Vendor-Specific Microbiome Controls both Acute and Chronic Murine Lung Allograft Rejection by Altering CD4⁺Foxp3⁺ Regulatory T Cell Levels

Abstract

Despite standardized post-operative care some lung transplant patients suffer multiple episodes of acute and chronic rejection while others avoid graft problems for reasons that are poorly understood. Using an established model of C57BL/10 to C57BL/6 minor antigen mismatched single lung transplantation we now demonstrate that the recipient microbiota contributes to variability in the alloimmune response. Specifically, mice from the Envigo facility in Frederick Maryland contain nearly double the number of CD4⁺Foxp3⁺ regulatory T cells (T_regs) than mice from the Jackson facility in Bar Harbor Maine or the Envigo facility in Indianapolis Indiana (18 vs. 9 vs. 7%). Lung graft recipients from the Maryland facility thus do not develop acute or chronic rejection. Treatment with broad spectrum antibiotics decreases T_regs and increases both acute and chronic graft rejection in otherwise tolerant strains of mice. Constitutive depletion of regulatory T cells, using Foxp3-driven expression of diphtheria toxin receptor, leads to the development of chronic rejection and supports the role of T_regs in both acute and chronic alloimmunity. Taken together our data demonstrate that the microbiota of certain individuals may contribute to tolerance through T_reg-dependent mechanisms and challenges the practice of indiscriminate broad-spectrum antibiotic use in the perioperative period.

1 |. Introduction

Despite standardized treatment some lung allograft recipients experience multiple episodes of acute rejection while others remain rejection-free. In addition many lose their allograft 5–10 years post-engraftment due to chronic rejection¹. Standard of care management includes routine scheduled perioperative biopsies for the diagnosis of acute rejection, which is then treated with accentuation of immunosuppression in an attempt to offset chronic rejection-mediated graft loss. Nevertheless many people with acute rejection never develop chronic rejection and many of those without acute rejection episodes develop chronic rejection².

Animal modeling offers utility for understanding factors controlling alloimmunity but, similar to human observations, murine models vary widely. In the C57BL/10→C57BL/6 orthotopic left single lung transplant model we and others noted a wide variability in the outcome for both acute and chronic rejection between different laboratories^3–5. Using this strain combination here we report that the outcomes of orthotopic left single lung transplant, performed in the same laboratory, vary widely based on the recipient facility of origin. We specifically demonstrate that the recipient microbiota plays a uniquely dominant role in controlling both acute and chronic rejection. Contrary to established data demonstrating that antibiotic treatment leads to a uniform downregulation of both allo- and tumor-specific immunity^6–9 we now demonstrate that in certain individuals dysbiosis can drastically alter pulmonary and systemic levels of CD4⁺Foxp3⁺ regulatory T cells, thus increasing both acute and chronic rejection.

2 |. Materials and Methods

2.1. Animals

Male C57BL6/J (B6J), C57BL/6NJ (B6NJ), C57BL10/J (B10J), Foxp3^{tm3(DTR/GFP)Ayr}/J (B6^Foxp3DTR), Balb/cJ and B6.SJL/BoyJ CD45.1 congenic mice were purchased from the Jackson Laboratory (Bay Harbor, ME). C57BL10/NHsd (B10E) and C57BL6/NHsd mice were purchased from ENVIGO Indiana or Maryland facility (defined as B6E^Indiana or B6E^Maryland). Germ-free C57BL6/NTac mice were from Taconic (Albany, NY).

2.2. Orthotopic mouse lung transplantation

All animal procedures were performed in compliance with IACUC. Orthotopic left lung transplantations was performed with the combination of C57BL10/J donors and C57BL/6 recipients of different origins as described throughout the text¹⁰. For some studies of chronic rejection mice were treated with diphtheria toxin (50 ng/g intra-peritoneal) as previously described¹¹.

2.3. Antibiotic ablation and transplantation of gut microbiota

Vancomycin hydrochloride was from GoldBio (St. Louis, MO) and Ampicillin, Metronidazole and Neomycin were from Sigma (St. Louis, MO). Antibiotic cocktail (VMNA) was made by dilution in phosphate buffered saline (PBS) to the final concentrations of 500mg/L for Vancomycin and 1g/L for all other antibiotics. 200μl of this mixture was administered to C57BL6 recipient mice via an 18-gauge gavage needle. Antibiotic cocktail was given for 5 consecutive days followed by 2 days’ rest before transplantation or fecal transfer. For some studies, Germ-Free mice or VMNA-induced dysbiosis mice were reconstituted with fecal samples derived from B6J, B6E^Indiana or B6E^Maryland mice. Fecal samples were collected freshly from anus and dissolved in sterile PBS prior to being orally gavaged into mice recipients¹².

