Genetic Control of Lyme Arthritis by Borrelia burgdorferi Arthritis–Associated Locus 1 Is Dependent on Localized Differential Production of IFN-β and Requires Upregulation of Myostatin

This article is featured in In This Issue, p.3385

Lyme disease, caused by infection with the bacteria Borrelia burgdorferi, affects 300,000 Americans each year (1). Disease outcomes range from acute to chronic, sometimes resulting in irreversible damage to the nervous (2) and cardiovascular (3, 4) systems. Arthritis is the most common late disease manifestation (5), affecting up to 60% of patients and persisting in 10% of patients despite antibiotic therapy (6). Lyme arthritis is frequently characterized by synovitis at one or both knee joints and may persist for years postinfection (6, 7). Although bacterial and host factors contribute to the spectrum of Lyme disease severity (8), many studies have shown that severe Lyme arthritis can result from genetically controlled dysregulated immune responses (9–17).

The spectrum of disease severity in humans can be modeled in inbred strains of mice that display different disease outcomes upon infection with B. burgdorferi (13). Specifically, C57BL/6 (B6) mice develop mild arthritis and C3H mice develop severe arthritis in response to the same B. burgdorferi inoculum. We have successfully used these mice during empirical (18–20) and forward genetic (21–23) approaches to identify genetic determinants of Lyme arthritis severity. Empirical approaches comparing the global gene-expression profiles of B6 and C3H joint tissue revealed a type I IFN signature in C3H mice that is absent from B6 mice (18), and a receptor-blocking Ab and receptor ablation showed that type I IFN is required for maximally severe Lyme arthritis (19, 20). Independently, forward genetic approaches identified a quantitative trait locus (QTL), B. burgdorferi arthritis–associated locus 1 (Bbaa1), which contains the type I IFN gene cluster and controls Lyme arthritis severity (21). An interval-specific congenic line in which the Bbaa1^C3H allele was introgressed onto the B6 background (B6.C3-Bbaa1) displays increased Lyme arthritis relative to B6 mice (21), unless the animals are pretreated with the type I IFNR-blocking Ab (22), functionally linking the expression of type I IFN within Bbaa1 to increased Lyme arthritis. Interestingly, Bbaa1^C3H was also found to control disease severity in the K/B×N serum-transfer model of rheumatoid arthritis (RA) through type I IFN production (22), indicating that insight gleaned from our model extends to a distinct inflammatory arthritis.

Our finding of a pathologic type I IFN profile is consistent with several publications that have shown a type I IFN signature in the serum of human Lyme patients during active disease (2, 24) and in the synovial fluid of treatment-refractory RA patients (25). Arthritis is also a transient side effect in hepatitis C and multiple sclerosis patients treated with IFN-α/β (26, 27), further supporting the linkage between elevated levels of type I IFN and arthritis. However, type I IFN is a key element in the innate and adaptive response against a variety of microbial infections, posing an inherent problem to treatment blockade. Because type I IFN regulation requires a sensitive balance and because IFN-α and IFN-β members play unique roles in various infections, we sought to identify the specific type I IFN and downstream mediators in Lyme arthritis pathogenesis. In this study, the B6.C3-Bbaa1 mouse allowed us to identify the specific type I IFN member that causes Bbaa1 Lyme arthritis, segregate cells that initiate type I IFN in B. burgdorferi infection from those that respond to it, and identify myostatin as an unexpected mediator of IFN-regulated Lyme arthritis development.

B6 and C3H mice were obtained from The Jackson Laboratory. B6.C3-Bbaa1 mice (Chr4: 11.6–93.46 Mbp) were generated previously (21) and maintained as a colony in the University of Utah Animal Research Center. Interval-specific recombinant congenic lines (ISRCL1–4) with Chr4 Bbaa1 intervals 11.6–77.8, 76.48–93.46, 83.7–93.46, and 88.3–93.46 Mbp were independently generated through repeated backcrosses of B6.C3-Bbaa1 mice to the parental B6 line. Filial offspring were selected based on single nucleotide polymorphism identification by high-resolution melting analysis, as described (28), and homozygous lines were fixed by mating littermates. All mice used in this study were housed in the University of Utah Animal Research Center and handled with protocols approved by the Institutional Animal Care and Use Committee at the University of Utah in accordance with the National Institutes of Health guidelines for the care and use of animals.

