TNF-Induced Interstitial Lung Disease in a Murine Arthritis Model: Accumulation of Activated Monocytes, Conventional Dendritic Cells, and CD21⁺/CD23⁻ B Cell Follicles Is Prevented with Anti-TNF Therapy

Rheumatoid arthritis (RA) is a chronic and systemic inflammatory disease estimated to affect 1% of the world’s population (1–4). Outside of the well-known inflammatory articular involvement, RA is often accompanied by a number of systemic manifestations. These include inflammatory nodules, cardiovascular disease, and pulmonary involvement (5). In particular, interstitial lung disease (ILD) is one of the primary extra-articular manifestations in RA. RA-associated ILD (RA-ILD) is a leading cause of morbidity and mortality in RA patients (5–9). This process is progressive and considered to be largely irreversible.

RA-ILD has two predominant subtypes as determined by chest computed tomography (CT) imaging and histologic features. These subtypes include the usual interstitial pneumonia (UIP) pattern, which exhibits a predominantly fibrotic disease process, and the nonspecific interstitial pneumonia (NSIP) pattern, which is more inflammatory in nature (9–20). The UIP pattern is the predominant subtype identified in RA-ILD cases and is composed of notable dense fibrosis with remodeling of the lung architecture. NSIP, the second-most prevalent subtype, is defined by the presence of diffuse and symmetric chronic interstitial inflammation, with retained lung architecture and less fibrotic change (21).

The factors that underlie the development and progression of ILD remain enigmatic. A large body of evidence supports the complex role of both genetic and environmental factors in the pathogenesis of RA-ILD, but the cause remains widely unknown (22–27). One theory of pathogenesis postulates that the pulmonary disease manifests in two sequential stages over time. This progressive pathologic condition involves a primary exacerbation, which initiates an inflammatory state predominated by cellular infiltration, leading to an NSIP pattern of pathologic conditions. This is followed by secondary environmental triggers, leading to fibrotic change in the peripheral pulmonary tissue, consistent with a UIP pattern (28, 29). The idea of distinct stages of progressive disease fits well with the previously addressed epidemiologic studies looking at the specific subsets of RA-ILD. However, clinical evidence for this model is difficult to attain because most patients are diagnosed following symptom presentation, when more advanced fibrotic changes are already established and lung function is significantly compromised. Given the limitations related to screening for ILD in RA patients, identifying a cohort of patients with early changes on lung imaging to study disease progression prior to symptom manifestation presents a formidable challenge.

TNF is an appealing target in RA-ILD because it is a pivotal cytokine in the development and perpetuation of both articular (27, 30, 31) and pulmonary inflammation (32–35). Furthermore, elucidating the function of TNF in pulmonary pathology associated with underlying RA is particularly crucial because of the central importance of TNF inhibition in treating joint inflammation. As previously discussed, TNF overexpression is associated with a variety of phenotypes; yet it is known that RA-ILD can progress and even worsen (particularly the fibrotic component) in patients with RA taking anti-TNF agents (36–42). Regrettably, the reports on progressive fibrosis do not comment on alterations of interstitial inflammation in the lung of RA-ILD patients on anti-TNF therapy. It is certainly plausible that anti-TNF therapy during the inflammatory phase of ILD proposed in the previously described model (28) may block the progression to lung fibrosis and challenge the irreversibility of this disorder. Understanding how TNF influences lung inflammation and fibrosis is critical because of the role of anti-TNF therapy as a cornerstone of RA treatment.

We work with the TNF-transgenic (TNF-Tg) mouse model (43, 44), which produces systemic TNF overproduction that causes inflammatory-erosive arthritis and ILD. Data support the concept that TNF mediates some aspects of pulmonary pathological conditions, but the specific aspects induced through TNF signaling are unknown. To address this gap, we did the following: 1) identified the inflammatory cells that accumulate in the lungs of TNF-Tg mice during the genesis of ILD and 2) examined the irreversible nature of inflammatory ILD in this mouse model of RA using anti-TNF therapy on established disease. Our results demonstrate a significant increase in activated monocytes and conventional dendritic cells (cDC2) in the lungs isolated from TNF-Tg mice with inflammatory ILD. We also identified CD21⁺/CD23⁻ B cells, which are phenotypically distinct from the CD21^hi/CD23⁺ B cells in inflamed nodes (Bins) that accumulate in joint-draining lymph nodes of TNF-Tg mice (45) and RA patients (46). In contrast to human RA-ILD, anti-TNF treatment significantly alleviates both joint and lung inflammation. Taken together, these results identify a potential role for activated monocytes, cDC2, and CD21⁺/CD23⁻ B cells in the genesis of RA-ILD, which are restricted to a previously unknown, reversible prefibrotic stage of the disease.

