Introduction
Osteoarthritis (OA) is a common and debilitating condition that similarly affects horses and people (1, 2). Because chronic synovial inflammation is a hallmark of OA and often the single driver of related degenerative changes (3–7), the use of anti-inflammatory drugs (steroidal and non-steroidal) has been a logical and long-accepted approach for the treatment of many arthritic conditions (8, 9). However, acute inflammation is not simply a clinical sign to alleviate. Acute inflammation is a critical event in promoting tissue repair and setting the stage for endogenous resolution of the inflammatory process and recovery of homeostasis (10). Importantly, anti-inflammatory and pro-resolving effects are not the same, and resolution is not merely the passive termination of the inflammatory process. Anti-inflammation is based on inhibiting key pro-inflammatory mediators, such as chemokine and cytokine production and leukocyte extravasation to the site of injury. Resolution is an active process driven primarily by macrophages and their derived cytokines and lipid mediators, which shift the phlogistic phase of inflammation into a non-phlogistic process that culminates with tissue repair and recovery of homeostasis (11, 12). Most importantly, the recruitment of macrophages and the production of pro-resolving mediators is triggered by enzymes synthesized during the acute inflammatory process (13). Macrophages play such a fundamental role in resolving inflammation and promoting tissue repair that impaired macrophage chemotaxis and/or macrophage depletion results in inefficient healing or chronic inflammation (14–16). Blocking acute inflammation with anti-inflammatory medications interferes, at least to some degree, with macrophage recruitment and the pro-resolving response, and often prevents effective resolution and recovery of homeostasis (12, 13). Targeted therapies for chronic joint inflammation should therefore have pro-resolving properties, which precisely combine pro- and anti-inflammatory mechanisms (12).
Synovial macrophages are the central drivers of the inflammatory response in osteoarthritic synovium (17, 18). In fact, synovial macrophage activation is directly related to disease activity, severity, and pain in OA-affected patients (19). However, this relationship is not causative. Synovial macrophages are also essential keepers of synovial homeostasis through phagocytic clearance and secretion of anti-inflammatory and pro-resolving cytokines, chemokines, enzymes, and growth factors (20–23). Following injury, synovial macrophages proliferate to form a protective immunological barrier in the synovial lining for intra-articular structures (24, 25). When regulatory functions are overwhelmed by the amount of damage, synovial macrophages upregulate inflammation, signaling to monocytes and other leukocytes (e.g., neutrophils and lymphocytes) to help counteract the increased demands for tissue repair and restore homeostasis (17, 26). During the progression of OA, the recruitment of myeloid monocytes into joints seems to be impaired (27), which combined with continuous joint damage, overwhelms the pro-resolving mechanisms of synovial macrophages, leading to degeneration (17, 27, 28).
The mononuclear cell fraction of bone marrow aspirates (bone marrow mononuclear cells -BMNC) is a rich source of pro-resolving macrophages that have been used therapeutically to improve tissue repair and inflammation resolution (29–39). Macrophages within BMNC are the main drivers of such effect, which’s documented pro-resolving functions include increased production of IL-10 (30, 33), diverse types of prostaglandins and specialized lipid mediators (13, 40, 41). The production of these molecules induce decreased production of IL-6 (33), increased phagocytic clearance of debris and apoptotic cells (efferocytosis) (30, 39) and enhanced PPAR-gamma signaling (42–47). Increasing the numbers of myeloid macrophages present in osteoarthritic knees, by injection of BMNC, restored joint homeostasis with long-lasting effects (48). Similarly, BMNC therapy increased counts of pro-resolving macrophages and induced marked resolution of joint inflammation in in vivo and in vitro models of equine synovitis (49, 50). In these models, there was a coordinated spectrum of pro-inflammatory, pro-resolving and anti-inflammatory events, including increased IL-10, IGF-1, and PGE2 production, and self-limiting IL-1 signaling (α and β). These events are all innately required for efficient synovial homeostasis and tissue repair and are commonly antagonized by therapeutic corticosteroids (30, 51–53). While these findings partially explain the durable effects of BMNC in the treatment of OA, little is known about BMNC-related mechanisms of resolution. Therefore, our purpose was to identify cellular mechanisms from BMNC driving joint homeostasis that could be used for developing targeted pro-resolving joint therapies and uncovering biomarkers of arthritis resolution. The aim of this study was to identify transcriptional signatures of BMNC leading to inflammation resolution using RNA-sequencing and relate these to the expression of key gene products in the synovial membrane. We hypothesized that gene networks linked to macrophage proliferation, negative regulation of inflammatory response, and to a lesser extent, NF-ƙB signaling, would be temporally upregulated in response to inflammation.