B. burgdorferi strain N40 was cultured for 4 d in Barbour-Stoenner-Kelly II medium containing 6% rabbit serum (Sigma-Aldrich). Mice were infected with 2 × 10⁴ live spirochetes intradermally into the skin of the back (29). Arthritis was assessed by rear ankle joint measurements obtained with a metric caliper at days 0 and 28 postinfection and by histopathological analysis of the most severely swollen rear ankle joint following fixation, decalcification, and H&E staining, as described previously (30). Histopathological scores were determined blindly and ranged from 0 to 5 for various aspects of disease, including severity and extent of the lesion, PMN leukocyte and mononuclear cell (e.g., monocyte, macrophage) infiltration, tendon sheath thickening (e.g., synoviocyte and fibroblast hypertrophy/hyperplasia), and reactive/reparative responses (e.g., periosteal hyperplasia and new bone formation and remodeling), with 5 representing the most severe lesion and 0 representing no lesion. Infection was confirmed in mice euthanized 28 d postinfection by ELISA quantification of B. burgdorferi–specific IgG concentrations in serum (31) and 16S rRNA B. burgdorferi transcripts in joints (18, 32) or B. burgdorferi DNA in ear tissue (33).

mAbs were used to neutralize murine IFNAR1 (MAR1-5A3), IFN-α (TIF-3C5), or IFN-β (HDβ-4A7) (Leinco Technologies). TIF-3C5 is considered a “pan–IFN-α” mAb because it neutralized all IFN-α subtypes available for testing: αA, α1, α4, α5, α11, and α13 (34). Mice received mAbs or isotype controls by i.p. injection following a dosage regimen described previously (34, 35). Briefly, anti-IFNAR1 was administered in a single 2.5-mg dose the day before infection with B. burgdorferi (19, 22, 35), anti–IFN-β was administered in two doses (300 μg each, 600 μg total) the day before and 2 d following infection, and anti–IFN-α was administered in three doses (333 μg each, 1 mg total) the day before and days 1 and 3 postinfection. Boosts were determined based on previously reported pharmacokinetics (34) and in consideration that the IFN profile peaks at 7 d postinfection with B. burgdorferi (18). For bone marrow–derived macrophage (BMDM) experiments, 10 μg/ml of each mAb was administered in combination with the stimulation. The myostatin inhibitor (MBP-fMSTNpro45-100-Fc) is a recombinant peptide containing a C-terminal mouse Fc domain fused to the fish truncated propeptide-1, the production of which was described previously (36). The Fc fusion did not affect the myostatin-inhibitory activity of the truncated propeptide-1, with the potency comparable to commercial full-length mammalian myostatin propeptide (Supplemental Fig. 4). Mice received four i.p. injections of vehicle (PBS) or myostatin inhibitor (400 μg each, 1.6 mg total) once weekly beginning 1 d postinfection.

Chimeras were generated in all pairwise combinations among B6 (CD45.1), B6 (CD45.2), and B6.C3-Bbaa1 (CD45.2) using a rapid reconstitution protocol, as described (20, 37). Rapid reconstitution involves transplanting donor splenocytes into irradiated recipient mice and allows for the infection of young mice < 8 wk old, which is necessary for a reliable arthritis phenotype. Chimerism was evaluated by flow cytometric analysis of peripheral blood leukocytes 25 d posttransplant (Supplemental Fig. 2A).