All animal research was conducted using protocols approved by the University of Rochester Institutional Animal Care and Use Committee. The 3647 line of TNF-Tg mouse systemically overexpresses a single copy of the human TNF gene. The 3647 TNF-Tg was originally obtained from Dr. George Kollias (Institute of Immunology, Alexander Fleming Biomedical Sciences Research Center, Vari, Greece) (43, 44, 47). Littermate controls were used in all experiments. Mice were screened via microcomputed tomography (μCT) starting at 3 mo of age for severe lung disease, as determined by a threshold of >400 mm³ postexpiratory lung tissue volume. All animals were treated and analyzed using a variety of outcome measures as outlined in Supplemental Fig. 1.

All animals were randomized into cohorts (n = 6 each) comparing male and female, anti-TNF versus placebo treatment at time of recruitment based on the previously stated threshold of lung disease. All animals were paired at recruitment, with matching initial lung volumes placed in opposite treatment groups. Murine anti-human TNF mAb (CNTO12; Centocor R&D, Radnor, PA) or IgG1 isotype placebo control mAb (CNTO151; Centocor R&D) were administered at 10 mg/kg/wk via i.p. injection for 6-wk total, as previously reported (48, 49).

Grip strength assessment was performed similarly to previous methods (50). Grip strength was assessed at all longitudinal timepoints via a force meter attached to a mesh grate (Chatillon DFE II Force Gauge; Metek, Largo, FL). Mice were placed on the grate such that their paws could grip the wires while the investigator performed low-force tail pulling to elicit a gripping response. The force generated from the resistance to pull was recorded. The mean value of five repeated measures per mouse for each time point was used for statistical analysis. Mouse weight was obtained and recorded together with grip strength.

Three-dimensional (3-D) ultrasound imaging of the popliteal lymph nodes (PLN) was performed using the VisualSonics Vevo3100 (Fuijifilm, Toronto, ON, Canada). Each mouse was followed longitudinally, with imaging every 2 wk beginning at week 0 of treatment. B-mode images were collected for each animal at each time point. Stacked B-mode images were imported into Amira Software (Thermo Fisher Scientific, Hillsboro, OR), allowing for identification and segmentation of the lymph node from the surrounding tissue. All lymph node volumes were collected and calculated as previously described (49, 51–53).

Cross-sectional μCT imaging was performed on all mice prior to initiation of treatment and at completion of treatment (at weeks 0 and 6). μCT of the lungs was performed with the vivaCT 40 (SCANCO Medical, Bruttisellen, Switzerland) as previously described (54). Briefly, mice were anesthetized using 1.5–2% isoflurane (VetOne; Fluriso, Boise, ID) and placed in the prone position in the imaging tube. A blood-pressure transducer (BPS-BTA and SensorDAQ; Vernier, Beaverton, OR) was placed underneath the chest cavity to monitor the chest wall movements as a surrogate measure of the respiratory cycle. A custom, previously cited, nonpredictive gating method developed in LabVIEW (National Instruments, Austin, TX) was applied to selectively obtain images during the postexpiratory phase of the breathing cycle to minimize motion artifacts (54). The μCT imaging parameters were as follows: 1000 projections over 180°, 300-ms integration time, 45 kVp, and 176-mA current. DICOMs at 35 μm³ isotropic voxel resolution were exported to Amira for image analysis using semiautomated segmentation. All 3-D reconstructions and volume quantifications were performed in Amira Software (54, 55).

