Thrombospondin-1 Restricts Interleukin-36γ-Mediated Neutrophilic Inflammation during Pseudomonas aeruginosa Pulmonary Infection

Pseudomonas aeruginosa pulmonary infection can lead to exaggerated neutrophilic inflammation and tissue destruction, yet host factors that regulate the neutrophilic response is not fully known. IL-36γ is a proinflammatory cytokine that dramatically increases in bioactivity following N-terminal processing by proteases.


RESULTS
Thrombospondin-1 limits excessive proinflammatory cytokine production and neutrophil-dominant immune cell recruitment during acute P. aeruginosa intrapulmonary infection. We previously reported increased lung bacterial burden and exaggerated neutrophilic inflammation in Thbs1 2/2 mice at 20 h postinfection with P. aeruginosa compared to wild-type (WT) mice (27). To better understand the mechanism underlying the early inflammatory response, we examined the kinetics of infection at 5 h postinfection (hpi), when no differences in bacterial burden were detected, and at 1 day postinfection (dpi), where the absence of TSP-1 resulted in increased bacterial burden in the lungs (Fig. 1A). At 5 hpi, both WT and Thbs1 2/2 mice experienced a rapid increase in several proinflammatory cytokines, such as IL-6, CXCL-1, CXCL-2, granulocyte-macrophage colony-stimulating factor (GM-CSF), granulocyte colony-stimulating factor (G-CSF), IL-1b, and IL-17A (Fig. 1B to H). Although both Thbs1 2/2 and WT mice produced similar amounts of most cytokines analyzed at 5 hpi, WT mice produced higher levels of IL-6 at this time postinfection (Fig. 1B), indicating early differences in the immune response of these mice against P. aeruginosa. At 1 dpi, however, Thbs1 2/2 mice showed increased levels of IL-6, CXCL-1, CXCL-2, GM-CSF, G-CSF, IL-1b, and IL-17A in the lungs at 1 dpi compared to WT mice ( Fig. 1B to H). We observed sustained myeloperoxidase (MPO) content in the lungs at 5 hpi and 1 dpi in Thbs1 2/2 mice, in contrast to WT mice, where lung MPO peaked at 5 hpi but was downregulated by 1 dpi (Fig. 1I). Furthermore, Thbs1 2/2 mice showed increased lung microvascular permeability, as evidenced by increased total bronchoalveolar lavage fluid (BALF) protein compared to that in WT mice (Fig. 1J). The early production of cytokines and chemokines induced a robust recruitment of leukocytes to the BALF of WT and Thbs1 2/2 mice. Flow cytometry analyses enabled determination of immune cell infiltration into the airspaces of WT and Thbs1 2/2 mice, where neutrophil numbers showed a mild and nonsignificant increase at 5 hpi compared with 0 hpi but were significantly elevated by 1 dpi (Fig. 1K; see also  Table S1 in the supplemental material). Notably, other cells, such as resident alveolar macrophages, eosinophils, classical (Ly6C 1 ) and alternative (Ly6C 2 ) monocytes, dendritic cells (DCs), B cells, T cells, and monocyte-derived macrophages, were found in the BALF of both WT and Thbs1 2/2 mice (Fig. 1K). However, Thbs1 2/2 mice exhibited increased numbers of neutrophils, alveolar macrophages, eosinophils, Ly6C 1 monocytes, and monocyte-derived macrophages ( Fig. 1L to P). Thbs1 2/2 mice also showed elevated numbers of T cells and B cells but equivalent numbers of CD11b 1 DCs, CD11b 2 DCs, and Ly6C 2 monocytes (see Fig. S3 in the supplemental material). These data suggest that TSP-1 restrains proinflammatory response by 1 dpi in the lungs and limits excessive recruitment of neutrophils and other myeloid cells into the airspaces, enhancing P. aeruginosa clearance and reducing lung injury.