2.4. Histology

A lung pathologist blinded to the experimental condition graded acute rejection according to the International Society of Heart and Lung Transplantation (ISHLT) criteria¹³. Chronic rejection evaluation was modified from a recently described grading system by two separate individuals blinded to the experimental conditions³.

2.5. Flow Cytometry

For flow cytometric analysis, native or allograft lung tissue was well minced with a homogenizer and digested by placing it into RPMI 1640 medium (ThermoFisher,Waltham, MA) containing 0.5mg/ml collagenase II (Worthington Biochemical Corporation, Lakewood, NJ) and 5U/ml DNAse (Sigma, St. Louis, MO). The digested lung tissue was then stained using surface and intracellular primarily conjugated antibodies (full list of clones in supplemental methods).

2.6. Microbiome 16S rRNA gene sequencing and analysis

The fecal samples were collected from mice and underwent total DNA extraction with the DNeasy PowerSoil Kit (QIAGEN, Hilden, Germany) following the manufacturer’s instruction. Details of qPCR and analysis are presented in supplemental methods.

2.7. Statistics

Student’s t-test was used for continuous variable comparisons while the Mann-Whitney U test was used for categorical variable comparisons. Differences were considered significant at p < 0.05. p <0.01 is used as cut-off value in taxonomic differentiation between groups. Data visualization in all figures was accomplished by R (v.3.5.3) and GraphPad Prism 8.1.

3 |. Results

3.1. Antibiotic-induced dysbiosis eliminates vendor-specific differences in acute lung allograft rejection

B6 mice provide a unique platform to study murine lung transplantation due to consistent availability from multiple vendors and availability of multiple transgenic strains. We utilized donors and recipient mice from The Jackson Laboratory facility in Maine (B6J), or ENVIGO facilities in Indiana (B6E^Indiana) or Maryland (B6E^Maryland) based on their use in previous publications^3,14,15. We specifically focused on immune responses in the C57BL10(B10) to C57BL6(B6) model of left single lung transplantation due to well-described phenotypic changes of acute and chronic rejection that resemble human pathology^3,7,14,15.

Seven days post-engraftment none of the grafts had gross or histologic evidence of collapse, parenchymal or obliterative airway fibrosis (O-AF)(Figure 1a). B10 grafts placed into B6J or B6E^Indiana recipients, however, demonstrated severe inflammatory changes with high levels of vascular cuffing and increased numbers of graft infiltrating T cells compared to grafts placed into B6E^Maryland recipients (Figure 1b). Lungs engrafted into B6E^Maryland recipients, however, were virtually free of inflammation (Figure 1a, b). Donor origin did not affect immune responses (Supplemental Figure 1a). We next performed whole exome sequencing of genomic DNA to determine if strain-specific genetic differences could play a role in disparate immune responses. We noted significant genetic similarity but also some differences between the three strains with minor variants, including SNVs (single nucleotide variation) and InDels (insertions and deletions) unique to each strain (Supplemental Figure 1b). Nevertheless, limited polymorphism was evident overall and none in loci of genes described as important to the transplant immune response such as CD4, CD8, Foxp3 or IL-17a (Supplemental Figure 1c). We thus decide to explore other factors that may contribute to strain-specific differences.

FIGURE 1.

Allo-immune response in the three strains of C57BL/6 recipients in presence or absence of VMNA-induced dysbiosis. (A) Gross pictures (first row), H&E stained histology pictures (second row) and ISHLT A grades (third row). (B) Total CD8⁺ and CD4⁺ T cell counts in lung grafts. Lung allografts were harvested day 7 post-engraftment. TXP: lung allografts. *p < 0.05, **p < 0.01, ***p < 0.001, NS: not significant.