Mouse rear ankle joints were gently digested into single-cell suspensions, as described (20). Briefly, skin was removed, and tibiotarsal tissue was teased away from bone using 20-gauge syringe needles, followed by 1 h incubation at 37°C in RPMI 1640 containing 0.2 mg/ml endotoxin-free Liberase TM (Roche) and 100 μg/ml DNase I (Sigma-Aldrich). Single-cell suspensions were filtered through a 100-μm cell strainer, RBCs were lysed in ammonium-chloride-potassium buffer, and cells were labeled with biotinylated anti-CD45.2 (BioLegend), followed by streptavidin magnetic bead labeling (Miltenyi Biotec). Magnetic bead separation was performed on MS columns (Miltenyi Biotec), according to the manufacturer’s instructions. Flow cytometric analysis revealed >85% purity in the CD45⁻ fraction, as reported previously (20). CD45⁻ cells from both rear ankle joints of two mice were pooled for each n sample to increase the RNA concentration for transcript analysis (Figs. 6, 7A). For ex vivo stimulation, CD45⁻ cells from both rear ankle joints of more than eight mice were pooled, plated in RPMI 1640 containing 2% FBS (20), and stimulated with live B. burgdorferi (6 × 10⁶ per milliliter) or 100 U/ml IFN-β (PBL Laboratories) for 3 h (Fig. 7B, Supplemental Fig. 3).

BMDMs were prepared by culturing bone marrow isolated from the femurs and tibias of mice for 7 d in L929 cell–conditioned media as a source of M-CSF, as previously described (38). Harvested macrophages were then replated in 24-well dishes at a density of 6 × 10⁵ per milliliter in media containing 1% of the serum replacement Nutridoma (Roche). Cultures were stimulated with live B. burgdorferi (6 × 10⁶ per milliliter) for 6 h at 37°C with 5% CO₂.

p-Stat1 was stained as described (34, 39), following a 15-h incubation with B. burgdorferi and type I IFN blocking mAbs. Briefly, cells were incubated in 0.05% trypsin (Fisher Scientific) and scraped to detach from the plate. Cells were fixed using 1.5% paraformaldehyde and permeabilized with 100% methanol. p-Stat1 was stained using an unconjugated Phospho-Stat1 mAb (pY701; Cell Signaling Technology) at a 1:200 dilution and incubated for 1 h at room temperature, followed by a secondary stain with Goat anti-Rabbit IgG conjugated to Alexa Fluor 647 (Fisher Scientific) at a 1:500 dilution and a 30-min incubation at room temperature. Data were collected on a FACSCanto II (BD Biosciences) flow cytometer and analyzed using FlowJo v.10.0.8 software.

Total RNA was recovered using TRIzol Reagent (Invitrogen) and purified using the Direct-zol RNA MiniPrep kit (Zymo Research). For RNA sequencing (RNA-seq), libraries were prepared using PolyA enrichment and sequenced with Illumina HiSEquation 50 Cycle Single-Read Sequencing version 4 at the University of Utah High Throughput Genomics Core Facility. Sequences were annotated with help from the University of Utah Bioinformatics Core Facility. Briefly, reads were aligned to the mouse genome from Ensembl release 87 using STAR (40) and assigned to genes using featureCounts (41). Differentially expressed genes were identified using a 5% false discovery rate with DESeq2 (42). The RNA-seq data presented in this article have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus repository (https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE102748. For quantitative RT-PCR (qRT-PCR) analyses, RNA was reverse transcribed, and transcripts were quantified using a Roche LC-480, according to our previously described protocols (21). Primer sequences used in this study for β-actin, Iigp (18), Oasl2, Cxcl10, Tyki (19), Gbp2 (43), Tnfa, and Ifnb (44) can be found in the indicated citations. Mstn primer sequences designed in this study were as follows: forward, 5′-GATCTTGCTGTAACCTTCCC-3′ and reverse, 5′-CTCCTGAGCAGTAATTGGC-3′.