At termination, the left lobe of the lung was inflated with 10% neutral buffered formalin (NBF) using a 23-guage needle and 5-ml syringe through the trachea. The left lobe was subsequently incubated in 10% NBF for 3 d before histology processing. The lungs were processed and embedded in paraffin blocks by the University of Rochester Medical Center, Center for Musculoskeletal Research Histology Core Facility. Lungs underwent coronal sectioning from paraffin in three consecutive regions, each no less than 100 μm apart. Each region was ∼100-μm thick, each sectioned into a series of 5-μm slices. One representative slide from each region was stained with H&E. All slides were scanned with an Olympus VS120 (Olympus, Tokyo, Japan). A board-certified pathologist (M.J.V.) was blinded and graded each slide on a scale of 0–3 (none, mild, moderate, severe) for perivascular inflammation, interstitial inflammatory infiltrates, alveolar macrophages, and vascular luminal thickness. A total histology score for each slide was generated using the sum of the scores from each subcategory for a total scale range of 0–12. Values from each individual subject were averaged together for data analysis.

Knees collected at termination were fixed in 10% NBF for 3 d prior to histologic processing and paraffin embedding. Samples were sectioned on a sagittal plane in three distinct regions 50-μm apart. Each region had five 5 μm-thick sections. Representative slides from each region were similarly stained with H&E and scanned with an Olympus VS120. Quantification of the synovial area was performed manually on Visiopharm software (Visiopharm, Hoersholm, Demark).

B220 immunohistochemical staining was performed on all lungs using PE-conjugated B220 Ab (BioLegend, San Diego, CA) and NucBlue Live cell nuclear staining (Invitrogen, Eugene, OR). Semiautomated histomorphometry was performed to manually identify and gate B220⁺ follicle-like structures within the total lung area of each stained slide. Histomorphometric analysis was used to quantify both the total number of follicles per lung and the follicle area. All B220-staining histomorphometric analysis was performed using QuPath software (Queen’s University, Belfast, U.K.).

Lungs were extracted for the use of both flow cytometric analysis and histologic evaluation. Prior to inflation for histologic analysis, the right lobes were tied off using surgical sutures, and peripheral lung tissue was removed for processing into a single-cell suspension. The right lobes were taken and manually digested using a razor blade before being incubated with 1 ml 0.2 mg/ml Liberase (Sigma-Aldrich, St. Louis, MO) with 0.75 KU/ml DNase IV (Sigma-Aldrich) in RPMI 1640 buffer at 37°C for 30 min. Homogenized lungs were passed through a 70-μm nylon mesh filter to create a single-cell suspension and washed in 2% FBS plus 2.5 μM EDTA in PBS. ACK lysis buffer (recipe from Cold Spring Harbor) was used to lyse the remaining RBCs in the suspension. Cells were counted on a hemocytometer (Hausser Scientific, Horsham, PA) using trypan blue in a 1:1 dilution.

Cells were blocked with 2 μg/ml Fc block (Bio X Cell, West Lebanon, NH) for 10 min prior to labeling. One milligram per milliliter LIVE/DEAD Aqua (Invitrogen) was used as a viability stain. Cells were subsequently labeled with a set of fluorochrome-conjugated Abs for cell surface markers at various optimized concentrations based on previously published panels (56, 57). These included CD3-allophycocyanin-Cy7 (clone: 17A2; BioLegend, San Diego, CA), CD19-allophycocyanin (clone: 1D3/CD19; BioLegend), CD11b-PE (clone: M1/70; BD Biosciences, San Jose, CA), CD11c–eF450 (clone: F418; eBioscience, San Diego, CA), CD45–BV786 (clone 30-F11; BD Biosciences), MHC class II (MHCII)–FITC (clone M5/114.15.2; eBioscience), Siglec F-PE-CD594 (clone E50-2440; BD Biosciences), anti-F4/80–FITC (clone BM8; BioLegend), anti-CD88–allophycocyanin (clone 20/70; BioLegend), anti-CD64 BV605 (clone X54-5/7.1; BioLegend), anti-XCR1 BV650 (clone ZET; BioLegend), and anti-CD172a PE-Dazzle-594 (clone P84; BioLegend). A second panel examining the B cell populations followed the same protocol, using CD1d-allophycocyanin (clone: 1B1; BioLegend), CD3-allophycocyanin-Cy7 (clone: 17A2; BioLegend), CD19-PE-Cy5 (clone: 6D5; BioLegend), CD45R/B220-BV785 (clone: RA3-6B2; BioLegend), CD21/CD35-PE (clone: 7E9; BioLegend). CD23-FITC (clone: B3B4; BioLegend).