To explore environmental factors, such as the microbiome, we next pre-treated recipient mice with oral vancomycin, metronidazole, neomycin and ampicillin (VMNA) antibiotics. Interestingly differences in cellular rejection, as well as graft T cell infiltration, disappeared when dysbiosis was induced with such antibiotic treatment (Figure 1a, b). This was due to a prominent increase in inflammation of B10 grafts placed into B6E^Maryland recipients where both cellular infiltration as well as rejection grade increased after antibiotics (Figure 1a, b). Inflammatory changes remained unchanged in B6E^Indiana recipients irrespective of antibiotic use while B6J recipients demonstrated a trend towards increased CD8⁺ T cell infiltration of the lung graft after antibiotic treatment (571,786± 78,043 without antibiotics vs. 782,515 ± 196,356 with antibiotic treatment). Taken together, these data indicate that the microbiota plays a prominent role in controlling acute cellular rejection. These data further suggest that antibiotic mediated alteration of the microbiome may not uniformly decrease the degree of the cellular immune response to allografts, as previously described^6,7, but may be context, recipient, and facility dependent.

3.2. Vendor-specific CD4⁺Foxp3⁺ regulatory T cell abundance is microbiome dependent

We next considered the possibility that low-level inflammatory changes present in the recipients prior to transplantation may be responsible for differences between the three sets of mice. We evaluated resting lungs and saw no inflammation in any of the strains at baseline (Figure 2a). Flow cytometric analysis of resting lungs showed similar numbers of CD8⁺ T cells but B6E^Maryland mice showed somewhat fewer lung resident CD4⁺ T cells (Figure 2a). In addition neither CD4⁺ nor CD8⁺ cells in the B6E^Maryland mice underwent significant expansion after transplantation. (Supplemental Figure 2a). Resting lung CD4⁺ T cells from B6E^Maryland mice had a significantly higher proportion of Foxp3 expressing CD4⁺ regulatory T cells (T_regs) compared to B6J or B6E^Indana mice (Figure 2b). B6J mice had a higher proportion of T_regs compared to B6E^Indiana mice (Figure 2b). Such differences were evident systemically in the spleen as well as peripheral lymph nodes and not just limited to the lungs (Supplemental Figure 2b). Transplantation accentuated these differences even further with close to 30% of CD4⁺ T cells expressing Foxp3 in B10J to B6E^Maryland grafts (Figure 2c). Pretreatment of recipients with antibiotics decreased Foxp3 expression in CD4⁺ T cells of B6E^Maryland as well as B6J-resident grafts but did not alter the expression in B6E^Indana mice with already very low levels of T_regs (Figure 2c). It has been described that bacterial flora, specifically in the gut, can influence the generation of Foxp3⁺ regulatory T cells by modulating levels of systemic TGF-β¹⁶. In our system antibiotic treatment decreased TGF-β levels in B6E^Maryland mice but did not alter its expression in the other strains. It is thus unlikely that microbiome-mediated TGF-β is solely responsible for vendor specific differences in T_reg levels (Supplemental Figure 2c).

FIGURE 2.

Quantitative differences of CD4⁺Foxp3⁺ regulatory T cells in resting recipient lungs and lung allografts. (A) H&E stained histology pictures, total CD4⁺ and CD8⁺ T cell counts in resting lungs of the three different recipient mice. (B) Percentage of Foxp3⁺ cells among CD4⁺ T cells in resting lungs. (C) Percentage of Foxp3⁺ cells among CD4⁺ T cells in lung grafts harvested at day 7 after engraftment. (D) Expression of CD25 and GITR on CD4⁺Foxp3⁺ T cells. (E) Mixed lymphocyte reactions demonstrating inhibition of B6 CD8⁺ T cell proliferation, as determined by Ki-67 expression, in the presence of Balb/c dendritic cells and T_regs isolated from B6J and B6E^Maryland mice. *p < 0.05, **p < 0.01, ***p < 0.001 NS: not significant.

Phenotypic analysis of T_regs from B6J and B6E^Maryland mice demonstrated similar expression of CD25 and GITR (glucocorticoid-induced TNFR-related protein) indicating similar level of activation (Figure 2d). In an in vitro inhibition assay no significant differences in inhibitory capacity were evident between B6J and B6E^Maryland -derived T_regs. Thus, we can conclude that strain-specific microbiota controls the quantitative but not necessarily qualitative differences in T_regs from select facilities.