Thrombospondin-1 does not alter IL-36c expression induced by P. aeruginosa infection but restrains the proteolytic activity of NE and LasB that can mediate IL-36c cleavage at distinct sites. IL-36 cytokines are major effectors of the immune response in the lungs during P. aeruginosa and other bacterial infections (30,31,34). Il36a transcript level was increased in the lungs by 1 dpi (Fig. 2A), but no changes in expression were detected for Il36b (Fig. 2B). Il36g transcript level peaked at 5 hpi but remained increased above baseline at 1 dpi (Fig. 2C). However, there were no differences in Il36a, Il36b, and Il36g transcriptional responses between WT and Thbs1 2/2 mice. In addition, we noted increased levels of IL-36g protein in the lungs at 1 dpi in both WT and Thbs1 2/2 mice ( Fig. 2D and E). As N-terminal processing of IL-36g is required for full bioactivity and the triggering of proinflammatory cytokines IL-6 and CXCL-1 by IL-36R in bone marrow-derived dendritic cells (BMDCs) and human keratinocytes (39,40), we show that cleaved IL-36g (cIL-36g) just proximal to S 18 but not full-length IL-36g (fIL-36g) leads to a robust production of IL-6 and CXCL-1 by BMDCs in vitro (Fig. 2F). The protease responsible for IL-36g cleavage in keratinocytes is cathepsin S (CatS), which cleaves IL-36g into the potent S 18 isoform (35). However, we were unable to identify substantial differences in BALF CatS activity in WT and Thbs1 2/2 mice following monocytes, and (P) monocyte-derived macrophages from WT and Thbs1 2/2 mice at 5 hpi and 1 dpi. *, P , 0.05 for single comparisons; the Shapiro-Wilk test was used to assess normal distribution followed by a Mann-Whitney U test or a parametric t test. A two-way analysis of variance (ANOVA) test was followed by a post hoc test for multiple comparisons over time. Each data point represents an individual mouse, combined from two independent experiments. Lines indicate the median. , and (C) Il36g transcripts were measured at 5 hpi and 1 dpi by quantitative reverse transcription-PCR (qRT-PCR) using gadph as the internal housekeeping gene. (D and E) IL-36g expression in the lungs measured by Western blot at 1 dpi. Density expression of IL-36g is normalized to b-actin. (F) IL-6 and CXCL-1 production by bone marrow-derived dendritic cells (BMDCs) after stimulation with full-length (fIL-36g) or cleaved IL-36g (cIL-36g, S 18 isoform). (G) Cathepsin S (CatS), (H) neutrophil elastase (NE), and (I) LasB activity were measured in the BALF of WT and Thbs1 2/2 mice at 5 hpi and 1 dpi using the specific substrates 2-aminobenzoyl-L-alanyl-glycyl-L-leucyl-L-alanyl-para-nitro-benzylamide, N-methoxysuccinyl-Ala-Ala-Pro-Val p-nitroanilide, and Mca-GRWPPMG;LPWEK(Dnp)-D-R-NH2, respectively. *, P , 0.05, for single comparisons, the Shapiro-Wilk test was used to assess normal distribution, followed by a Mann-Whitney U test or a parametric t test. A two-way ANOVA test was followed by a post hoc test for multiple comparisons over time. Each data point represents an individual mouse, combined from two independent experiments, except for the Western blot and in vitro BMDC stimulation, which were performed once. Lines indicate the median. P. aeruginosa infection (Fig. 2G), prompting us to hypothesize that other proteases may be involved in the N-terminal processing and activation of IL-36g in the lungs. Notably, Thbs1 2/2 mice showed higher levels of BALF free NE activity at 1 dpi (Fig. 2H), and higher BALF P. aeruginosa LasB activity at 5 hpi and 1 dpi (Fig. 2I). Although TSP-1 does not regulate IL-36 cytokine gene and protein expression, these data show that TSP-1 tempers the proteolytic environment of the lung, which is potentially required to boost the biological activity of IL-36g protein.