Discussion
In this study, we identified differential transcriptional signatures of BMNC in response to ISF and SF. These same BMNC were previously shown to resolve synovitis following exposure to an inflamed synovial environment in vivo and in vitro (49, 50). We demonstrate a temporal behavior of co-expressed gene networks and their association with traits from our previous studies (49, 50), as well as with the expression of key proteins in the synovium by immunohistochemistry. Our findings illustrate the elaborate balance of pro- and anti-inflammatory mechanisms shifting dominance through the recovery of joint homeostasis. BMNC responded to ISF with an early pro-inflammatory response (green module), characterized by a short spike in the expression of NF-ƙB-related genes, coincident with the peak of IL-1β secretion in conditioned ISF (50). This response was associated with increased expression of the blue and brown modules, 2 gene networks with homeostatic functions comprising known drivers of resolution, which were more highly expressed in ISF and were dominant among activated upstream regulators. Significant differences in the expression of phosphorylated PPARγ and mevalonate kinase in synovial membranes from inflamed joints treated with BMNC, and equivalent IL-1β staining between BMNC- and DPBS-treated joints, emphasize the fine tuning of so-called pro-inflammatory pathways that must remain active at physiological levels during the resolution process. These observations highlight the differences between the pro-resolving effects associated with BMNC therapy compared to the anti-inflammatory effects observed following clinical treatment with corticosteroids (30, 49–51, 69, 70).
The short spike of expression of the pro-inflammatory green module observed in ISF, is essential to trigger a cascade of events that culminates in a pro-resolving response. NF-ƙB-related and mitogen genes co-expressed in the green module (Supplementary Table 3) play a key role in promoting the proliferation of macrophages necessary to counteract damage (24, 25). NF-ƙB-related genes also induce increased expression of genes with anabolic and anti-inflammatory functions within the blue and brown modules, which have critical roles in driving resolution of joint inflammation. The top four overrepresented BPs in the green module relate to a group of genes sharing important mitogenic activity. Genes such as MAD1L1 and CHAMP1 encode proteins that interact and regulate cell structure organization preceding mitosis (71, 72). NR4A3 and NR1D2 encode transcriptional activators involved in proliferation, survival and differentiation of myeloid progenitor cells, and in adjusting myeloid progenitor cell metabolism to oxidative stress (73, 74). NR1D2 does so by activation of IL-6 transcription, which is also required to induce expression of the IL-4 receptor and related downstream regulatory functions of macrophages, including their self-renewal (75–77). The BP “response to estradiol” was characterized by the expression of the amphiregulin (AREG) and STAT3 genes. Macrophages are an important source of amphiregulin produced during acute inflammation. The AREG/ERK/STAT3 signaling axis is required for the differentiation of progenitor cells during tissue repair and establishing a pro-resolving response (78–81). AREG was more highly expressed in ISF and exhibited progressively decreasing expression over time, as the resolving response progressed (Supplementary Table 6). Additionally, estrogen accelerates the resolution of inflammation through the regulation of IL-10/STAT3-mediated deactivation of pro-inflammatory responses, such that post-menopausal women are prone to developing chronic inflammation (82). STAT3 was the activated upstream regulator in ISF cultures with the highest connectivity. The NIK/NF-ƙB signaling in the green module was highlighted by the expression of RELB and NFKB2. Both RELB and NFKB2 are subunits of the non-canonical NF-κB signaling, which in macrophages, can exert both pro- and anti-inflammatory effects (83, 84). Non-canonical NF-κB signaling is critical to produce SDF-1α and recruit monocytes to the site of damage immediately following injury (83). Further, during monocyte-macrophage differentiation, non-canonical NF-κB signaling prevents hyperactivation of new macrophages by accelerating the removal of RelA and c-Rel (canonical NF-κB subunits) from pro-inflammatory gene promotors preventing overt inflammation. As such, blocking non-canonical NF-κB by inactivation of its IKKα subunit results in increased inflammation (84). In both SF and ISF cultures, RELB expression was positively regulated, with decreasing expression over time, while RELA was downregulated (Figure 6; Supplementary Table 6). Combined, these signatures illustrate a fraction of molecular drivers of the acute response of BMNC to inflammation, which also sets the stage for establishing a pro-resolving response.
