C-terminal truncation of IFN-γ inhibits proinflammatory macrophage responses and is deficient in autoimmune disease

Controlled macrophage differentiation and activation in the initiation and resolution of inflammation is crucial for averting progression to chronic inflammatory and autoimmune diseases. Here we show a negative feedback mechanism for proinflammatory IFN-γ activation of macrophages driven by macrophage-associated matrix metalloproteinase 12 (MMP12). Through C-terminal truncation of IFN-γ at 135Glu↓Leu136 the IFN-γ receptor-binding site was efficiently removed thereby reducing JAK-STAT1 signaling and IFN-γ activation of proinflammatory macrophages. In acute peritonitis this signature was absent in Mmp12–/– mice and recapitulated in Mmp12+/+ mice treated with a MMP12-specific inhibitor. Similarly, loss-of-MMP12 increases IFN-γ–dependent proinflammatory markers and iNOS+/MHC class II+ macrophage accumulation with worse lymphadenopathy, arthritic synovitis and lupus glomerulonephritis. In active human systemic lupus erythematosus, MMP12 levels were lower and IFN-γ higher compared to treated patients or healthy individuals. Hence, macrophage proteolytic truncation of IFN-γ attenuates classical activation of macrophages as a prelude for resolving inflammation. IFN-γ is central in inflammatory pathogenesis, response to infection and autoimmune diseases. Here the authors show that MMP12 expression is reduced in patients with SLE and that MMP12 post-translationally truncates IFN-y, inhibiting its function and affecting pathogenesis of mouse models of peritonitis, SLE and rheumatoid arthritis.

MMP12 cleaves substrates important for macrophage migration 21,22 , for example, elastin, fibronectin, laminin, entactin, type I and IV collagens, and proteoglycan core proteins [22][23][24] . Mmp12 -/mice are relatively healthy and, consistent with the extracellular matrix substrates of MMP12, have a defect in elastinolytic activity and a reduction in macrophage invasion to penetrate basement membranes, and a decrease in IL-13 22,25 . MMP12 also has antimicrobial activity 26 and a function in antiviral immunity by cleavage of IFN-α, but not IFN-β 6 , and regulates neutrophil influx via chemokine processing 23,27 .
Conflicting reports suggest IFN-γ has dichotomous timedependent activity and actions in autoimmunity that is not well understood 4,16 . We hypothesized that MMP12 may contribute to the temporal regulation of IFN-γ activity by proteolytic processing. Here we show an inverse correlation between mRNA levels of IFN-γ and macrophage MMP12 in SLE patients and upon treatment. We characterize inactivation of IFN-γ by C-terminal proteolytic processing that contributes to attenuation of proinflammatory macrophage activation in acute inflammation. This negative feedback mechanism is driven by MMP12 secretion by proinflammatory IFN-γ-activated macrophages, with elevated MMP12 expression in IL-4-activated immunosuppressive macrophages reinforcing inflammation resolution. Contrary to the detrimental functions classically ascribed to MMPs in inflammation, low-MMP12 levels may be a risk factor underlying excessive proinflammatory IFN-γ macrophage activation in disease.
Associated with increasing systemic lupus erythematosus disease activity index (SLEDAI) and clinical deterioration of SLE patients over time was increased IFN-γ associated with significantly decreased MMP12 mRNA levels (two-tailed paired Student's t-test: p = 6 × 10 −3 ) (GSE11909, n = 114) 28 (Fig. 1e, Supplementary Fig. 3). Conversely, in patients showing clinical improvement with decreasing SLEDAI over time, MMP12 mRNA was markedly increased up to~16-fold (two-tailed paired Student's t-test: p = 3 × 10 −2 ) (Fig. 1e, Supplementary Fig. 3). In sum, our reanalysis of human clinical sample transcript data revealed that low levels of MMP12 were associated with autoimmune disease and elevated IFN-γ signature genes. Thus, MMP12 may exert clinically significant protective roles that are deficient in the development of SLE. The return of higher MMP12 levels upon treatment and resolution of clinical symptoms are consistent with this hypothesis.