All graphical data depict the mean ± SEM. Statistical calculations were performed using GraphPad Prism 7.0b software. Continuous variables were analyzed by one-way ANOVA with the Dunnett post hoc test for pairwise comparisons or by the Student t test (two-tailed). Categorical variables were analyzed by the Mann–Whitney U test (two-tailed). Statistical significance (*p <0.05, **p <0.01, ***p <0.001, ****p <0.0001) is indicated in the figures.

Both empirical and forward genetic studies from our laboratory have converged on the finding that a pathologic type I IFN profile drives Lyme arthritis in C3H mice (18–22). In this regard, we identified differential expression of type I IFN as a positional candidate for an expression QTL (eQTL) underlying Bbaa1 (22). Nevertheless, there are multiple members of the type I IFN family, with the IFN-α and IFN-β subclasses playing the largest roles in pathogenesis, and although each signals through a shared receptor heterodimer, their unique roles in host protection and pathogenesis are becoming increasingly appreciated (34, 45).

To assess the specific contributions of IFN-α and IFN-β to Lyme arthritis development, B6.C3-Bbaa1 mice were treated with a pan-acting mAb that blocks multiple IFN-α subtypes (TIF-3C5) or a mAb that targets IFN-β specifically (HDβ-4A7). We showed previously that the anti-IFNAR1 mAb (MAR1-5A3) suppresses arthritis development to the level of B6 mice (22). Because blocking IFN-β resulted in the same reduction in Lyme arthritis as blocking IFNAR1 (Fig. 1A), we conclude that IFN-β, and not IFN-α, is the proarthritic cytokine generated in B6.C3-Bbaa1 mice. Analysis of anti-B. burgdorferi IgG in the serum and B. burgdorferi 16S rRNA transcripts in the joint confirm that blocking type I IFN does not impact the ability of the host to generate a B cell response or to control infection (Fig. 1B). Taken together, these findings suggest that differential expression of IFN-β, and not IFN-α, may be the eQTL within Bbaa1 controlling the severity of Lyme arthritis.

We previously showed that BMDMs from B6.C3-Bbaa1 mice express higher levels of IFN-inducible transcripts in response to B. burgdorferi than do BMDMs from B6 mice (22). Thus, B6.C3-Bbaa1 BMDMs were used as a surrogate for myeloid cells in joint tissue to assess the specific contribution of IFN-α or IFN-β to the Bbaa1-directed induction of type I IFN. BMDMs were stimulated with B. burgdorferi and simultaneously treated with each mAb for 6 h. Isotype-control mAbs had no effect, as shown for murine IgG_2a (Supplemental Fig. 1A). Treatment with anti–IFN-β resulted in complete suppression of IFN-inducible genes, which was indistinguishable from levels of macrophages treated with IFNR blockade (Fig. 2A). In contrast, treatment with anti–IFN-α had no effect on the expression of IFN-inducible genes, indicating that IFN-β is the critical eQTL within Bbaa1 driving Lyme arthritis development. Importantly, none of these treatments impacted the expression of Tnfa (Fig. 2A), which is downstream of the critical MyD88-dependent host defense pathway (46, 47). This highlights the in vivo observation that type I IFN is not required for host defense in Lyme arthritis (Fig. 1B), and its suppression does not influence the expression of initial MyD88-dependent cytokines (20, 22)

Because Stat1 is phosphorylated when IFN-α and IFN-β proteins bind the receptor (34, 48), flow cytometric analysis of p-Stat1 protein was used as a complementary approach to determine the impact of IFN-α and IFN-β on feed-forward IFN amplification. To allow for transcription and translation of type I IFN proteins, Stat1 phosphorylation was assessed in BMDMs 15 h poststimulation with B. burgdorferi and mAb treatment. Strikingly, IFN-β neutralization completely inhibited Stat1 phosphorylation, similar to levels in macrophages treated with receptor-blocking mAb, whereas blocking IFN-α had no effect (Fig. 2B). Because anti–IFN-α did not impact arthritis severity or macrophage responses to B. burgdorferi, mAb functionality was assessed with rIFN-α (Supplemental Fig. 1B). Confirmation of anti–IFN-α activity supports the conclusion that IFN-β (but not IFN-α) is the type I IFN made in response to B. burgdorferi in vivo and in vitro and that not even low levels of IFN-α contribute to the feed-forward response. Interestingly, Stat1 deficiency was previously reported not to influence arthritis severity in B. burgdorferi–infected C3H mice (49), suggesting that the macrophage responses of congenic mice may allow a focused appreciation of Stat1 involvement in type I IFN signaling that may be limited by the activation of compensatory pathways in the joint tissue.