Increased expression of the blue module in response to inflammation, in parallel with the surge of the green module, and over a timeline associated with resolution in our previous studies (49, 50), suggests a pro-resolving identity. Genes encoding established drivers of joint inflammation resolution (IL-10, IGF-1) allocated to the blue module (Supplementary Table 2). A major functional signature of this module was the activation of the mevalonate pathway and isoprenoids biosynthesis, comprised by the expression of genes encoding central drivers of the mevalonate/isoprenoid pathway (COQ2, HMGCR, FDPS, HMGCS1, GGPS1, MVK, PDSS1, PDSS2, GGDPS1, FDPS, ACAA2, HADHA and ACAT2. In agreement, the super pathway of cholesterol biosynthesis, geranylgeranyl biosynthesis (an isoprenoid) and mevalonate pathway had the highest activation scores in ISF cultures between 0 and 48 hours, as detected by IPA. The roles of the mevalonate pathway in steroidogenesis, counteracting oxidative stress and inflammation resolution, are well documented in the macrophage response to damage and inflammation resolution (85–88). Deficiency of mevalonate kinase (MVK), a key enzyme in the mevalonate pathway, causes reduced synthesis of isoprenoids, leading to mitochondrial damage, subsequent oxidative stress and severe inflammation (89, 90). Importantly, exogenous isoprenoid treatment in models of inflammation induces decreased oxidative stress and production of inflammatory markers, by increasing expression of the NF-κB inhibitor IκBα and antioxidant selenoproteins (86, 91–93). Gene expression for MVK and HMGCS1, key enzymes in the mevalonate pathway, exhibited a similar expression profile between themselves (Figure 6; Supplementary Table 6). These enzyme genes were more highly expressed in SF, suggesting that inflammation negatively affects this pathway. A similar pattern of gene expression for IL10 and IGF1 was observed in our previous study, in which both genes were more highly expressed in SF than ISF (50). However, concentrations of IL-10 and IGF-1 in ISF were higher than in SF, because the decreased relative production in ISF was compensated for by higher macrophage counts (50). Our ongoing lipidomic study on BMNC-conditioned SF and ISF from the same samples used in this study may help elucidate if a similar context applies for isoprenoid production. Interestingly, it was recently evidenced that transcriptional dysregulation of the mevalonate pathway is a key signature of the overt inflammation caused by SARS-CoV-2 infection, highlighting its importance for inflammation resolution (88). Differences in MVK expression detected in synovial membranes suggests that the in situ activity of MVK in synovitis resolution happens earlier in time, as suggested by our pathway analysis and in vivo study (49).
The second overrepresented BP in the blue module “proteasome ubiquitin-dependent protein catabolism” was comprised of genes that characterize the formation of the 26S proteasome. Inflammation-derived oxidative stress damages nascent proteins that become misfolded and targeted for degradation (94). The 26S proteasome is essential for the degradation of these proteins, preventing aggregate formation, which is part of the pathogenesis of several conditions (94). Such observation from GO analysis agrees with the IPA findings where the “Unfolded Protein Response” was identified among the most activated pathways following isoprenoid biosynthesis through the mevalonate pathway. The next overrepresented BP was “protein localization to the Cajal body”. These are coiled bodies found in the nucleus of proliferating or metabolically active cells and are implicated in telomere homeostasis (95, 96). This BP was represented primarily by genes encoding chaperonin containing tailless (CCT) proteins that are critical regulators of telomerase folding and trafficking (97). Depletion of CCT proteins cause Cajal body and telomerase mislocalization and failure of telomere elongation (97). CCTs are also required for folding of cytoskeletal proteins during cell proliferation (98). The fourth dominant BP, “response to thyroid hormone stimulus”, reflects the effects of triiodothyronine on regulation of macrophage maturation and responses. Such responses include controlling cell migration and conferring protection against endotoxemia and LPS exposure, in great part through proliferation of tissue resident macrophages (99, 100). In addition, half of the genes characterizing this BP encode for cathepsins that mediate endolysosomal protein degradation. In summary, the dissection of a minute part of the blue module, illustrates the effect of isoprenoids and thyroid hormones in improving metabolism and performance of BMNC-derived macrophages to counteract the effects of inflammatory oxidative stress (85, 89, 90, 99, 100). This response is, at least partially, achieved by adjusting proteostasis through the 26S proteasome and Cajal bodies, preventing degenerative protein aggregate formation during increased cell transcription, proliferation and metabolism in response to inflammation (94, 97, 98).