The major neutrophil protease responsible for IL-36g activation in vitro is NE (37), and NE has been reported to cleave IL-36g proximally to Y 16 and Q 17 (here referred to as Y 16 and Q 17 isoforms) (35), although only the Y 16 isoform has been previously shown to be biologically active (35,37). We evaluated whether LasB, a pathogen-derived metalloprotease with elastase activity, can cleave IL-36g. Incubation of human full-length IL-36g (18.7 kDa) with cell-free supernatant of P. aeruginosa grown in culture resulted in the cleavage of IL-36g to a smaller product of approximately 17 kDa (Fig. 3A). Supernatant obtained from a transposon insertion mutant of P. aeruginosa strain PA14 deficient in LasB (PA14lasB::Tn5) (27) or PA14 WT in the presence of a LasB inhibitor reduced the cleavage of IL-36g (Fig. 3A). These findings suggest that the PA-protease LasB can cleave IL-36g. We next compared the cleavage of IL-36g by purified LasB (pLasB) and NE. Our data show that LasB and NE cleave IL-36g at different positions of the N-terminal region (Fig. 3B). N-terminal sequencing was conducted by Edman degradation and showed that LasB cleaved IL-36g just proximally to M 19 (M 19 isoform), whereas NE-mediated cleavage of IL-36g yielded several products with the largest truncated product at Y 16 (Fig. 3C). The latter finding is consistent with previous reports (35)(36)(37)(38).
Sequential N-terminal truncation models in silico predict the bioactivity of the M 19 isoform. The N-terminal truncation of IL-36g just proximal to S 18 (S 18 isoform) amplifies the bioactivity of IL-36g by ;1,000-fold in vitro (33,35). To gain molecular insight into why the bioactivity of IL-36g is dependent on the removal of N-terminal residues, we set out to evaluate in silico how a panel of IL-36g sequential truncation models could bind to the IL-1Rrp2/IL-1RAcP heterodimer (IL-36R) complex. To avoid any steric clashes with IL1RAcP during docking runs, we used a previously reported homology model of the IL-36R complex (41). Since the N terminus before S 18 was unresolved in the IL-36g crystal structure (PDB identifier 4IZE), we employed the deep learning algorithm RaptorX (http://raptorx.uchicago.edu) to compute a model of the fulllength IL-36g (fIL-36g) molecule for truncation, as the three-dimensional (3D) structure of the S 18 -D 169 amino acid sequence of the model generated by RaptorX was a close match to the 4IZE crystal structure with the same sequence. The binding patterns of our collection of N-terminal deletion models were determined in a series of docking trials using ClusPro 2.0. In the first round of docking, we tested the 23, 26, 29, 212, 215, 218, and 221 amino acid IL-36g truncation models. The fIL-36g model (as well as the 23, 26, 29, and 212 models) did not bind in a way deemed productive based on the work of others (41,42). After the coarse sampling, we proceeded to analyze the docking between IL-36R with the Y 16 , S 18 , and M 19 isoforms, produced by NE, CatS, and LasB, respectively.
The M 19 isoform [ClusPro 2.0 job identifier [ID] 456743: cluster 0 (95 members), 21099 weighted lowest energy score] sits in a groove composed of parts from the IL-1Rrp2 D2 and D3 domains. Importantly, the N terminus, next to a short helix (I 104 -G 109 , indicated with a red star), faces the receptor D3 domain (Fig. 3D). In this arrangement, the bound cytokine is upside down relative to the orientation of the IL-36R complex. Moreover, loop L 155 -N 160 is nestled in a shallow pocket in the D2 domain of IL-1RAcP, while loop T 61 -D 72 faces the D3 domain of the accessory protein (Fig. 3D). Interestingly, our upside-down orientation bears a strong resemblance to an earlier prediction of how the 4IZE crystal structure binds to the receptor complex (42). Last, there is a network of favorable electrostatic interactions between the M 19 isoform and the IL-36R complex (Fig. 3E). Our binding model also contains a set of destabilizing electrostatic repulsions (Fig. 3E, black star), potentially suggesting the existence of a complex The repose of the M 19 isoform is almost identical to how the S 18 isoform binds (Fig. 3D) [job ID 456742: cluster 5 (46 members), 2892 weighted lowest energy score]. Thus, we conclude that the S 18 and M 19 isoforms may interact with IL-1Rrp2 in a close, if not identical, fashion, offering a structural rationale for why the pLasB cleavage product, M 19 isoform, might possess strong bioactivity. Finally, we asked how the Y 16 isoform could associate with the IL-36R complex. Y 16 isoform [job ID 456740: cluster 3 (43 members), 2883 weighted lowest energy score] can adopt the basic upside-down binding orientation, but is out of sync by a counterclockwise rotation compared to the position of M 19 isoform helix (I 104 -G 109 ) (Fig. 3F). As for packing, there are fewer atomic contacts between the Y 16 isoform and IL-1RAcP in the IL-36R complex, suggesting the Y 16 isoform has a looser upside-down fit. Our docking-modeling data suggest that electrostatic interactions may contribute to the docking stability of the Y 16 isoform. Unlike S 18 and M 19 isoforms, we do not see any electrostatic repulsions that could influence any potential movement of the IL-1Rrp2 D3 domain (Fig. 3G, black star). This raises the possibility that favorable electrostatic contacts may counteract the looser fit of IL-36g to its receptor complex. This suggestion could prove to have implications for the strength of the IL-36g-mediated inflammatory response.
