Hungatella hathewayi serine protease K04772 transcript levels and Romboutsia ilealis K13275 were assayed using RT-qPCR. The primers for the serine proteases were designed with Primer 3 from NCBI-Refseq NZ_WNME01000035.1 and NZ_LN555523 (Table 2). Mouse RNA was extracted as described for Nano String gene expression and quality was assayed by visualization in an agarose gel. Mouse cDNA was prepared using the iScript cDNA synthesis kit (Bio-rad) and transcripts were assayed using RT-qPCR using the primers described in Supplementary Table S2. Relative quantitation was performed using standard curves constructed from serial dilutions of PCR products as previously described.⁶⁸ Expression levels were normalized to overall 16S rRNA transcript levels.

Table 2. List of primers for the RT-qPCR.

Primer	Sequence (5’→ 3’)	Orientation	Target
K04772- Hhat-F	TAATACCTCTCCCGGTGTCG	Forward	Serine protease (K04772)
K04772- Hhat-R	TGTTGCCATGGCTATCTCTG	Reverse
K13275-Rile-F	AATAGGTGTGGCTCCGGATG	Forward	Serine protease (K13275)
K13275-Rile-R	ACCTGCTGCACATACCACTA	Reverse
UnivF	TCCTACGGGAGGCAGCAGTG	Forward	Bacterial 16S rRNA gene
UnivR	TTACCGCGGCTGCTGGCACG	Reverse

In vitro PAR1 and PAR2 cleavage

Chinese hamster ovary (CHO) cells, stably infected with constructs expressing PAR1 or PAR2 tagged with an N-terminal nano luciferase (nLuc) and a C-terminal enhanced yellow fluorescent protein eYFP, were used to detect cleavage of PAR1 or PAR2.^69–71 Cells were cultured in Ham’s F-12 Nutrient Mix (Gibco, ThermoFisher Scientific) with 1 mM sodium pyruvate, 100 U/mL penicillin, 100 µg/mL streptomycin, 1 mM L-glutamine, 10% heat-inactivated fetal bovine serum (FBS, Gibco, ThermoFisher Scientific), and 600 ug/Ml G418 sulfate (Invitrogen, ThermoFisher Scientific). Cells were seeded in a 96-well culture plate at a density of 1.0 × 10⁴ cells per well and incubated for 48 hr at 37°C 5% CO₂. Cells were rinsed three times with Hank’s Balanced Salt Solution (HBSS, Gibco ThermoFisher Scientific) containing CaCl₂ and MgCl₂, then incubated with 100 µL HBSS at 37°C 5% CO₂. Basal luminescence was measured by taking 50 µL of supernatant. Cells were then treated with 50 µL of fecal suspensions from either germ-free, CD-HPA, CD-LPA, HC-LPA, or HC-HPA colonized mice as well as porcine trypsin (PAR1 and PAR2 agonist, Sigma Aldrich) or HBSS for 15 min at 37°C 5% CO₂. Luminescence of the treated wells was then measured on a plate reader by taking 50 µL of the supernatant in the presence of the nano luciferase substrate furimazine according to manufacturer’s instructions (Promega).

Statistical analysis

Data analysis was carried out using GraphPad Prism 9 (GraphPad Software, USA). One-way ANOVA with Tukey post-hoc test for multiple comparisons was used for comparisons between more than 2 groups. For microbiota analysis, data transformation was used when required and possible to achieve a normal distribution (logarithmic, square root, inversion, and inverted logarithm). Unpaired Student’s t-test or Wilcox rank sum test was performed when only comparing two groups or by one-way ANOVA or Kruskal–Wallis tests were performed when comparing more than two groups, followed by Tukey or Dunn’s post-hoc tests with Holm correction, as appropriate. False Discovery Rate (FDR) with a Q-value of 0.1 was used for the multiple testing of bacterial taxa or a Q-value of 0.15 for exploratory analysis of predicted proteases. Extreme outliers were defined as any data point more than 3 interquartile ranges (IQRs) below the first quartile or above the third quartile. Permutational multivariate analysis of variance (PERMANOVA) was used to evaluate statistical differences in beta-diversity between the groups. Spearman correlation was calculated to evaluate the linear association between two nonparametric variables. For Nano String data, normalization was performed using nSolver 4.0 software (Nanostring Technologies) and further statistical analysis was carried out with R package using the same approach as used for microbiota analysis, except that FDR was not performed given the low number of samples per group. P < 0.05 was considered as statistically significant.

Acknowledgments

The authors thank the staff at the Axenic Gnotobiotic Unit in McMaster University, Joe Notarangelo, Michael Rosati, and Sarah Armstrong for their assistance in mouse care. The authors thank the clinical team at McMaster University, Pedro Miranda, Andrea Nardelli, and Rajka Borojevic. We also thank the McMaster Genome Facility for technical support with 16S rRNA gene sequencing and NanoString assays.

Disclosure statement

EFV is Member of the Biocodex International and National (Canada) Scientific Review Boards, Member of the Center for Gut Microbiome Research and Education Scientific Advisory Board of the AGA, Secretary of the International Society of the Study of Celiac Disease, and holds grants from Kallyope and Codexis, unrelated to this study.

Data availability statement

All sequencing data have been deposited in the Sequence Read Archive (SRA). 16S rRNA gene sequencing data used in this study can be accessed under BioProject ID PRJNA861238 at https://www.ncbi.nlm.nih.gov/bioproject/PRJNA861238.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19490976.2023.2205425

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This study was funded by a CIHR Project Grant (PJT-478700) and by a Weston Family Foundation Proof-of-Principle grant to EFV and HJG and a CCC-GIA to EFV who also holds a Tier 1 Canada Research Chair in Microbial Therapeutics and Nutrition in Gastroenterology. AS received a Farncombe Postdoctoral Fellowship Award from McMaster University and an Elevate Postdoctoral Fellowship Award. AH received an Elevate Masters research scholarship. JL received an IMAGINE-CIHR-CAG postdoctoral research fellowship. KJ holds a Vanier Canada Graduate Scholarship. WR is supported by the German Research Foundation (Project Number 318346496, SFB1292/2 TP02). AC holds a Paul Douglas Chair in Intestinal Research. PB holds the Richard Hunt-AstraZeneca Chair in Gastroenterology. RR is supported by a CIHR grant (376560).