Multiple lines of evidence indicate that differential expression of IFN-β may be the eQTL underlying Bbaa1, leading to the question of whether the proarthritic effect of IFN-β expression acts through hematopoietic or resident cells. MHC compatibility between mildly arthritic B6 mice and more severely arthritic B6.C3-Bbaa1 mice allowed reciprocal radiation chimeras to be generated (Fig. 3A). Using a rapid reconstitution protocol previously developed by our laboratory (see 2Materials and Methods), myeloid compartments were sufficiently reconstituted by donor-derived cells (∼80%) prior to infection (Supplemental Fig. 2A), and reconstitution was determined to be adequate for host defense based on the similar abilities of irradiated mice to control infection (Supplemental Fig. 2B).

As expected, B6 mice reconstituted with autologous splenocytes (B6→B6) developed mild arthritis, and B6.C3-Bbaa1 mice reconstituted with autologous splenocytes (Bbaa1→Bbaa1) developed more severe arthritis 4 wk postinfection with B. burgdorferi (Fig. 3B). Notably, B6.C3-Bbaa1 mice reconstituted with hematopoietic cells from B6 mice (B6→Bbaa1) displayed mild arthritis compared with Bbaa1→Bbaa1 chimeras, and B6 mice reconstituted with cells from B6.C3-Bbaa1 mice (Bbaa1→B6) developed more severe arthritis than B6→B6 chimeras (Fig. 3B). Thus, genes encoded within Bbaa1 regulate Lyme arthritis severity through the radiosensitive hematopoietic cellular constituents of the joint.

Considering that differential expression of IFN-β may be the eQTL underlying Bbaa1, we can further deduce that radiosensitive cells in the joint, which are likely myeloid (20), initiate the production of proarthritic IFN-β (Fig. 3C). This is consistent with our previous findings that CD45⁺ cells harvested from a naive mouse joint possess the unique ability to generate an IFN profile upon ex vivo stimulation with B. burgdorferi, whereas CD45⁻ cells are capable of responding to, but not initiating, type I IFN in response to B. burgdorferi (Supplemental Fig. 3, 20). Another striking finding is that Bbaa1→B6 chimeras develop the full arthritis phenotype, strongly suggesting that, regardless of genotype, radioresistant resident cells are fully competent to choreograph arthritis manifestation in the joint (Fig. 3B, 3C). This is consistent with the previous finding that endothelial cells and fibroblasts, both of which are radioresistant cell types in the joint, are major responders to, and amplifiers of, type I IFN, resulting in the production of chemokines in infected C3H mice (20). To our knowledge, this study provides the first in vivo evidence for the “pass off” that occurs between cells that initiate IFN-β in response to B. burgdorferi and cells that respond to IFN-β to direct arthritis development (Fig. 3C).