Neutralization of IL-36c improves lung immunity and inflammatory response against P. aeruginosa in the absence of thrombospondin-1. To evaluate whether the hyperinflammatory response observed in Thbs1 2/2 mice at 1 dpi is mediated by IL-36g, we intraperitoneally (i.p.) injected Thbs1 2/2 mice with a neutralizing rabbit antimouse IL-36g antibody or rabbit IgG (30) at the time point of 5 h during the peak of Il36g expression. IL-36g neutralization at 5 hpi had a major protective effect in Thbs1 2/2 mice, as evidenced by improved P. aeruginosa burden in the lungs (Fig. 4A), a significant reduction of proinflammatory cytokines and chemokines such as CXCL-1, CXCL-2, GM-CSF, and IL-1b (Fig. 4B to E), and a nonsignificant reduction of IL-17A, G-CSF, or IL-6 ( Fig. 4F; see also Fig. S4A and B in the supplemental material). Moreover, IL-36g neutralization reduced free BALF NE (Fig. 4G) and lung tissue MPO (Fig. 4H) activity. Although the infiltration of neutrophils and other immune cells such as eosinophils, macrophages, Ly6C 1 monocytes ( Fig. 4I and J; see also Fig. S1B in the supplemental material), DCs, Ly6C 2 monocytes, and T and B cells (Fig. S4C) were not significantly reduced in the BALF after IL-36g neutralization, these data show that IL-36g mediates the hyperinflammatory response in the lungs of Thbs1 2/2 mice during P. aeruginosa infection and supports the hypothesis that TSP-1 tempers IL-36g activity.
Thrombospondin-1 regulates neutrophil function and proinflammatory cytokine production induced by N-terminally processed IL-36cin the lungs. We next evaluated whether TSP-1 directly regulates the inflammatory effects triggered by IL-36g in the lungs. As cleaved IL-36g, but not full-length IL-36g, induced the production of IL-6 and CXCL-1 in murine BMDCs (Fig. 2F), we intratracheally delivered cIL-36g (S 18 isoform) to WT and Thbs1 2/2 mice. One day posttreatment, we evaluated the influx and activation of neutrophils, as well as cytokine and chemokine production. Instillation of cIL-36g induced robust neutrophil recruitment to the airspaces, as measured in the BALF of WT and Thbs1 2/2 mice (Fig. 5A). However, Thbs1 2/2 mice showed higher airspace neutrophil counts (Fig. 5B) and elevated free NE activity (Fig. 5C) and lung tissue MPO activity (Fig. 5D) compared to that in WT mice. In addition, Thbs1 2/2 mice showed exaggerated proinflammatory chemokine and cytokine response in the lungs, including CXCL-1, CXCL-2, GM-CSF, and IL-1b (Fig. 5E to H). G-CSF and IL-6 ( Fig. 5I and J) were induced by cIL-36g but were not significantly different between WT and Thbs1 2/2 mice, whereas IL-17A was not induced by cIL-36g (Fig. 5K). The findings indicate that, in the absence of  the Shapiro-Wilk test was used to assess normal distribution, followed by a Mann-Whitney U test or a parametric t test. A two-way ANOVA test was followed by a post hoc test for multiple comparisons over time. Each data point represents an individual mouse and two independent experiments. Lines indicate the median. TSP-1, neutrophil recruitment, activation, and inflammatory cytokine production induced by cIL-36g are amplified in the lungs.