Data acquisition was performed on a BD LSR II 18-color flow cytometer using BD FACSDiva software (BD Biosciences). Data analysis and compensation calculations were performed using FlowJo software (TreeStar, Ashland, OR). Cell populations were evaluated based on a sequential gating strategy (see Fig. 1). The data were analyzed and presented as a percentage of live cells based on LIVE/DEAD staining.

All statistics were performed in Prism 7.00 (GraphPad Software, La Jolla, CA). Unless otherwise stated, all continuous variables were analyzed using a two-way ANOVA with Tukey post hoc on multiple comparisons to examine the effects of both sex (female versus male) and treatment (placebo versus anti-TNF).

Flow cytometric analysis was performed on the pulmonary tissue of both TNF-Tg mice and their wild-type (WT) littermates to identify the cellular populations responsible for the lung pathologic condition. Sequential gating (Fig. 1A–G) was performed to find differences in specific cell populations. The identified populations were labeled as outlined in Fig. 1H. These populations include alveolar macrophages, two subsets of cDCs (cDC1s and cDC2s), and monocytes at various stages of activation. For proportionate changes in the total cellular composition, representative contour plots for WT and TNF-Tg mice of the CD11b versus CD11c profile demonstrated gross differences in the myeloid populations in disease (Fig. 1I, 1J). More specifically, large increases were found in the CD11b⁺⁺ CD11c⁺⁺ (population 2) and CD11b⁺⁺ CD11c⁺ (population 3) cells (Fig. 1I, 1J). Positive staining with the cell surface markers MHCII and Siglec F found these cells were consistent with cDC2 cells and activated monocytes/macrophages (populations 2 and 3, respectively) (Fig. 1H). Additional characterization markers for dendritic cell subsets and macrophages confirmed these findings (Supplemental Figs. 2, 3).

Previous studies have monitored the effects of anti-TNF therapy on the systemic and articular pathologies in the TNF-Tg model (48, 49, 58, 59). For the 3647 line specifically, it was found that the maximum efficacy of treatment was attained at 6 wk of therapy (48, 49). To confirm the effectiveness of anti-TNF therapy on systemic effects, total body weight was monitored at each week of the treatment course beginning at week 0 just prior to initiation of treatment. Baseline weight at beginning of treatment was identical within sex between the cohorts (Fig. 2A). Total body weight continuously declined over time in the placebo-treated TNF-Tg females, consistent with disease progression. In contrast, the male TNF-Tg placebo-treated cohort maintained a consistent body weight, consistent with previous reports (50), whereas both anti-TNF–treated cohorts continuously gained weight over the 6-wk course (week 0 versus week 6, Males: p = 0.0025; Females: p < 0.0001, Fig. 2A).

Grip strength was monitored over time on a biweekly basis to monitor gross arthritic change in the smaller peripheral joints. Grip strength was identical within each sex between treatment groups at baseline but significantly increased in the anti-TNF–treated animals compared with placebo for both male and female cohorts over time (Fig. 2B). Mice treated with anti-TNF therapy demonstrated improved grip strength compared with baseline at points as early as 2 wk into treatment (Fig. 2B).

Histomorphometry performed on H&E sections of the knee joint revealed that total synovial area, a measure of active synovitis, was significantly reduced in anti-TNF–treated mice as compared with their placebo control counterparts (p < 0.0001 for both sexes) (Fig. 2C–G).

Assessment of total volume of the PLN is another noninvasive marker of arthritic activity in the TNF-Tg model (53, 59, 60). During the development of active inflammation in these mice, the draining lymph node of the affected joint expands in size because of the increase in lymphatic trafficking from the synovial space. We used ultrasound imaging and virtual 3-D reconstruction to quantify the total volume of the lymph node as a marker of disease progression and activity in the context of both animal models and clinical patients (53, 60, 61). Representative 3-D reconstructions of PLNs shown in Fig. 2H–K demonstrate a marked difference between placebo and anti-TNF mice by the end of treatment. Similar to the grip strength results, longitudinal quantification of the total PLN volumes demonstrated a marked initial decrease in PLN volume in the anti-TNF cohorts, with a significant difference in volume between treatment groups maintained over the course of treatment for both sexes (Fig. 2L). The improved arthritis assessments indicate the alleviation of symptoms and confirm efficacy of treatment with anti-TNF therapy in this model.