C-terminal truncation of IFN-γ. We hypothesized that MMP12 cleaves IFN-γ and attenuates classical activation of macrophages in vivo. Enzyme kinetic analyses revealed that human and mouse IFN-γ were efficiently cleaved proximal to the C-terminus by human and mouse MMP12, respectively ( Fig. 2a; Supplementary  Fig. 4a). IFN-γ cleavage was prevented by the MMP12-specific inhibitor, Rxp470.1 6,29,30 . The first cleavage product of human IFN-γ was apparent within 15 min of co-incubation, consistent with a k cat /K M (1) of 301 M −1 s −1 (Fig. 2b). A second MMP12 cleavage rapidly followed (k cat /K M = 75 M −1 s −1 (1)) and both products displayed an intact N-terminus (Met23 of the recombinant protein) as shown by Edman microsequencing (Fig. 2a). High-mass accuracy top down mass spectrometry identified the two cleavage sites in human IFN-γ at 157Met↓Leu158 and 135Glu↓Leu136 (Fig. 2b, Supplementary Figs. 4b and 5a-c). The 135Glu↓Leu136 scissile bond is conserved in human, monkey, mouse, and rat (N/HE↓LIR/QV) suggestive of a conserved function ( Supplementary Fig. 4b). In contrast, MMP12 did not cleave the immunosuppressive macrophage alternative activator, IL-4 ( Supplementary Fig. 6).  MMP12, C3, C4, C5, CD74, IFNA2, IFNB1,  IFNG, IFNGR1, IFNGR2, ITGAM, ITGAX S100A8, S100A9, STAT1, and NOS2 mRNAs in SLE patients (n = 102) or healthy control subjects (n = 12) in the GSE11909 transcript dataset 28 . Vertical bars are the mean values in all bean plots. c Comparison of these mRNA levels in PBMCs of drug-treated SLE patients (n = 40) or healthy control subjects (n = 32) in the GSE37356 mRNA dataset 15 . Statistical significance in b and d was determined by calculating the q-values in a and in c. e Longitudinal comparison (31 time points) in 20 individual subjects of the fold changes of MMP12 and IFNG PBMC mRNA levels upon clinical deterioration (increasing SLEDAI in individual patients over time, n = 7 time points), or upon clinical improvement (unchanged or decreasing SLEDAI in individual patients over time, n = 24 time points). Patient information (ethnicity, gender, age, SLEDAI and drug treatment or not) from GSE11909 is in Supplementary Tables S4-S5 28 . Statistical significance was determined by a two-tailed paired Student's t-test: p = 3 × 10 −3 IFNGRs are comprising two ligand binding IFNGR1 chains and two signal-transducing IFNGR2 chains 3,31 . We related the cleavage sites in the IFN-γ dimer to the IFNGR chain structures using in silico structural analysis (Fig. 2c). The IFN-γ nonstructured C-terminal strand was modeled from Ala146 since it is absent from the crystal structure 32,33 . Our analysis predicted that the strand remaining after cleavage at 157Met↓Leu158 would retain binding to both homodimers of IFNGR1 and IFNGR2, consistent with activation 34,35 (Fig. 2d). However, both cleavage products were consistently identified together at all time points by mass spectrometry indicating rapid processive cleavage leading to the final shorter truncated product ending at Glu135 (Fig. 2b   Supplementary Fig. 5a-c). As the loss-of-IFNGR2 binding is predicted by truncation at 135Glu↓Leu136, this suggests the net effect of MMP12 is loss-of-IFN-γ signaling.
Reduced IFN-γ activation of macrophages by MMP12 in vivo. We next examined IFN-γ levels in acute thioglycollate-induced peritonitis from 0 to 4 days in Mmp12 +/+ B10.RIII (n = 20) and Mmp12 -/-B10.RIII (n = 20) mice (n = 4, for each genotype for each day). Consistent with the initiation of IFN-γ clearance by MMP12 in the Mmp12 +/+ mice, ELISA analyses revealed that IFN-γ protein was~3-fold higher in the peritonitis exudate at days 1 to 4 in the Mmp12 -/mice (Fig. 3a). This difference was also found in peritoneal macrophage lysates by western blotting (Fig. 3b) and was associated with increasing amounts of MMP12 protein over time in the Mmp12 +/+ B10.RIII mice (Fig. 3b). The IFN-γ-activated macrophage response markers 5 , iNOS ( Fig. 3b) with attendant ROS production (Fig. 3c), were also up at day 2 and then later declined at days 3 and 4 in the Mmp12 +/+ B10.RIII peritoneal macrophages. However, these levels were always greater in the Mmp12 -/-B10.RIII primary macrophages, with iNOS and ROS continuing to significantly increase over time; whereas, iNOS protein was undetectable in the Mmp12 +/+ B10. RIII macrophages at days 3 and 4. As no one marker is indicative of either proinflammatory classically activated or immunosuppressive alternatively activated macrophages, we supported the iNOS findings in peritonitis by a spectrum of other markers associated with IFN-γ activation. Thus, STAT1, MHCII, S100A8, and S100A9 were also at higher levels in primary Mmp12 -/-B10.RIII macrophages compared with Mmp12 +/+ B10.RIII cells, particularly at the later time points (Fig. 3d). In contrast, Mmp12 +/+ B10.RIII mice displayed higher levels of CD36 on day 4, which is a marker of IL-4 alternative activation. Thus, Mmp12 -/-B10.RIII mice displayed elevated IFN-γ protein and prolonged classical macrophage activation in acute inflammation compared with Mmp12 +/+ B10. RIII mice. In other words, the absence of MMP12 shifted peritoneal macrophage populations to one enriched with IFN-γclassically activated proinflammatory macrophages.
IFN-γ cleavage reduces STAT1 and iNOS during phagocytosis. IFN-γ activates JAK-STAT1 signaling with phosphorylation of STAT1 at Tyr701 leading to IFN-γ response gene transcription occurring within 15-30 min and a characteristic upregulation of total STAT1 protein expression after 18-24 h that is highly correlated with IFN-γ activation of macrophages 3,36 . To examine the effect of MMP12 cleavage on IFN-γ signaling, we treated mouse RAW264.7 cells lines for 15 min or 24 h with full-length mouse IFN-γ, or with mouse MMP12-cleaved mouse IFN-γ terminating at Glu133 that was mass spectrometry confirmed ( Fig. 4a, b). We also treated PMA-matured THP-1 human monocytes with human IFN-γ or with human MMP12-cleaved human IFN-γ terminating at Glu135 (mass spectrometry confirmed) (Fig. 4c,  d). In both macrophage cell lines, IFN-γ led to the characteristic phosphorylation of STAT1 at Tyr701 after 15 min with subsequent upregulation of total STAT1 protein at 24 h (Supplementary Fig. 7a, b). Tyr701 phosphorylation was associated with an increase in iNOS protein (Fig. 4b, d). In contrast, incubation with MMP12-truncated IFN-γ neither induced STAT1 phosphorylation nor increased total STAT1 or iNOS proteins in either cell line.