DISCUSSION
Our findings indicate TSP-1 regulates the inflammatory response in the lungs mediated by IL-36g during P. aeruginosa lower respiratory tract infection by restraining the extracellular proteolytic environment. Compared to WT mice, Thbs1 2/2 mice developed a hyperinflammatory response in the lungs during P. aeruginosa infection that is characterized by enhanced production of proinflammatory cytokines and chemokines, as well as by increased influx of neutrophils and other leukocytes. While IL-36g is induced in the lungs early during infection, WT and Thbs1 2/2 mice showed similar levels of IL-36g in both transcript and protein expression. Others have shown that the bioactivity of IL-36 cytokines requires N-terminal processing by proteases such as NE in vitro (37), and here we show that cleaved IL-36g (S 18 isoform), but not full-length IL-36g, induced IL-6 and CXCL-1 production in BMDCs. Instillation of cIL-36g recapitulated the amplified proinflammatory cytokine and chemokine response and enhanced neutrophil influx and activation observed with P. aeruginosa infection in Thbs1 2/2 mice. Moreover, IL-36g neutralization reduced the production of proinflammatory cytokines and free neutrophil NE activity in the lungs of Thbs1 2/2 mice and paradoxically improved the ability of Thbs1 2/2 mice to clear P. aeruginosa in the lungs. Together, our data provide evidence that TSP-1 tempers the hyperinflammatory response during P. aeruginosa lung infection by regulating IL-36g bioactivity and restraining feed-forward inflammation.
Once full-length IL-36g is secreted to the extracellular space, host proteases cleave the protein and thereby increase its bioactivity ;500to 1,000-fold, allowing IL-36g to bind to the IL-36R complex and trigger inflammation (33,(35)(36)(37). NE, proteinase-3, and CatS are host proteases that cleave and activate IL-36g in vitro (33,(35)(36)(37). However, little is known regarding pathogen-derived proteases that can directly cleave the IL-36 family of cytokines and about the regulation of extracellular proteases in vivo. Given the notable increase in free NE and LasB activity in the lungs of Thbs1 2/2 mice, we examined N-terminal processing of IL-36g by NE and LasB and show that NE and LasB cleave IL-36g just proximally to Y 16 and M 19 , respectively. Sequential truncation experiments performed in silico predict that the M 19 isoform and bioactive S 18 isoform show a similar binding pattern to that of the IL-36R complex, a heterodimer formed by IL-1Rrp2 and IL1RAcP. Mechanistically, the binding of IL-36g to IL-1Rrp2 prompts the recruitment of IL1RAcP. One study has proposed that the binding of IL-36g can drive the intermolecular association of the D3 domains of the receptor and accessory protein (41). Consequently, their Toll/interleukin-1 receptor (TIR) domains (tethered to D3s), along with the TIR domain of the adaptor protein MyD88, form a signaling platform comprised of an intermolecular TIR domain trimer. This leads to the activation of NF-κB, which will traffic to the nucleus to modulate the transcription of a set of genes, including those that encode proinflammatory cytokines (41). Although our docking models are static snapshots, the concentration of like charges in close physical proximity raises the possibility that electrostatic interactions may play a role in D3 domain association. In our model, the activity of CatS-the protease responsible to process IL-36g into the S 18 isoform (35)-was not significantly increased in the airspaces following P. aeruginosa infection. CatS contributes to several processes in the extracellular space, including degradation of the extracellular matrix (43,44). However, CatS also participates in the intracellular antigen processing required for MHC-I and MHC-II class antigen presentation (45). Therefore, it is possible that CatS may be involved in the intracellular processing of full-length IL-36g, although this remains to be seen.