The signature of the brown module were BPs that comprised of a series of gene groups encoding proteins that regulate the mitochondrial respiratory chain and mitochondria-mediated regulation of energy metabolism (101, 102). There is growing evidence of the pivotal role of mitochondria in energy metabolism adjustments required for inflammation resolution as shown in the brown module (103–105). In the face of inflammatory challenges, enhanced cell respiration induces oxidative stress, activating alternative sources of energy driving gluconeogenesis and enhancing mitochondrial fatty acid oxidation (101, 102) as denoted by the increased expression of genes such ACAA2 and HADHA in ISF cultures (Figure 6). These genes were, however, associated with the isoprenoid biosynthetic pathway, highlighting the recently reported role of mitochondrial isoprenoid biosynthetic process (106). Enhancing the mitochondrial respiratory chain is a homeostatic mechanism that prevents mitochondrial DNA damage and its subsequent cytosolic and extracellular release signaling through damage-associated molecular pattern (DAMP) receptors (107). The peroxisome proliferator-activated receptor-gamma (PPAR-γ) co-activator 1-α (PPRGC1A; PPRGC1A gene), a master regulator of mitochondria biogenesis and liver gluconeogenesis was among the outstanding genes involved in mitochondrial regulatory functions listed in the brown module (101) and exhibited a trend for higher expression in BMNC treated joints (Figure 8).
Following interaction with PPRGC1A, PPAR-γ exhibits increased activity, interacting with a multitude of transcription factors and PPAR-γ responsive elements (43). PPAR-signaling was an important activated pathway identified by IPA in both SF and ISF cultures between 48 and 96 hours (Table 1, Figure 6), denoting its homeostatic functions. Among different PPARs, PPAR-γ was one of the most highly connected upstream regulators. In macrophages and other cells, PPAR-γ signaling is a cornerstone of tissue repair and inflammation resolution, exhibiting myriad functions that are either PPAR-γ-mediated or -dependent (43). Examples include shifting the production of pro-inflammatory cytokines towards anti-inflammatory and pro-resolving mediators, driving apoptosis and clearance of neutrophils, enhancing macrophage traffic, recruitment, phagocytosis and efferocytosis, improving mitochondrial respiratory performance, and the overall transcriptome of a regulatory response driving recovery of homeostasis (43, 47, 101, 108). Of note, some isoprenoids, the signature of the blue module, as well as some specialized pro-resolving molecules, signal through PPAR-γ, conferring increased production of IL-10, resistance to inflammatory stimuli and attenuated NF-ƙB activation following LPS stimulation (43, 81, 109–111). Gene expression for PPARG was higher in ISF compared to SF. Significantly lower staining for phospho-PPAR-γ expression in synovial membranes detected by IHC, combined with the timing at which PPARG was identified as an Upstream Regulator, suggest that its activity modeling synovitis resolution almost overlap with the acute phase of inflammation. Importantly, PPAR-signaling findings from IPA at 48-96 coincide with the time at which pro-resolving effects were observed in our previous in vivo and in vitro studies (49, 50). Histochemical findings for phospho-PPAR-γ and PPRGC1A in BMNC compared to DPBS-treated controls suggests PPAR-γ signaling was not only a BMNC response to the inflamed synovial environment, but also part of the beneficial effects of BMNC on treated joints. Moreover, there is recent evidence that PPRGC1A expression is required for chondrocyte metabolism and cartilage homeostasis, with PPRGC1A knockouts exhibiting delayed endochondral ossification, disruption of physeal morphology and severe premature osteoarthritis (112).