An important response to macrophage classical activation by IFN-γ is elevated phagocytosis with associated induction of microbicidal activity wrought by ROS 3 that is generated by elevated iNOS 5 . In a fluorescent bead phagocytosis assay with differentiated THP-1 cells, PBS-treated and IL-4-activated macrophages showed low levels of phagocytosis (Fig. 4e, f). Phagocytosis was increased 4.9-fold in IFN-γ-activated THP-1 cells, but when IFN-γ was cleaved by MMP12, the cells retained the lower phagocytic activity of non-activated macrophages. These functional data are consistent with the elevation in iNOS protein levels in both IFN-γ-activated RAW264.7 and THP-1 cells compared to PBS-treated or IL-4-alternately activated macrophages, as well as a corresponding increase in cellular ROS (Fig. 4b, d). In contrast, MMP12 cleavage of IFN-γ reduced the increase in the IFN-γ activation markers iNOS and ROS.
Mmp12 loss is associated with elevated IFN-γ markers in arthritis. We found that genetic deletion of MMP12 increased the severity of inflammatory arthritis in mice. Twenty-eight days after induction of collagen-induced arthritis, male Mmp12 -/-B10.RIII mice (n = 20) exhibited more severe arthritis ( Fig. 5a, b), with increased ankle width and histopathological scores compared with male Mmp12 +/+ B10.RIII mice (n = 18). The Mmp12 -/-B10.RIII mice also had higher levels of IFN-γ and the IFN-γactivated macrophage markers iNOS and MHCII, with concomitant near absent CD36 staining (Fig. 5c, d). Neutrophils and neutrophil extracellular traps (NETs) in the synovial space 23 further indicated an acute IFN-γ-driven proinflammatory response 38 . We confirmed these findings in female mice on another background. The autoimmune MRL/lpr mouse strain 39 spontaneously develops systemic autoimmunity with severe lymphadenopathy, arthritis and glomerulonephritis. In our substrain, we found it was necessary to initiate autoimmune arthritis earlier with Complete Freund's adjuvant (CFA), otherwise gross lymphadenopathy necessitated euthanasia too soon to derive meaningful insights into the development of severe arthritis. As CFA stimulates the hyperproduction of IFN-γ, expression of MHC class II genes, and Th1/classical activation of macrophages 39,40 we reasoned this strain was suited to assess the effects of MMP12 on IFN-γ-driven disease and macrophages in vivo.
In female Mmp12 -/-MRL/lpr mice, we saw similar associations as found in the B10.RIII mice. Hind ankle joints at baseline were normal with no evidence of disease as quantified by a 4-point histopathological grading scheme 41 , including no difference in paw edema in both the Mmp12 +/+ (n = 30) and Mmp12 -/-(n = 28) female MRL/lpr mice ( Supplementary Fig. 8b). With loss-of-MMP12 there was an increased incidence of arthritis as compared to Mmp12 +/+ MRL/lpr mice ( Supplementary Fig. 8a). In the acute disease phase immediately following CFA induction, the Mmp12 -/-MRL/lpr mice displayed a faster onset of arthritis that was detectable at day 6 by significant and more severe ankle swelling than Mmp12 +/+ MRL/lpr mice, that continued to exacerbate to day 25 (Fig. 5e). In contrast, in Mmp12 +/+ MRL/lpr mice, ankle edema started to decrease at day 19 and continued to fall to day 25-at which time the Mmp12 -/-MRL/lpr mice showed larger pannus formation, more synovial hyperplasia and subsynovial inflammation ( Supplementary Fig. 8c) with a histopathological score of 3.8 in comparison to 1.1 for Mmp12 +/+ MRL/lpr mice (Fig. 5f). In quantifying differences in IFN-γ-activated macrophage proteins at day 25 we found the Mmp12 -/-MRL/lpr ankles had elevated IFN-γ, as well as elevated iNOS and MHCII classical activation marker proteins, reflecting sustained IFN-γ-stimulated macrophage responses in vivo (Fig. 5g, h). This indicated that without MMP12, MRL/lpr mice experienced a loss-of-control of macrophage activation and inflammation that led to a sustained disease flare.
Elevated IFN-γ activation in Mmp12 -/mouse models of lupus. Mmp12 Peritoneal macrophages . Data were normalized to day 1 Mmp12 +/+ B10.RIII macrophages and expressed as fold differences. Error bars, s.d. Statistical significance was determined by a two-tailed unpaired Student's t-test: ***p < 0.005. d 10% SDS-PAGE western blot analysis of markers characteristic for macrophage activation by IFN-γ (MHCII, S100A8, and S100A9) and STAT1, or by IL-4 (CD36) and STAT6 in Mmp12 +/+ and Mmp12 -/-B10.RIII mouse macrophages harvested daily (n = 20 for each genotype, n = 4 for each time point, N = 2). Tubulin, loading control. Molecular weight marker positions in all blots are as shown chronic lymphoid-driven human diseases, such as SLE 36,39,42 and in mouse models of SLE, including the MRL/lpr strain 39,40 . The signature lymphoproliferative response in female MRL/lpr mice, as evidenced by enlarged lymph nodes, was seen by day 90 in female Mmp12 -/-(n = 18) and Mmp12 +/+ (n = 38) mice ( Fig. 6a, Supplementary Fig. 9a, b, N = 2) and later for males (n = 13 each for Mmp12 -/and Mmp12 +/+ male mice, Supplementary Fig. 9a). Consistent with the MMP12-dependent differences in lymph node size apparent by day 90 in the female mice, CFA synchronization further expanded the differential increase in lymphadenopathy manifest in the Mmp12 -/-MRL/lpr females (n = 18, N = 2) over that seen for the female Mmp12 +/+ MRL/lpr  (Fig. 6a, Supplementary Fig. 9b), supporting a role for MMP12 in dampening acute inflammatory responses. As in human autoimmune disease, where there is significantly increased prevalence in females versus males 43 , female mice in the absence of CFA had a shortened survival time than males (112 days versus 131 days) before lymphadenopathy led to humane endpoint euthanasia (Fig. 6b, Supplementary Fig. 9c-e, N = 2). Protein levels of IFN-γ and the IFN-γ-activated macrophage markers iNOS and MHCII were significantly increased in the superficial cervical lymph nodes of the female Mmp12 -/-MRL/lpr mice compared with female Mmp12 +/+ MRL/lpr mice (day 98) (Fig. 6c, d). As immunohistochemistry is only semi-quantitative we confirmed the elevation of these IFN-γ classical activation macrophage proteins, plus STAT1, S100A8, and S100A9, in lymph node extracts of the knockout versus the wild-type Mmp12 MRL/ lpr mice (Fig. 6e). Of note, we detected lower molecular weight forms of IFN-γ in Mmp12 +/+ but not Mmp12 -/mice (Fig. 6e), suggestive of in vivo processing of IFN-γ. Thus, the lack of MMP12 was associated with a shift in macrophage populations to one retaining ongoing IFN-γ-activation in vivo.