Because the original B6.C3-Bbaa1 congenic interval was very large (Chr4: 11.6–93.46 Mbp), with >450 genes in addition to Ifnb, we generated and studied a panel of overlapping interval-specific recombinant congenic lines across Bbaa1 (ISRCL1–4). This allowed us to further interrogate the relationship between the differential expression of IFN-β as an eQTL for Bbaa1 and the pathogenesis of Lyme arthritis (Fig. 4). ISRCL1 encompasses the largest portion of the Bbaa1^C3H interval but excludes the type I IFN locus, whereas ISRCL2–4 retain the C3H type I IFN gene cluster with further reduction in the size of the C3H donor sequence (Fig. 4). Phenotypic analysis following infection with B. burgdorferi revealed that arthritis severity cosegregates with the C3H-derived type I IFN locus (Fig. 4). Because blocking IFN-β in B6.C3-Bbaa1 mice completely suppressed Lyme arthritis (Fig. 1A), we deduce that the phenotype in ISRCL2–4 mice is similarly dependent on IFN-β production.

BMDMs were used to confirm that genes contributing to heightened IFN-β expression are retained in the narrowed Bbaa1 interval. Macrophages from ISRCL3 and ISRCL4 mice, which possess the C3H allele for the type I IFN locus, displayed greater IFN-inducible transcriptional responses for Cxcl10 and Gbp2 than did macrophages from B6 mice, with trending differences for other transcripts (Fig. 5). However, macrophages from the congenic mice failed to attain the maximal level of IFN signature transcripts seen in macrophages from C3H mice, indicating that additional genetic loci contribute to the fully expressed IFN response in C3H mice. As shown, Ifnb expression is low, but none of the IFN-α transcripts were detectable, further supporting the role of IFN-β in driving Lyme arthritis pathogenesis. The narrowed Bbaa1 congenic lines ISRCL3 and ISRCL4 have provided a refined tool to study the involvement of type I IFN in Lyme arthritis.

Now that we have established the role of radioresistant resident cells in directing Bbaa1 Lyme arthritis downstream of IFN-β, we pursued a better understanding of the mechanism of arthritis development through transcriptome analysis of CD45⁻ (resident) joint cells. Newly developed ISRCL3 and ISRCL4 lines were compared with B6 mice to capture all genes related to IFN-β production and arthritis development with a minimal physical interval. To capture the stage of active arthritis development (as opposed to full-blown arthritis with confounding wound repair pathways), joint cells were harvested from mice 22 d postinfection with B. burgdorferi (29). Following recovery of single-cell suspensions, the CD45⁻ population was isolated by magnetic bead separation, and RNA-seq was performed.

Not surprisingly, comparing the transcriptomes of these highly similar genotypes at a time point submaximal to arthritis development revealed very few differences in the gene-expression profiles (Fig. 6A). Of those differences, only five genes were independently identified in ISRCL4 versus B6 and ISRCL3 versus B6 comparisons with a ≥1.5-fold change (in either direction) and an adjusted p value < 0.05 (Fig. 6A, Table I), supporting their involvement in Bbaa1-directed Lyme arthritis. Importantly, myostatin (Mstn) had the greatest induction and achieved the highest level of significance in independent analyses with both interval-specific congenic lines (Fig. 6A, Table I), identifying it as a strong candidate for Bbaa1 Lyme arthritis development. qRT-PCR analysis of Mstn expression in CD45⁻ cells from uninfected animals revealed similar expression among uninfected mice of all three genotypes (data not shown) and revealed that genes within Bbaa1 specifically regulate the induction of myostatin following infection with B. burgdorferi (Fig. 6B). This was a striking finding for Bbaa1 Lyme arthritis development in light of recent evidence that myostatin is involved in RA development in humans and mice (50).

To ascertain a direct effect between IFN-β production and myostatin expression in Bbaa1 Lyme arthritis, IFN-β was blocked in ISRCL4 and ISRCL3 mice infected with B. burgdorferi (as described in Fig. 1). CD45⁻ cells were isolated from joint tissue 22 d postinfection, and myostatin expression was assessed by qRT-PCR. Importantly, CD45⁻ cells from ISRCL4 and ISRCL3 mice infected with B. burgdorferi and treated with anti-IFN-β mAb displayed a 2-fold reduction in Mstn transcripts compared with mice treated with isotype control (Fig. 7A). The fold suppression by blocking IFN-β is internally consistent with the level of induction found in our RNA-seq (Fig. 6). This directly connects the genetic control of differential expression of IFN-β to the regulation of myostatin expression as a downstream effector.