3-D μCT is a validated in vivo method for measuring total tissue volume of the lung (54). Increased lung tissue volumes were observed in TNF-Tg mice compared with WT littermates (54). To document the presence and severity of lung disease and the effects of anti-TNF therapy on pulmonary inflammation, μCT imaging was performed prior to treatment initiation and prior to termination. Representative 3-D reconstructions of female placebo-treated (Fig. 3A, 3C) and anti-TNF–treated lungs (Fig. 3B, 3D) at week 0 (Fig. 3A, 3B) and week 6 (Fig. 3C, 3D) revealed that anti-TNF–treated lungs demonstrated a markedly decreased tissue mass. Quantification of the lung tissue volume via thresholding on μCT Hounsfield units shows significant reduction in lung tissue volume and improvement in the anti-TNF–treated cohorts of both sexes at the end of treatment (Fig. 2E, anti-TNF week 0 versus week 6, female, 469.0 ± 28.6 versus 292.0 ± 31.7 mm³; male, 437.5 ± 33.1 versus 292.2 ± 38.8 mm³; mean ± SD. p < 0.0001).

It should be noted that μCT imaging is limited by resolution thresholds. Tissue volume changes for these scans could therefore be attributed to alterations in fluid accumulation, cellular mass, or collagen deposition. To qualitatively interrogate the change in composition, histology was performed on the lung tissue (Fig. 4A–G). H&E staining revealed an inflammatory pathologic condition with no signs of fibrotic change in the parenchymal tissue (data not shown). Four characteristic pathologic patterns were identified in the lung, including interstitial inflammatory infiltrate, vascular luminal thickening, perivascular inflammation, and the formation of follicle-like structures lacking a germinal center (representative examples in Fig. 4A), all consistent with previous reports (50, 62).

To demonstrate the level of inflammation and cellular accumulation present at 3 mo of age (Fig. 4B, 4E), we obtained representative H&E sections of both female and male age-matched TNF-Tg animals from a different experimental cohort. Following the 6-wk treatment course, histological analysis of placebo-treated and anti-TNF–treated lungs from both females and males confirmed the findings on μCT. Anti-TNF–treated lungs showed marked decreases in both the total cellular infiltrate and the previously noted pathologic patterns as compared with their placebo-treated counterparts (Fig. 4C versus Fig. 4D; Fig. 4F versus Fig. 4G). Clinical grading of representative slides from each animal for individual pathologic features confirmed the effects of anti-TNF therapy, with significant decreases in total histology scores for both female and male mice (Fig. 4H, placebo versus anti-TNF, females: 7.88 ± 1.88 versus 3.33 ± 1.20; males, 5.72 ± 1.24 versus 2.89 ± 1.29; mean ± SD. ****p < 0.0001). Notably, although total scores differed between females and males with placebo treatment (p < 0.0001), no significant difference was noted between the sexes with anti-TNF treatment (p = 0.7361). Subscoring for individual histologic patterns exposed to anti-TNF treatment showed a significant decline in interstitial inflammatory infiltrates (Fig. 4I) and alveolar macrophages (Fig. 4J). Vascular patterns were less drastically affected by treatment. Perivascular inflammation (Fig. 4K) in male cohorts and vascular luminal thickness in the female cohorts (Fig. 4L) each decreased significantly.

The interstitial inflammatory infiltrate, the perivascular inflammatory infiltrate, and the vascular luminal thickening appeared to be almost completely alleviated with treatment. The follicle-like structures alone remain in the treated lungs. However, these structures are attenuated, with decreased absolute numbers of these structures appearing in the anti-TNF– treated lungs (Fig. 5H). Importantly, for both calculated tissue volume measured by μCT and histologic evaluation, anti-TNF therapy ceases progression of the disease and treats the baseline pathologic condition.

Previous studies of the arthritic joint-draining lymph node from the laboratory implicated a specific B cell subset, Bins, as key effector cells that promote both structural and inflammatory changes in the draining lymph nodes of inflamed joints (45, 63). This unique population is a subpopulation of B cells defined by their CD21/35^hiCD23⁺ expression profile and were targeted to the sites of active articular disease activity. They are associated with the onset of inflammatory-erosive arthritis and reflect arthritic activity in the affected joint. Subsequent B cell depletion was unexpectedly found to alleviate symptoms, suggesting that they play an active role in disease pathogenesis (45, 63).