C-terminal cleavage of IFN-γ is reduced in lupus nephritis. We next analyzed MMP12 and IFN-γ protein levels and the IFN-γ cleavage status in patients with lupus nephritis (Supplementary Table 1). To do so we raised and affinity-purified three antibodies to human IFN-γ to determine its activation status in vivo (Fig. 7a). To profile total (active plus inactive) IFN-γ protein, we raised an antibody against the N-terminal IFN-γ peptide, 25DPYVKEAENLKKYFNAG41. To identify full-length IFN-γ protein, we raised antibodies against the C-terminal sequence 158LFRG161 (C-terminal-1 antibody) distal to the first cleavage site and therefore only present in non-cleaved IFN-γ. By recognizing a sequence (QVMA) between the 1st and 2nd MMP12 cleavage sites, the C-terminal-2 antibody discriminated between the two C-terminal MMP12-truncated forms. So, whereas positive staining for the N-terminal antibody identified all 3 forms of IFN-γ, including the inactive MMP12-truncated IFN-γ 23-135, the C-terminal-2 antibody confirmed the presence of both the active IFN-γ proteoforms past position 135. As the human and murine IFN-γ sequences diverge significantly in the C-terminus ( Supplementary Fig. 4b), the anti-human IFN-γ antibodies did not cross react with mouse IFN-γ.
After a time-course incubation of IFN-γ with MMP12 for up to 1080 min the presence of a C-terminal cleaved form of IFN-γ was apparent using the N-terminal antibody, even as soon as 15 min (Fig. 7a), confirming the kinetic data of efficient cleavage. The Cterminal-2 antibody will recognize both full-length and IFN-γ cleaved at 157Met↓Leu158. Decreasing immunoreactivity for both C-terminal epitopes over time in vitro confirmed cleavage removal of the short C-terminal peptides (Fig. 7a). Although cleavage at 135Glu↓Leu136 can occur first, the loss-of-C-terminal-1 immunoreactivity with a weak C-terminal-2 band still present at 1080 min indicated that 157Met↓Leu158 cleavage can also precede 135Glu↓Leu136 cleavage. Therefore, the prominent lower immunoreactive band detected by the strong N-terminal antibody is likely the 135Glu↓Leu136 cleaved inactive IFN-γ proteoform.
Finally, to directly link MMP12 catalytic activity with the controlled normal succession of IFN-γ-classically activated macrophage populations to alternately activated macrophages in vivo, we injected Mmp12 +/+ B10.RIII mice with 5 mg/kg MMP12-specific inhibitor (Rxp470.1) (n = 4) or vehicle (n = 4) daily for 4 days during the development of peritonitis. Although MMP12 is predominantly a macrophage-associated protease 21,22 , other cells also express the enzyme 6 , which systemic administration of Rxp470.1 will block. In primary macrophages of the vehicle controls harvested at day 4 we found that the cell population was enriched for macrophages with high CD36 expression (indicative of the alternative macrophage activation pathway), an absence of iNOS and MHCII proteins, and low levels of other IFN-γ-classically activated macrophage signature proteins (Fig. 8f). This represents the normal temporal evolution of macrophage phenotypes in acute inflammation. In contrast, treatment of the Mmp12 +/+ B10.RIII mice (n = 4) for 4 days with Rxp470.1 shifted the day 4 peritoneal macrophage population to one retaining the early proinflammatory IFN-γ-classically activated macrophage characteristics of Mmp12 +/+ B10.RIII at day 2 (Fig. 3b, d) or Mmp12 -/-B10.RIII mice at day 4 (Figs. 8f and 3b, d). Thus, IFN-γ, iNOS, MHCII, S100A8, S100A9, and STAT1 proteins were elevated in peritoneal macrophages harvested from the Rxp470.1-treated Mmp12 +/+ mice compared with vehicle-treated Mmp12 +/+ mice at day 4 and the profile now matched that of the Mmp12 -/mice at day 4. Notably, this lossof-controlled evolution in macrophage populations over time was also a feature of Mmp12 -/-B10.RIII arthritis at 28 days (Fig. 5) and lymphadenopathy in MRL/lpr mice at 96 days (Fig. 6e). Thus, injection of a specific MMP12 inhibitor in vivo in this loss-offunction experiment directly linked MMP12 catalytic activity with the normal controlled shift in proinflammatory to immunosuppressive macrophage cell phenotype ratios occurring in vivo.