3.3. Conventional mechanisms of fecal transfer do not recapitulate the B6E^Maryland T_reg high phenotype

Previous investigators have manipulated the murine microbiome through cohousing or forced fecal gavage in order to introduce select bacteria into a host^6,12,17. We thus gavaged germ-free mice with fecal pellets obtained from either B6J, B6E^Maryland, or B6E^Indiana mice and established colonies from such reconstituted strains. Surprisingly such therapy failed to recapitulate the microbiome-dependent T_reg differences described in Figures 2b above (Figure 3a). We next performed similar experiments in conventional mice through VMNA-induced dysbiosis in B6J and B6E^Indiana mice followed by fecal reconstitution. No differences were evident in B10 graft recipients reconstituted with B6^Maryland vs. autologous facility-identical fecal samples (Figure 3b). To expand on this further, we next co-housed B6J and B6E^Maryland mice for a period of 2 months prior to challenging these two strains with a B10J allograft. However, we noted that the trend for increased inflammation and rejection in B6J recipients persisted while B6E^Maryland recipients still manifested only low levels of inflammation (Supplemental Figure 3a, b). Taken together we can conclude that the vendor-specific differences in our system are complex in nature and are not easily or completely transmitted through fecal transfer.

FIGURE 3.

Conventional mechanisms of fecal transfer to recapitulate the B6E^Maryland phenotype. (A) Percentage of Foxp3⁺ cells among CD4⁺ T cells in germ-free C57BL/6 mice colonies, which were reconstituted with fecal samples derived from B6J, B6E^Maryland and B6E^Indiana mice. (B) B6J or B6E^Indiana recipients were treated with VMNA antibiotics and reconstituted with B6^Maryland vs autologous facility-identical fecal samples, respectively, prior to engraftment of B10J donor lungs. Allograft rejection is evaluated by gross appearance, H&E staining, ISHLT A grade and flow cytometry analysis of lung allograft-infiltrating T cells and T_reg differentiation. NS: not significant.

3.4. B6^Maryland mice demonstrate bacterial diversity with a preponderance of unique tolerogenic Firmicutes

Bacteroides fragilis is known to facilitate T_reg generation in the gut and can be difficult to inoculate as part of fecal transfer¹⁸. We thus suspected that differences in B. fragilis may be responsible for altered T_reg levels but we were unable to detect differences in this bacteria between the three facilities (Supplemental Figure 4a). Taken together with the data described above we deduced that the relationship between the microbiome and tolerance in our system was more complex than a simple one to one relationship.

We next utilized 16S rRNA metagenomics to characterize fecal flora by both quantitative and qualitative analysis of OTUs (operational taxonomic units) as previously described^19–21. We first looked at the total abundances of all the OTUs in the fecal samples and noted no significant differences in total bacterial load (Figure 4a). However by evaluating the Shannon diversity index, which measures both the abundance and evenness of the species present²², we noted that both B6E^Maryland and B6E^Indiana mice have significantly more diverse composition of gut microbiota than the B6J strain (Figure 4a). Principal coordinates analysis (PCoA) of beta diversity, or similarity between the samples, also demonstrated that mice from the two ENVIGO facilities share greater similarity with each other compared to the B6J strain (Figure 4b). Relative phylum abundance analysis, filtering out the phyla that comprise <1% of the microbiome, demonstrated that Bacteroidetes and Firmicutes were the dominant phyla in all three strains of mice (Figure 4b). B6E mice contain more Bacteroidetes and less Firmicutes compared to B6J (Figure 4b). Given that B6E^Indiana mice’s gut microbiota seemed highly analogous to B6E^Maryland mice, but the two strains demonstrate major differences in T_reg profiles (Figure 2b, c), we next performed a more detailed analysis at the order level and noted that B6E^Maryland mice possessed an undefined population of Bacteroidales that is significantly less abundant in B6J or B6E^Indiana mice (Supplemental Figure 4b).

FIGURE 4.

16s-rRNA analysis for the gut microbiota. (A) Total abundances of OTUs and Shannon Diversity Indices of resting mice (n=5). (B) Principle Component Analysis (PCoA) and relative abundances of phylum in resting mice (n=5). (C) Total abundances of OTUs and Shannon Diversity Indices of non-treated vs antibiotic treated mice (n=5). (D) Algorithm of identifying T_reg related OTUs. The volcano plots demonstrate the two comparisons, cut-off values are −2 and +2 for the x-axes and 2 for the y-axes. The heat-map shows OTU abundances in different groups. Highlighted bacterial families were previously reported to induce T_reg in a TGF-β dependent manner. *p < 0.05, **p < 0.01, ***p < 0.001, NS: not significant.