Several proinflammatory cytokines, TNF-α, IL-1α, and IL-17, were previously shown to directly induce expression of myostatin in the context of RA, but the impact of IFN-β was not addressed (50). To test whether IFN-β directly induces expression of myostatin, CD45⁻ cells were isolated from the joints of naive B6 mice and stimulated ex vivo with IFN-β, B. burgdorferi, or IFN-β and B. burgdorferi. Mstn expression assessed by qRT-PCR revealed that IFN-β works synergistically with B. burgdorferi to directly induce Mstn expression (Fig. 7B). To our knowledge, this is the first report of IFN-β induction of myostatin.

To determine whether myostatin is a marker or a mediator of Bbaa1 Lyme arthritis, infected ISRCL4 and ISRCL3 mice were treated with an inhibitory propeptide of myostatin that is known to efficiently bind myostatin and block it from interacting with its receptors in vivo and in vitro (Supplemental Fig. 4, 36, 50, 51), and the impact on arthritis development was assessed. Consistent with the dual requirement of B. burgdorferi and IFN-β for myostatin upregulation (Fig. 7), myostatin inhibition did not impact ankle swelling in uninfected animals (Fig. 8A). However, myostatin inhibition in infected ISRCL4 and ISRCL3 mice led to a remarkable suppression in arthritis 4 wk postinfection with B. burgdorferi (Fig. 8A, 8B). Importantly, myostatin inhibition did not impact host defense, because anti–B. burgdorferi IgG in the serum and B. burgdorferi numbers in tissues were similar in treated and untreated groups (Fig. 8C). These findings demonstrate a direct effect of myostatin on joint-specific inflammatory responses.

In this study, we established IFN-β as the Lyme arthritis–driving type I IFN that is regulated by a locus, Bbaa1, previously identified by forward genetics (Fig. 1). This is the culmination of a decade of our work: two parallel pathways of investigation revealed a pathologic type I IFN profile in C3H mice infected with B. burgdorferi that is suppressed by IFNAR1 blockade (14, 18–22). Other investigators have corroborated the association between pathologic type I IFN production and Lyme disease pathogenesis in murine studies (52, 53), as well as in human patients at various stages of disease (2, 24), and many investigators have examined the type I IFN response to B. burgdorferi in murine and human cells (19, 20, 43, 53–62). However, the exact pathogenic member has remained elusive because of the transient expression of IFN-α/β (63, 64), overlapping IFN-stimulated gene pathways induced by IFN-α/β (65), and differing contexts of B. burgdorferi infection in which IFN-α/β transcripts have been detected. The strength of this study centers on the use of B6.C3-Bbaa1 mice, which allowed the impact of type I IFN dysregulation to be assessed in isolation from the five other B. burgdorferi arthritis–associated QTLs contained within the C3H genome (21), coupled with the use of newly available mAbs to assess IFN-α and IFN-β individually.

Multiple mAb experiments, along with the transcriptional detection of Ifnb in BMDMs from refined congenic lines (and lack of evidence for Ifna transcripts), revealed that IFN-β is the sole contributor to the type I IFN profile and arthritogenesis in Bbaa1-directed Lyme arthritis (Figs. 1, 2). Another major finding was that partial control of IFN-β is intrinsic to the Bbaa1 locus (Figs. 4, 5). The intrinsic regulation of IFN-β within Bbaa1 was surprising because of the lack of coding or regulatory single-nucleotide polymorphisms within 10,000 flanking base pairs of the Ifnb genetic sequence (66, 67); it suggests a novel mechanism for IFN-β regulation, which is a topic of future investigation. In this study, the congenic mice provided a unique tool to segregate IFN-β initiation from feed-forward amplification, and radiation chimeras demonstrated that the Bbaa1 arthritis–initiating lineage is radiation sensitive in the joint (Fig. 3). This finding is in contrast to our previous publication on the Bbaa2 QTL, in which the hypomorphic allele of Gusb drives arthritis through accumulation of glycosaminoglycans in the radiation-resistant resident joint population (23). Thus, our forward genetic approach has identified two distinct QTLs that regulate Lyme arthritis through independent mechanisms and responsible initiating cell lineages.