Discussion
It is important to recognize all levels of control of key cytokines contributing to the transition from proinflammatory to immunosuppressive macrophage dominated inflammation, and therefore disease risk, susceptibility and resolution. IFN-γ is one of the main effectors in autoimmune diseases, such as SLE, rheumatoid arthritis, and multiple sclerosis 19,46 . However, the role of posttranslational modification of IFN-γ in these diseases has not been previously recognized. We demonstrated that MMP12, primarily a monocyte-macrophage protease 21,22 , downregulates IFN-γ activity by precise C-terminal cleavage at an evolutionary conserved site that prevents phosphorylation of STAT1-Tyr701, and thus IFN-γ signaling and classical activation of macrophages. Initial MMP12 cleavage then sets the stage for IFN-γ clearance in vivo that reduces total IFN-γ levels over time. By forming a post-translational negative feedback mechanism, we found that the normal succession from IFN-γ-classically activated proinflammatory to alternately activated macrophages was dampened when MMP12 was genetically deficient or therapeutically inhibited, especially in acute inflammation, but also in chronic inflammatory autoimmune models experiencing disease flares. Thus, there was a clear shift in inflammatory cell infiltrates to populations enriched for IFN-γ-activated macrophages with attendant tissue pathology in Mmp12 knockout mice in three different models of inflammation, in both male and female mice, and on two different genetic backgrounds.
Numerous pathways and cell types are important in the pathogenesis of chronic inflammation and autoimmune diseases.
Multifactorial control of normal macrophage polarization leads to many possible break points that can fail, increasing the risk of chronic inflammation and autoimmunity. Such complexity renders diagnosis and treatment challenging. We were able to translate our findings identifying proteolytic control of IFN-γ activity and IFN-γ activation of macrophages to human autoimmune disease. By analyzing previous data, we found that SLE diagnosed by multiple clinical parameters, including elevated anti-double stranded DNA antibody levels, was associated with lower PBMC MMP12 mRNA levels than in healthy subjects. Further, with clinical improvement upon treatment during the chronic inflammatory phase of the disease, MMP12 mRNA expression increased to near healthy control levels. In a different study in which clinical responses after treatment of 40 SLE patients were examined, we found a strong association in the reduction of IFN-γ response gene expression that coincided with elevated expression of MMP12 mRNA and IL-4 alternatively activated macrophage markers. We found that MMP12 normalization occurs after instituting treatment and precedes clinical improvement, whereas patients whose MMP12 levels are not affected appeared not to improve. Although this clinical association does not prove a mechanistic link for MMP12 in protection from autoimmunity in human disease, it supports the findings from our animal model studies.
The temporal integration of opposing cytokine signals important in differentiation fate decisions for IFN-γ classical activation versus IL-4 activation and other alternate macrophage activation pathways, and the spectrum of cell populations occurring in between, is a multi-faceted, highly nuanced immune mechanism that underpins the development of chronic inflammation and autoimmunity and leads to acute clinical relapses. As MMP12 expression is higher in IL-4 activated macrophages than in IFN-γ-activated cells, MMP12 expression by alternatively activated macrophages may amplify in trans the reduction in IFN-γ activity by IFN-γ-activated macrophage MMP12 to reinforce the decrease in proinflammatory cells. Conversely, low-MMP12 activity favors chronic disease progression and so may be a risk factor for SLE, and potentially other autoimmune diseases. Our findings also further emphasize that using genomic or transcriptomic data alone risks overlooking the key importance of post-translational modifications of immune regulatory and autoimmune signature cytokines. The present data establishing the proteolytic regulation of IFN-γ activity in acute inflammation, including disease flares of chronic disease, suggest a more generally applicable mechanism not limited to SLE or lupus nephritis, but also potentially at play in other IFN-γ-linked phenotypes. Having identified that MMP12 expression provides a benefit in mouse models of IFN-γ-driven inflammatory diseases, it will be important to prospectively evaluate MMP12 activity in a larger cohort of patients with diseases such as SLE. If activity is significantly lower in a large patient cohort with active disease, it will provide crucial information that potentially may point to new therapeutic modalities that restore normal cytokine cellular control to augment the production of MMP12.

Methods
MMP12 cleavage assays of IFN-γ. Recombinant human and mouse MMP12, expressed and purified as described 49 , were incubated with 10 or 100 ng human or mouse MMP12, 37°C in assay buffer (100 mM Tris-HCl, 50 mM NaCl, 5 mM CaCl 2 , pH 8.0) at enzyme/substrate ratios from 1:10 to 1:100. IFN-γ cleavage was determined by silver staining of 15% Tris-Tricine gels and western blots. To identify cleavage sites, cleaved peptides were analyzed by the Edman sequencing, liquid chromatography tandem-mass spectrometry (LC-MS/MS), and MALDI-TOF MS on a Applied Biosystems Voyager DE STR Mass Spectrometer as described 50 . The specific MMP12 inhibitor RXP470.1 6,30 was a control in the cleavage assays. The k cat /K M values were quantified according to (1) from IFN-γ fragment generation over time as described 50 : Top down mass spectrometry. IFN-γ and MMP12 cleavage fragments were analyzed by top down liquid chromatography tandem-mass spectrometry (LC-MS/MS). After cleavage (15 min or 240 min), samples were first purified by solid phase extraction on Empore Octyl C8 disks (3M Empore) packed as a C18 StageTips. Tips were activated with 50% acetonitrile, equilibrated with 0.1% formic acid, samples were acidified, loaded and then washed with 2 column volumes of 0.1% formic acid, and eluted in 50% acetonitrile, 0.1% formic acid. Eluates were directly infused into an Impact II high resolution, high-mass accuracy quadrupole time-of-flight (QTOF) system using a Captive-Spray ion source (Bruker Daltonics) at a flow rate of 3 µL/min, ionized with 4.5 kV, 0.3 bar nebulizer gas pressure, heated to 180°C and analyzed in positive ion mode. MS1 spectra were acquired in the mass range of 110-5000 m/z at a scan rate of 1 Hz. Following acquisition, a composite MS1 spectrum was generated by summing all MS1 spectra during acquisition time using Compass DataAnalysis v4.3 (Bruker Daltonics). Spectra were deconvoluted by mMass and used to determine the mass of peptide and protein ions. Default mMass baseline correction and smoothing parameters were used.