To narrow this down even further we took advantage of the fact that T_regs are attenuated by antibiotic treatment in both B6E^Maryland and B6J mice. We thought it likely that the unique bacterial signature responsible for T_reg expansion may be deduced by studying fecal samples in the presence or absence of antibiotics. We thus examined in parallel B6E^Maryland and B6J mice in the presence or absence of antibiotics by total OTU abundance and Shannon diversity index (Figure 4c). We noted that the overall OTUs were not impacted while diversity was altered by antibiotic treatment, specifically in B6E^Maryland mice (Figure 4c). After antibiotic treatment the Shannon diversity index of B6E^Maryland mice became similar to that of resting B6J mice. We additionally performed overlapping component analysis (OCA) by analyzing taxonomy to identify antibiotic sensitive bacteria unique to the B6E^Maryland strain that are not present in either B6J or B6E^Indiana mice. Such an algorithm narrowed our search to 32 unique OTUs encompassing various phylum including similar Clostridium family members of the Firmicutes phylum that have been described to induce T_regs²³ (Figure 4d). Taken together we can conclude that polymicrobial factors are likely to contribute to differences evident in T_reg mediated immune responses and seen in various strains of B6 mice from different vendors but it is likely that Clostridia or other members of the Firmicutes phylum play a dominant role.

3.5. Chronic lung allograft rejection varies based on facility of origin and antibiotic treatment

The B10 to B6 model of lung transplantation can be used to study not only acute but chronic rejection as well, since changes resembling obliterative bronchiolitis (OB) or obliterative airway fibrosis (O-AF) occur in this strain combination^3,14. We grafted B10J lungs to B6J, B6E^Maryland or B6E^Indiana mice and evaluated changes by histology 21 days after engraftment. Similar to previously described data B6E^Indiana recipients demonstrated evidence of chronic rejection but neither B6J nor B6E^Maryland recipients demonstrated evidence of O-AF (Figure 5a). Lack of chronic rejection in B6E^Maryland recipients was understandable due to the lack of acute inflammatory changes. However, it was somewhat puzzling for the B6J recipients not to have evidence of chronic rejection due to severe acute rejection in this strain on day 7 (Figure 1a) and day 21 post engraftment (Supplemental Figure 5).

FIGURE 5.

The severity of chronic rejection differs between the three strains of mice and is altered by antibiotic treatment. (A) Gross pictures, low power/200X Masson’s Trichrome staining and H&E staining of lung allografts and levels (O-AF scores plotted at the bottom) of chronic allograft rejection on untreated mice. (B) Gross pictures, low power/200X Masson’s Trichrome staining and H&E staining of lung allografts and O-AF scores on antibiotic-treated B6J and B6E^Indiana recipients. Arrow: typical lesions of obliterative airway fibrosis. TXP: lung allografts. **p < 0.01, ***p < 0.001.

To further expand on this model we next evaluated O-AF in VMNA-treated B6 recipients of B10J lung grafts 21 days after engraftment. As the B6E^Maryland mouse colony was terminated by ENVIGO and could not be utilized for further experiments²⁴ we relied on B6J and B6E^Indiana mice to obtain these data. Treatment with antibiotics induced severe O-AF changes in B6J recipients, including grade 2 to 4 airway lesions which we did not detect in any other transplant combination (Figure 5b). The severity of chronic rejection decreased in B6E^Indiana recipient mice (Figure 5b) consistent with previously published results in mice from this facility⁷. Taken together we can conclude that early inflammatory changes do not necessarily lead to chronic lung allograft rejection and antibiotic-induced dysbiosis can have variable microbiome-specific effects on chronic lung graft rejection.

3.6. IL-17a/Foxp3 ratio can predict chronic lung allograft rejection

IL17a production and Th17 polarization of the microenvironment has been linked to chronic lung allograft fibrosis in both murine models and human observational studies⁵. It has also been demonstrated that neutralization of IL-17a can ameliorate graft fibrosis and development of O-AF⁴. We hypothesized that IL-17a-dependent O-AF development is controlled by vendor origin and dysbiosis. While T cells of both B6E^Indiana and B6E^Maryland-derived mice elaborated significant levels of IL-17a, B6J mice had much lower levels (Figure 6a). Furthermore, and unlike the case for Foxp3, IL-17a expression did not change after antibiotic treatment (Figure 6a). We found it somewhat surprising that B6E^Maryland recipients did not develop O-AF despite the presence of high numbers of IL-17a producing T cells, similar to those found in B6E^Indiana mice which did develop BO-like lesions (Figure 5a). In addition, B6J mice demonstrated low numbers of IL-17a producing T cells but developed O-AF after antibiotic treatment.