In the case of Bbaa1, the discrimination among cell types, along with refined congenic lines, was critical to understanding the mechanism of arthritis development. An unbiased RNA-seq on resident (CD45⁻) joint cells led to the third major finding that IFN-β acting as an eQTL underlying Bbaa1 regulates myostatin expression (Figs. 6, 7, Table I). This was surprising given that myostatin is widely appreciated for its role as a negative regulator of skeletal muscle growth and regeneration (68) and has been intensely investigated for its positive impact on food production in the animal agriculture industry. Nevertheless, we were able to use a newly developed reagent in the field of muscle development to inhibit myostatin protein activity in vivo and discovered that myostatin is a direct mediator of Bbaa1 Lyme arthritis development (Fig. 8, Supplemental Fig. 4). Strikingly, myostatin was recently found to be a mediator of a distinct inflammatory arthritis using the TNF-α overexpression model of RA in mice and was noted to be expressed by synovial fibroblasts from patients with RA (50).

We do not have a complete picture of the means by which the IFN-β to myostatin axis results in Lyme arthritis development and find it surprising that myostatin is not encoded by any of the QTLs previously identified by forward genetics between B6 and C3H mice (14, 21). Clearly, the impact of IFN-β on the pathogenesis of Lyme disease involves a complex web, and the development of congenic mice with reduced physical intervals provides a unique tool for identifying the tissue-specific components of pathogenesis. Of great interest is the recent demonstration that myostatin promotes RA in TNF-α–overexpressing mice by activating osteoclast differentiation (50). Although bone pathologies have not been widely investigated in Lyme arthritis patients or mice, Tang et al. (69) recently identified B. burgdorferi infection in mice bone and found that inhibition of osteoblast activity led to bone destruction. Because B. burgdorferi also infects quadriceps muscle (70, 71), our identification of myostatin suggests that a more global involvement of bone and muscle in the response to B. burgdorferi may be involved in Lyme arthritis development than previously appreciated. Future studies are needed to investigate the role of myostatin in bone and other tissues of the joint in Lyme arthritis development and the ability to reverse clinical disease once established.

This study capitalized on the power of forward genetics in studying pathogenesis in an animal model, wherein refined congenic lines afforded the opportunity to assess transcriptional differences limited to extremely narrow genomic differences using RNA-seq, and radiation chimeras identified the “hand off” of cells involved in arthritis development. Additionally, myostatin is not encoded within the Bbaa1 QTL, illuminating the vast potential of control exerted by this limited genetic region. Future studies will be directed at delineating the important sources and inflammatory targets of myostatin in vivo, which could uncover unrecognized signaling pathways in this inflammatory arthritis. Myostatin-directed therapies are an appealing distinction from classic anti-inflammatory targets, because antibacterial responses are expected to remain intact. Additionally, myostatin inhibition is in various stages of clinical trials for patients with muscle-wasting conditions and is receiving attention as a potential therapeutic target for many metabolic disorders (68); these studies could also provide insight into the potential for inflammatory disorders, such as arthritis.

We thank Scott Hale, Dean Tantin, and members of the Weis laboratory for helpful discussions during the course of these studies. We also thank Brian Dalley and the High Throughput Genomics Core Facility for expert guidance and for performing the RNA-seq, Chris Stubben (High Throughput Genomics Core Facility) for analyzing the sequencing data, and Anna Dilger (University of Illinois at Urbana-Champaign) for insightful discussions on myostatin-inhibitory reagents.

The authors have no financial conflicts of interest.