To analyze collagen-induced arthritis, we backcrossed Mmp12 -/mice on the 129Sv/Ev background for seven generations onto the B10.RIII-H2 r H2-T18 b background (The Jackson Laboratory, stock #000457 23 ). Genotyping was performed by the HotShot DNA isolation followed by polymerase chain reaction with Mmp12 primers (forward) and (reverse) and a primer specific for the neomycin insert (reverse). The PCR products were 1100 bp for Mmp12 +/+ mice and 550 bp for Mmp12 -/mice. The UBC Animal Care Committee approved all mouse breeding and animal experimental procedures and, established and supervised humane end points.  Thioglycollate-induced sterile peritonitis. At day 90, male and female mice were injected intraperitoneal with 0.5 mL 4% Brewers thioglycollate broth (Sigma-Aldrich, St. Louis, MO) to establish acute sterile peritonitis. Mice were excluded from further studies if obvious adhesions were observed between the peritoneum and skin or if marked increases were observed in the thickness of the peritoneum. Mice were killed at day 0, 1, 2, 3, or 4 and 5 mL of phosphate buffered saline (PBS) was administered via intraperitoneal injection to collect infiltrating cells and peritoneal exudate. Cells were plated for 4-18 h in RPMI 1640 medium, 10% cosmic calf serum and non-essential amino acids. For in vivo MMP12 inhibition, we administered 5 mg/kg Rxp470.1 intraperitoneal once per day for 4 days in Mmp12 +/+ B10.RIII (n = 4) and Mmp12 -/-B10.RIII (n = 4) mice and collected the peritonitis fluid and cells by PBS lavage on day 4.
Cells. Human THP-1 monocytic cells (ATCC, Manassas, Virginia, USA), mouse RAW264.7 macrophages (ATCC, Manassas, Virginia, USA) and primary mouse peritoneal macrophages were cultured in 10% fetal bovine serum (v/v) in RPMI. All cell lines were tested and found to be mycoplasma negative.
Phagocytosis assay. THP-1 cells (5 × 10 6 cells in 10 mL) were stimulated with 200 ng/mL PMA for 24 h to induce cell attachment and differentiation and then the medium was changed. A 2.69% Fluoresbrite® 2-μm microparticle suspension (12 μL) (Polysciences, Inc., Warrington, PA) was incubated in 175 μL normal human serum (Complement Technology Inc., Tyler, TX) for 30 min at 37°C before addition to the PMA-activated cells for 1 h. Phagocytosis was stopped by adding 2 mL of ice-cold HBSS. The bead solution was removed, and the cells were washed twice with HBSS, then harvested by trypsinization and washed. The phagocytosis index was determined in each experiment (n = 3 repeats per experiment, and the experiment itself performed twice, N = 2) by counting the number of beads per cell in 20 separate fields containing 5-30 cells/image for each repeat. Statistical analysis was determined by a two-tailed unpaired Student's t-test.
Raising and affinity purification of IFN-γ epitope antibodies. Three IFN-γ peptides (GenScript, Piscataway, NJ): DPYVKEAENLKKYFNAG-GC; C-GGLFRG, and CRGG-QVMA (Arg was added for solubility), all synthesized with either a G or GG added as a flexible spacer to a cysteine to enable conjugation to succinimidyl 4-(N-maleimidomethyl) cyclohexane-1-carboxylate/Keyhole Limpet Hemocyanin (Sigma-Aldrich, St-Louis, MO). The conjugates were emulsified in CFA and injected in New Zealand white rabbits. Antibody expression was boosted with peptide in incomplete Freund's adjuvant at 3 weeks. Serum antibody titers to the respective peptides were quantified by ELISA. SulfoLink Immobilization (Pierce Biotechnology, Rockford, IL) was used to conjugate the peptides to affinity-purify antibodies from rabbit sera. Peptides (5 mg) were dissolved in 2 mL of coupling buffer (50 mM Tris, 5 mM EDTA, pH 8.5), reduced with 25 mM TCEP, and incubated for 30 min. The sulfhydryl-containing peptides were coupled to 3 mL SulfoLink columns by mixing with the SulfoLink resin for 15 min and incubated for 2 h. Non-specific binding sites were blocked with 50 mM cysteine for 1 h. The columns were then equilibrated with 50 mM Tris, 5 mM EDTA, pH 8.5 before affinity purification of anti-peptide antibodies from rabbit sera. After column washes, high-affinity antibodies were eluted using 0.1 M glycine, pH 2.5. Antibody specificity was verified by ELISA and western blotting to IFN-γ and MMP12cleaved IFN-γ. Western blotting of human kidney extracts was used to assess specificity and non-specific binding before translation to ELISA assays.