Mice deficient in TSP-1 show exaggerated neutrophilic response to cleaved IL-36g, which suggests the existence of a feed-forward mechanism in which neutrophils that arrive into the airspaces release more proteases to further amplify inflammation. Following intratracheal administration, cleaved IL-36g leads to enhanced production of chemokines and cytokines, presumably by resident IL-36R 1 cells in the lungs (31,46,47). We suggest that recruited neutrophils release proteases into the airspaces that cleave and activate de novo synthesized IL-36g that further drives neutrophil recruitment. We further suggest that TSP-1 limits this feed-forward inflammation mediated by IL-36g by restraining the extracellular proteolytic environment in the lung during P. aeruginosa infection. What might trigger the initial N-terminal processing of IL-36g during infection? It is conceivable that a pathogen-secreted protease could serve this role. Indeed, our findings show that a pathogen protease, LasB, can cleave IL-36g to generate the M 19 isoform, which is predicted to have a similar binding pattern to IL-36R as the bioactive S 18 isoform of IL-36g. Together, these data indicate that the restraint of host and pathogen proteases that cleave full-length IL-36g into different isoforms is a key checkpoint that regulates the biological activity of IL-36g. In this context, our data suggest that, during P. aeruginosa infection, the inhibitory properties of TSP-1 over NE and LasB control the generation of IL-36g isoforms implicated in runaway neutrophilic inflammation.
P. aeruginosa-infected patients with ARDS show increased BALF and plasma levels of IL-36g (31). ARDS is a heterogeneous syndrome, and at least two endotypes of ARDS have been identified, a hyperinflammatory and a hypoinflammatory endotype (48). The hyperinflammatory endotype is characterized by a robust production of IL-8, IL-6, and TNF-a and is associated with a high mortality (48,49). These three cytokines are related to IL-36g either as downstream or upstream effectors (33,46,50), suggesting that IL-36g could be a potential contributor to the development of the hyperinflammatory endotype of ARDS. IL-36g neutralization in Thbs1 2/2 mice reduced levels of proinflammatory cytokines and chemokines, improved pathogen clearance, and reduced neutrophilic activity, evaluated as lung MPO and BALF free NE activity. However, IL-36g neutralization did not significantly reduce neutrophil recruitment to the airspaces, nor the increased production of IL-17A in the lungs. IL-17A is a master regulator of neutrophil chemotaxis (51). In a murine bacterial pneumonia model, lung IL-17 contributed to neutrophil recruitment through the induction of downstream chemokines different from CXCL1 and CXCL2, such as CXCL5 (52). Therefore, the robust production of IL-17A during P. aeruginosa infection in Thbs1 2/2 mice treated with anti-IL-36g may explain why IL-36g neutralization did not reduce the increased neutrophil numbers found in the airspaces and also suggest the existence of an alternative inflammatory pathway involved in neutrophil recruitment that should be further studied.
Collectively, these data support the idea that IL-36g neutralization could be exploited as an adjunct therapy against dysregulated inflammation observed in acute and chronic PA infections. As targeting the IL-36 signaling pathway is a viable strategy to block excessive inflammation of pustular psoriasis and other autoinflammatory disorders (53), a neutralizing antibody against IL-36R (ANB019) is currently in phase 2 clinical trial (ClinicalTrials.gov registration no. NCT03633396). One possible therapeutic application is in a subset of patients with runaway inflammation as a sequela of P. aeruginosa infection-induced tissue injury or conditions with protease/antiprotease imbalance such as in cystic fibrosis where excessive inflammation is a key feature. A better understanding of how extracellular processing of IL-36g is regulated by the host could provide a working framework in the design of new therapeutic strategies targeting pathogenic inflammation in the lungs.

MATERIALS AND METHODS
Mice. C57BL/6J (WT, stock no. 000664) mice and B6.129S2-Thbs1tm1Hyn/J (Thbs1 2/2 , stock no. 006141) mice were originally obtained from Jackson Laboratories (Bar Harbor, ME) and maintained in the animal facility of University of Pittsburgh as previously described (27,28). Thbs1 2/2 mice were further backcrossed an additional 5 generations before experiments. Thbs1 2/2 and WT mice were cohoused in the same vivarium and fed the same chow for at least 4 weeks prior to in vivo experiments as previously described (27). All experimental protocols were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Pittsburgh.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.