Also in the brown module, CSF1 was identified as an upstream regulator gene and was more highly expressed in ISF than SF cultures. The marked proliferation of BMNC in ISF observed in vitro (50) may also be a response from CSF-1 signaling (50). Following a spike of proliferation, stimulation of macrophage progenitors with interferon-gamma (IFN-γ) or LPS induce maturation and cell cycle arrest, increasing MHC-II expression and developing the capability to quickly respond to inflammatory stimuli and antigen presentation (113). Given that synovitis in our model was induced by injection of LPS, it is not surprising that “antigen processing and presentation of polysaccharide antigen via MHC class II” was a dominant BP in the brown module, particularly considering that IFN-γ was not detected by immunoassay in conditioned SF or ISF in a preliminary screening performed in our study (50). Combined, these observations from the brown module highlight the importance of macrophage proliferation and maturation, and PPRGC1A/PPAR-γ signaling during macrophage-mediated joint homeostasis, identifying the need for further investigation of the therapeutic roles of PPAR-γ-agonists in the recovery of joint health.
The combined overrepresented BPs in the turquoise module, exclusive to ISF, reflect the dynamic of serine/threonine phosphorylation, autophosphorylation and dephosphorylation during cytokine signaling and subsequent mRNA transduction (114). In our study, these events happened in response to IL-4 (50). IL-4 signaling regulates the establishment of a pro-resolving response in macrophages by adjusting chromatin conformation and access for RNA transcription initiation (45). It also represses the expression of classical pro-inflammatory genes, decreasing inflammasome activation, IL-1β production and pyroptosis (44). Alternatively, IL-4 signaling enhances DNA binding of RNAPII-pS2 and RNAPII-pS5, and H3K27Ac as a positive epigenetic modification that leads to increased expression of several gene networks with a pro-resolving function (44). STAT6, which also allocates to the turquoise module, is a major regulator of IL-4 (likely through IL-4r expression) and PPAR-γ signaling (46). STAT6 facilitates binding of PPAR-γ to DNA, which, like IL-4, increases the number of regulated genes and the magnitude of response from macrophages (44, 46, 47, 115). IL-4r expression was increased in BMNC cultured in both SF and ISF, but was significantly higher in ISF (50), in agreement with the overall findings in this module. We did not, however, detect gene expression for IL-4 or the IL-4r agonist IL-13. Also, neither IL-4 nor IL-13 were detected in the synovial fluid (SF or ISF) in our in vitro and in vivo studies (29,31). Collectively, these findings suggest that, if produced by cells in the joint other than BMNC, their concentrations were either below detectable limits or absent, raising the possibility of signaling by an unidentified IL4r agonist or an unconsidered pathway with equivalent gene ontology. The decreasing mean expression of the turquoise module in relationship to the increasing mean expression of the blue and brown modules, indicates that the biological processes included in the turquoise module are primarily required for priming the macrophages in BMNC towards a pro-resolving response (45).
IL1B gene expression in BMNC and protein expression in synovial fluid coincided in time and duration, peaking at 24 hours and progressively decreasing over time (49, 50). The IL1B gene was identified as a centerpiece among inhibited upstream regulators in ISF. Interestingly, IL-1β expression in the synovial membrane at 6 days following LPS model induction remained evident in both BMNC- and DPBS-treated joints. This finding suggests that the physiological levels of IL-1β expression required for homeostasis were conserved and not blocked, as commonly observed with the use of corticosteroids (51). Staining patterns for the other selected markers were inconclusive due to variability between individual horses. Although our experimental design was aimed at minimizing variability by including only horses of the same breed and a narrow range of age, variability is expected when working with human populations and animal models. Inbred mice poorly mimic the inflammatory reaction of people (116). While this study comprised a small cohort, the horse is an excellent model for the translation of inflammation (117) and an established model for the study of degenerative and inflammatory joint disease (1, 118). In agreement with our in vivo (49) and in vitro (50) findings using the same horses, the combined effects of gene networks comprised by each module depict a pro-resolving response, dissecting some important signatures of this process, and identifying candidate biomarkers of synovitis resolution, such as PPAR-γ and MVK expression and synovial fluid quantification of isoprenoids. The overall agreement between different analytical tools used in this study (between gene ontology of WGCNA-derived modules with IPA pathways and upstream regulator analysis) reinforce the significance of our data. However, future mechanistic studies are necessary to further determine the specific roles of pathways and biological processes identified in our study. A study of the transcriptome analysis from synovial samples from naturally inflamed joints treated with BMNC and DPBS would complement our current observations and further illustrate the molecular drivers of the synovial response to BMNC injection.