ELISA. Plates were coated with 5 μg/mL of each of the 3 IFN-γ peptides in Vollers buffer (15 mM Na 2 CO 3 , 35 mM NaHCO 3 , pH 9.6) overnight and then blocked with 1% BSA in PBS for 1 h. The wells were rinsed three times with PBS-0.05% Tween 20 and 100 μL of a serial dilution of antibodies (1:10 2 -1:10 9 ) were added and incubated at room temperature for 2 h. The wells were rinsed with PBS-Tween 20 (0.05%) before 100 μL of 1:1000 goat anti-rabbit HRP antibody (Bio-Rad Laboratories, Hercules, CA) was added and incubated for 1 h. Wells were aspirated and incubated in the dark with SIGMAFAST™ OPD HRP substrate (Sigma-Aldrich, St-Louis, MO) until visible coloration was seen. At this time 50 μL of 1 M acetic acid was added to stop the reaction and the absorbance was measured at 450 nm on a POLARstar optima microplate reader (BMG Labtech, Durham, NC).
Collagen-induced arthritis in B10.RIII mice. Collagen-induced arthritis was induced by injection with type II collagen (4 mg/mL) at two inguinal sites in 49day-old-male Mmp12 -/-(n = 20) and Mmp12 +/+ mice (n = 18) all on the B10.RIII background. Day 0 was defined as 21 days after injection and was typically the date of onset of arthritis. At 18 and 28 days later, the mice were killed and the hind ankles were dissected and prepared for immunostaining and histological staining by hematoxylin and eosin (H&E) and toluidine blue to visualize cartilage proteoglycan 23 . NETs were previously characterized in collagen-induced arthritis by a variety of stains, including hematoxylin and eosin in combination 23 or alone 59, 60 , the histological appearance and distribution of which are consistent with the NETs identified here by hematoxylin and eosin.
Arthritis in MRL/lpr mice. Spontaneous arthritis and lupus nephritis are a characteristic of the MRL/lpr strain. To ensure reasonably severe and consistent disease stage between animals before excessive lymphadenopathy necessitated animal euthanasia it was necessary to synchronize the initiation of systemic autoimmunity by CFA earlier than spontaneous disease was manifest. We injected 0.05 mL of a 1:1 emulsion of CFA (Sigma) in water intradermal at two inguinal sites in 90-day-old mice 41 . The width of both hind ankles was measured on three consecutive days before injection and every 2-3 days for 25 days following injection using electronic digital calipers (Marathon). Baseline ankle width was calculated as the average of the pre-injection measurements. Arthritis was deemed present with a 10% increase in average ankle width. Spleen weights were determined of female Mmp12 -/-MRL/lpr (n = 22) and Mmp12 +/+ MRL/lpr (n = 27) mice with or without CFA injection. Statistical analysis was performed with Prism 5 (GraphPad Software) using two-way analysis of variance with Bonferroni post-tests.
Immunostaining of biopsy sections. Mouse ankle, lymph node and kidney, and human kidney biopsies were dissected and formalin-fixed for 16 h. Hind ankles were decalcified in 10% formic acid for 72 h. Biopsies were then dehydrated through graded ethanol series and paraffin-embedded. Tissue sections (3-5 μm) were deparaffinized, rehydrated through an ethanol series ending in water and stained with hematoxylin and eosin, toluidine blue, trichrome, Jones, or Periodic acid-Schiff (PAS). For immunostaining, tissue sections were rehydrated, quenched to remove endogenous peroxidase, blocked in 1/10 normal goat serum for 1 h and then incubated with antibody in Tris-buffered saline. Primary antibodies and species appropriate secondary antibodies were used as performed by Wax-it Histology Services (Vancouver, BC, Canada).
Mouse glomerulonephritis activity and chronicity index scoring. The means of the activity and chronicity scores from Mmp12 +/+ MRL/lpr (n = 11) and Mmp12 -/- Fig. 7 IFN-γ epitope antibody staining of human lupus nephritis biopsies. a Left, amino acid sequences (underlined) of human IFN-γ peptides used to raise and affinity-purify rabbit anti-N-terminal, C-terminal-1, and C-terminal-2 IFN-γ epitope antibodies. Right, western blot analysis of human IFN-γ after timedependent cleavage to 1080 min by human MMP12. Note: disappearance of the C-terminal epitopes as MMP12 cleavage proceeds. Molecular weight marker positions in all blots are as shown. b Immunostaining of human kidney biopsies using anti-N-terminal, C-terminal-2, C-terminal-1, and MMP12 antibodies; and staining with hematoxylin and eosin (H&E), Trichrome, Jones, and PAS of kidney biopsies of healthy (n = 5), lupus nephritis at Grades III-(A) (n = 3) and IV-(A) (n = 2) as diagnosed in Supplementary Fig. 11. Scale bar, 100 μm. Original magnification, ×400. c Quantification of immunostaining intensities in healthy (n = 5) and lupus nephritis (n = 5) kidney biopsies (additional data are included in Supplementary Fig. 11) and expressed as means ± s. d. Statistical significance was determined by a two-tailed unpaired Student's t-test: ***p < 0.005   MRL/lpr (n =15) mice were calculated, for each biopsy, by analysis of 10 components divided into two scoring systems: Renal activity index (1-cellular proliferation, 2-leukocyte infiltration, 3-fibrinoid necrosis or karyorrhexis, 4-cellular crescents, 5-hyaline thrombi, wire loops and 6 -mononuclear cell infiltration), and Renal chronicity index (7-glomerular sclerosis, 8-fibrous crescents, 9-interstitial fibrosis, 10-tubular atrophy) 61,62 . Each individual component was scored 0 (for normal), 1, 2, or 3 (for severe abnormality, 3 being the worse). In calculating the activity index, fibrinoid necrosis and cellular crescents were weighed by a factor of 2. The renal activity index scale ranges from 0 to 24 points and the renal chronicity index scale from 0 to 12. For immunostaining analyses of mouse kidney glomeruli, and glomeruli plus the interstitium a minimum of 10 different fields, with each field having a minimum of 5 glomeruli were examined and quantified for each antibody, for each of these two separate analyses (N = 2). Three mice per group (Mmp12 +/+ (n = 3) and Mmp12 -/-MRL/lpr (n = 3)) were analyzed separately for the glomeruli and the glomeruli and interstitium staining quantifications.