FIGURE 6.

Relative IL-17a levels and IL-17a/Foxp3 ratios in lung allografts. (A) Flow cytometry analysis of IL-17a expression in different compartments of T cells in lungs engrafted into untreated or antibiotic-treated recipients. (B) IL-17a/Foxp3 ratio within T cells in lung grafts. Lung allografts were harvested at day 21 after engraftment. p < 0.05, **p < 0.01, ***p < 0.001 NS: not significant.

Since T_reg levels change after antibiotic treatment we next considered the possibility that the ratio of IL-17a producing T cells to T_regs, rather than the absolute number of Th-17 polarized T cells may determine the development of O-AF. We thus plotted the relative IL-17a/Foxp3 ratios in unmanipulated vs. antibiotic treated mice. We were forced to exclude B6E^Maryland mice from such analysis due to the loss of this strain as described above²⁴. Indeed this ratio correlated with O-AF as antibiotic-treated B6J mice with the highest IL-17a/Foxp3 ratio had the highest O-AF grade while untreated B6J recipients, that did not develop airway fibrosis, demonstrated very low IL-17a/Foxp3 ratios (Figure 6b). Intermediate IL-17a/Foxp3 ratio in B6E^Indiana recipient mice correlated with mild but evident O-AF.

In order to evaluate whether T_regs directly modulate chronic lung allograft rejection we took advantage of a transgenic strain of B6 mice (B6^Foxp3DTR)²⁵ where ablation of T_regs can be accomplished through the administration of diphtheria toxin in the absence of microbiome manipulation. Such a system allows for a targeted reductionist approach to study IL-17a/T_reg relationship in fibrosis without other confounding factors. We thus treated B6^Foxp3DTR or B6NJ wild-type control mice with DT post-engraftment of a B10 lung graft. Twenty-one days later B10 grafts placed into B6^Foxp3DTR recipients had evidence of both O-AF and pulmonary parenchymal fibrosis consistent with chronic rejection (Figure 7a-c) within the background of significant acute cellular rejection (Supplemental Figure 6). As expected elimination of T_regs was evident in B6^Foxp3DTR mice, resulting higher IL-17a/Foxp3 ratios (Figure 7d). Taken together with previously published data that neutralization of IL-17a can ameliorate airway fibrosis^4,5 we can now conclude that T_regs play a protective role against O-AF and that IL-17a/Foxp3 ratios can be used to predict the severity of chronic lung allograft rejection.

FIGURE 7.

Impact of T_reg ablation on chronic lung allograft rejection. (A) Gross pictures, low power/200X Masson’s Trichrome staining and H&E staining of lungs engrafted into B6^FoxpDTR and B6NJ control mice. Both strains received DT treatment after engraftment as described. Evaluation of chronic lung allograft rejection at day 21 post-engraftment by O-AF score (B) and parenchymal fibrosis score (C). (D) Flow cytometry analysis of mediastinal lymph nodes shows the ablation of CD4⁺Foxp3⁺ T_regs, stable IL-17a levels and significantly altered IL-17a/Foxp3 ratios. *p < 0.05, **p < 0.01, ***p < 0.001 NS: not significant.

4 |. Discussion

Commensal bacteria are a normal component of the mammalian gastrointestinal tract, as well as other mucosal tissues, and play an important role in both health and disease. Dysbiosis, or microbial imbalance, can be induced in otherwise healthy laboratory mice by treatment with antibiotics. It has been previously demonstrated that in C57BL/6 mice, specifically obtained from the ENVIGO Indiana facility, induction of dysbiosis with broad spectrum antibiotics can alter antigen presenting capacity of myeloid cells ameliorating rejection of both skin and heart grafts due to impaired priming of alloreactive T cells⁶. Similar findings have been described for chronic lung allograft rejection in a murine model⁷. These data are supported by human observational studies that use of antibiotics can downregulate the immune response and decrease the efficacy of tumor immunotherapy⁹. Based on such preclinical data it has become an assumption that alteration of the microbiome with broad spectrum antibiotics may be used as an easily implemented mechanism of immunomodulation in human transplant recipients.