The majority but not all of the pathologic patterns noted in terminal disease were obliterated with anti-TNF therapy. Of the four noted patterns, only the follicle-like structures remained after treatment administration. Analogous to the lymphatic system, we wondered whether these follicles contained cellular compositions similar to the latent populations found in the draining lymph nodes. B220 (CD45R) immunohistochemistry staining uncovered the presence of B cells in the follicle-like structures in the pulmonary tissue prior to treatment initiation (3-mo-old representative samples) and in both the placebo and anti-TNF–treated cohorts (Fig. 5A–C). Histomorphometry analysis was performed on the lungs using semiautomated gating to identify and quantify B220⁺ follicle number and area for each lung (Fig. 5D, 5E). For anti-TNF–treated mice, total follicle area (as a percentage of total lung area), average follicle area, and the number of follicles per lung were decreased compared with placebo-treated counterparts (Fig. 5F–H).

Further investigations regarding the existence and possible role of the Bin population in pulmonary manifestations in the TNF-Tg model were performed. Flow cytometric analysis interrogating the phenotype of these B cell accumulations was performed (45, 64), defining follicular B cells (CD21/35^lowCD23⁺), marginal zone B cells (CD21/35^hiCD23^low), and Bins (CD21⁺ CD23^hi). Flow cytometry revealed a different B cell profile in the diseased-untreated TNF-Tg lung than in the one observed in the peripheral lymph node (Fig. 5I–L). Most notably, the pulmonary B cells displayed a decreased expression of CD23 (Fig. 5L). Therefore, these structures remain in the lung despite therapeutic intervention, as hypothesized, but the profile of these B cell populations differs from the profile that is present on B lymphocytes found in the lymphatic system.

Total cell count for the lungs was significantly increased in the placebo-treated cohort when compared with both WT controls and anti-TNF–treated mice (Fig. 6G). It should be noted that the cell counts in the anti-TNF–treated cohorts declined to the level of WT animals, with no significant difference found between these groups for either female or male mice.

To further assess the proportionate shifts in the pulmonary cell populations with treatment, flow cytometric analysis was performed using the earlier panel and gating scheme (Fig. 1A–H) that identified populations of interest. Representative contour plots of the myeloid profile (again using CD11b versus CD11c) comparing age-matched WT littermates, placebo-treated, and anti-TNF–treated lungs demonstrate differences in the populations of each cohort (Fig. 6A–F). Consistent with our earlier analysis, the placebo-treated TNF-Tg mice demonstrated disproportionate increases in the CD11b⁺⁺ CD11c⁺⁺ MHCII⁺ Siglec F⁻ (population 2, cDC2) and CD11b⁺⁺ CD11c⁺ MHCII^int Siglec F⁻ (population 3, activated monocyte/macrophage) cells. In comparison, the anti-TNF–treated mice had cellular profiles similar to the WT littermate counterparts.

Analysis of the populations as percentages of live CD45⁺ cells (Fig. 6H–O) and absolute numbers of cell subsets (Supplemental Table I) confirm that the most significant changes are within the cDC2 and activated monocyte/macrophage populations (Fig. 6L, 6M). These findings support our initial impressions, confirming that these two populations disproportionately changed in the TNF-Tg lung and returned to WT levels with anti-TNF therapy.

In this study, we find the following: 1) anti-TNF therapy both prevented progression and alleviated pre-existing pathologic conditions in pulmonary tissue, and 2) treatment of pathologic conditions involved the shifting of the cellular composition of the diseased lung to WT-like levels.

Six weeks of anti-TNF therapy decreased lung tissue volume with resolution of inflammatory cellular infiltration in TNF-Tg mice of both sexes, which was not observed in placebo-treated animals. The treatment decreased interstitial cellular accumulation, and the cell populations returned to WT levels. We identified two distinct myeloid populations in the lung that were disproportionately upregulated in disease and remain elevated following anti-TNF therapy. Flow cytometric analysis of the lung found that these two populations were consistently elevated in the pathologic state and were identified as activated monocytes/macrophages and cDC2 cells. B cell–based follicle-like structures were noted to be alleviated in both absolute number and size, yet they remained despite anti-TNF treatment. These B cells were found to be distinct from the Bin population, a population which was previously implicated in promoting and facilitating active articular disease in this mouse model. Future studies aimed to further evaluate the role of these immune cell populations will include depletion of the key immune cells. Taken together, these results suggest that the inflammatory stage of ILD is mediated by a limited number of key populations, and that the cellular component of RA-ILD is responsive to anti-TNF therapy.