Mouse whole-blood leukocyte and monocyte counts. Whole blood was collected from mice in EDTA-collection tubes and analyzed on a Siemens Advia 120 Hematology System (Oakville, ON, Canada) with the Perox method. The whole blood of each mouse (Mmp12 +/+ (n = 15) and Mmp12 -/-(n = 27) MRL/lpr) was analyzed in technical duplicates.
MMP12 mRNA analyses. Relative expression levels of MMP12 mRNA was measured in PMA-matured THP-1 cells treated with 20 ng/mL IFN-γ, 30 ng/mL IL-4, or PBS for 24 h (n = 3, N = 2 for each condition). After isolating RNA using the RNeasy Mini system (Qiagen), total RNA (1 μg) was amplified (Ambion MessageAmp II amplification kit, Thermo Scientific) and labeled by ULS aRNA fluorescent labeling (Kreatech Biotechnology, Amsterdam, The Netherlands). ULS-Cy3 labeled RNA was mixed 1:1 with ULS-Cy5 labeled control RNA (equimix of all amplified RNA samples) and hybridized to the CLIP-CHIP TM human protease and inhibitor targeted microarray 63 for 18 h, 42°C in duplicate with three biological replicates (i.e., six data points/gene). The microarray slides were scanned (428 ArrayScanner, MWG Biotech, High Point, NC, USA) and analyzed using ImaGene 6.1 (BioDiscovery, El Segundo, CA, USA). Data were normalized using the Bioconductor-based CARMAWeb software package 44 and statistically analyzed and visualized with the MulitExperiment viewer 4.8.1 from TIGR (TM4.org) using the significance analysis of microarrays algorithm as described 45,63 .
RNA expression relationships in human systemic lupus erythematosus. We searched three major databases (19/07/2017) GEO (26,622 human array or sequencing-based RNA datasets) 64 , ArrayExpress (70,672 datasets of 2,242,940 assays) 65 , and Gemma 66 (4118 datasets)) data series relating to human lupus transcript analyses and individually examined more than 239 datasets. Although there are 152 data series for human lupus in GEO and 130 in ArrayExpress, only five in total had longitudinal data, which were downloaded from Gemma. We compared RNA expression levels of MMP12 and IFNG genes in untreated SLE patients vs. healthy subjects in the GSE11909 dataset; and of untreated healthy reference subjects vs. treated SLE patients in the GSE37356 dataset. For each gene, microarray probes with the lowest p-value were selected and the multiple testing FDR-corrected q-values calculated by a linear modeling approach and taken directly from Gemma. All plots and analyses were done in R.
Human lupus nephritis analyses and grade scores. Five kidney biopsies from non-SLE patients with mild to moderate arteriosclerosis and healthy, non-lupus, non-inflammatory kidney disease, and kidney samples from SLE patients were classified independently by one blinded pathologist and one nephrologist: Class I, minimal mesangial lupus nephritis; Class II, mesangial proliferative lupus nephritis; Class III, focal lupus nephritis; Class IV, diffuse segmental (IV-S) or global (IV-G) lupus nephritis; Class V, membranous lupus nephritis; Class VI, advanced sclerosing lupus nephritis. Healthy (n = 5) and active (A) (n = 5) SLE Grade III-(A), IV-Global-(A), and IV-Segmental-(A) kidney biopsies were selected for immunochemical analysis.
Glomeruli were examined and quantified for each antibody from all human biopsies of the healthy kidneys (n = 43 glomeruli) and SLE patient kidneys (n = 46 glomeruli) and scored according to these classes: Class I, normal glomeruli by light microscopy; Class II, mesangial hypercellularity or mesangial matrix expansion by light microscopy with immune deposits; Class III, active or inactive focal, segmental or global endocapillary or extracapillary glomerulonephritis involving <50% of all glomeruli; Class IV, active or inactive focal, segmental or global endocapillary or extracapillary glomerulonephritis involving >50% of all glomeruli; Class V, global or segmental subepithelial immune deposits or their morphologic sequelae by light microscopy; and Class VI, >90% of glomeruli sclerosed.
Statistical analyses. A normal distribution of the data were shown in all cases and hence only parametric tests were required. Standard deviation and analysis of variance were used to analyze data variability. The Student's t-test determined statistical significance between two treatments. A p-value <0.05 was considered statistically significant.
Human ethics approvals. The UBC/Providence Health Care British Columbia Research Ethics Board approved the patient consenting procedures and research protocols for studies using human samples (approval #H15-00953), informed consent was obtained from all patients and the research was conducted in accordance with the Declaration of Helsinki.
Data availability. All data are available from the authors upon reasonable request.