In contrast to this notion we now demonstrate that normal variability in the recipient microbiome may have drastically differential effects on graft acceptance or rejection. While we are able to replicate data from Chicago-based investigators, and demonstrate downregulation of chronic lung allograft rejection in antibiotic-treated B6^Indiana mice⁷ (Figure 5), similar treatment of Maryland and Jackson-derived mice resulted in an increase in acute and chronic rejection respectively (Figures 1 and 5). Thus, our data closely resemble recent reports describing that individual variability in the microbiome of melanoma patients contributes to the success or failure of PD-1 and CTLA-4-mediated immunotherapy²⁶ and supports the recent report by Rey and colleagues demonstrating exacerbation of acute vascular rejection after disruption of the gut microbiome with antibiotics²⁷. Our data further expand on the complexity that the commensal microbiome plays in orchestrating the homeostasis of the immune system^28–32. Unlike other investigators we were unable to completely recapitulate the microbiome-dependent tolerance of Maryland-derived mice through cohousing or fecal transfer (Figure 3, Supplemental Figure 3). Such data, combined with 16S rRNA sequencing identifying multiple bacterial species associated with tolerance, indicate a complex interrelationship between T_reg levels and the microbiome. Alternatively it is possible that unique genetic polymorphisms present in mice from different facilities may uniquely interact with the microbiome to facilitate graft acceptance³³. While we focused on analysis of gut specific microbiome by 16S rRNA sequencing of fecal pellets it remains to be determined whether the gut or pulmonary microbiome plays a dominant role in lung allograft survival. While multiple aspects of the gut microbota have been demonstrated to contribute to pulmonary as well as systemic immune responses^7,34, an ever-growing understanding of the lung microbiome and the unique lung-gut axis suggests that local interactions may shape graft outcome as well³⁵.

We found it interesting that microbiome-dependent downregulation of the immune response, in both Maryland and Jackson-derived mice, correlated with relatively higher proportion of Foxp3-expressing T_regs. Such data support the well-described, and at times bidirectional, interrelationship of this cell population with comensal organisms^18,23. In addition to such correlative studies we definitively demonstrate that T_regs can prevent chronic lung allograft rejection, including obliterative airway fibrosis or bronchiolitis. Such data shed light on the often conflicting human observational studies describing a role for T_regs in the development of chronic lung allograft dysfunction^36–38. Interestingly, our data suggest that it may not be the absolute number but rather the ratio between regulatory and pathogenic T cells that affects graft fibrosis. This is clearly evident in Maryland mice with high numbers of both IL-17a producing as well as Foxp3-expressing T cells as well as Jackson mice containing low numbers of both. Only when this balance is perturbed does chronic rejection set in. Combined with previous reports demonstrating that ablation of IL-17a production can prevent chronic lung allograft disfunction⁴, we can conclude that strategies aimed at either augmenting T_regs or inhibiting Th-17 differentiation may be utilized concurrently or interchangeably to improve long-term graft survival. However it is important to point out that even such principles maybe individual strain and context dependent and it is possible that in certain strains of mice, such as those from the Envigo Indiana facility, other factors may control fibrosis⁷.

It did not escape us that neither early nor late peri-bronchial inflammation uniformly predicted airway fibrosis. Specifically, mice from Jackson demonstrate severe inflammatory changes post-engraftment but fail to develop fibrosis in the absence of microbiome ablation (Figure 1a, 5a). While on the surface such data seem counterintuitive to reports demonstrating a correlation between the number of acute rejection episodes and bronchiolitis obliterans^39–41, it does support the notion– that the quality of cell infiltrates rather than just quantity may control long-term allograft fate. Specifically low levels of TGF-β and IL-17a in Jackson mice may set the stage for inflammation in the absence of a profibrogenic environment. Such data is in line with clinical strategies espoused by some that histologically detected A1 rejection may not uniformly lead to chronic lung allograft dysfunction^40,41. Such knowledge also begs for assessment of routine post-operative biopsies for factors other than simple histologic inflammation.

Abbreviations:

CTLA-4

cytotoxic T-lymphocyte-associated protein 4

GITR

glucocorticoid-induced TNFR-related protein

H&E

haemotoxylin and eosin

ISHLT

International Society for Heart and Lung Transplantation

O-AF

obliterative airway fibrosis

OB

obliterative bronchiolitis

OTU

operational taxonomic units

PBS

phosphate buffered saline

PD-1

Programmed cell death protein 1

rRNA

ribosomal ribonucleic acid

TGF-β

transforming growth factor beta

Tregs

regulatory T cell

VMNA

vancomycin, metronidazole, neomycin and ampicillin