RA-ILD is generally considered an irreversible pathologic condition, with variable amounts of fibrotic change, which is the component that is resistant to treatment. The previously discussed theory of the progression of RA-ILD through distinct stages of disease suggests that a prior inflammatory state may be necessary for further disease development and could be reversible with anti-inflammatory therapies.

That being said, it has previously been reported that anti-TNF therapy is ineffective and possibly detrimental in the context of RA-ILD (36–42). In the literature, the relationship between anti-TNF therapy and RA-ILD is complex and conflicting. Several articles report differences in ILD diagnosis or outcome with treatment (36, 38–41); however, others have found no associated risk with anti-TNF therapy (65–67). These studies clarify that pre-existing pulmonary impairment was strongly associated with the increased risk of adverse reactions, and anti-TNF treatment often exacerbated a previously diagnosed lung disease, although the inflammatory or fibrotic nature of the pathologic condition was not specified. When anti-TNF therapy accelerates disease progression, it is specifically in the context of fibrotic change in the lung (41, 68). One possible explanation for this phenomenon is TNF’s ability to inhibit TGF-β signaling (69–71), coupled with TGF-β’s role in promoting fibrotic disease (72). These findings led to the current guidelines stipulating the cessation of all anti-TNF therapies in RA patients diagnosed with ILD. This restriction severely limits the available treatment for the articular symptoms of RA.

One of the most notable results from this study was the reversal of pre-existing inflammation on top of the prevention of disease progression with the anti-TNF regimen. Given the two-stage hypothesis of RA-ILD development, the reversal of the inflammatory pulmonary pathologic condition in the TNF-Tg mice in this study could have broader implications on the current clinical guidelines regarding the risk of anti-TNF therapy in RA-ILD. The TNF-Tg model isolates an important inflammatory component of the RA-ILD phenotype. In the context of the theory of RA-ILD progression, our results regarding the reversal of inflammatory components of ILD implies a more nuanced interpretation of clinical data. Whereas anti-TNF therapy may be contraindicated for irreversible fibrotic changes in the lung, implementation of anti-TNF therapies at an earlier, purely inflammatory stage of disease may lead to complete alleviation of the pulmonary pathologic condition. Of course, this would require identifying patients with ILD prior to fibrosis onset, which has been difficult to date, and current screening guidelines have not sufficiently captured this population.

A second-line biological therapy for the articular disease in RA is rituximab, a chimeric anti-CD20 Ab that targets B lymphocytes for the treatment of RA (73). RA-ILD bronchoalveolar lavage fluid showed plasma cells, and this finding combined with the efficacy of B cell depletion for arthritis catalyzed several studies that evaluated rituximab in treating RA-ILD (74–76). Overall, positive responses to anti–B cell therapy were observed for both UIP and NSIP subtypes, although responses in patients with severe UIP were not as robust (75). The reasons for the responsiveness to anti–B cell therapy remain unclear. Our current findings suggest that the presence of these B cell accumulations in the lung tissue may reflect inducible bronchus-associated lymphoid tissue also found in other TNF-overexpressing mouse models (77). These tissues may be playing a role in either the onset of the inflammatory processes in the lung or the persistence of the disease state. However, further studies are necessary to fully elucidate the role of B cell accumulations in this mouse model.

Several limitations of this study are noted. First, whereas we have identified two target cell populations of interest from our myeloid panel, these characterizations are limited. Our flow panel was based on prior publications characterizing the cellular composition of pulmonary tissue to create a general profile of our TNF-Tg lungs. The original intention of the panels was to create a general, broad profile of the composition of the lung, which was also our goal for this study. However, to delve more deeply into the specific roles these cells play in mediating the pathogenesis of our model, more specific characterization is required. Future directions regarding the roles of these populations would ideally include longitudinal studies looking at the progression of disease over time as well as methods of accumulation (such as peripheral migration, proliferation of local populations, etc.).

The second issue regards the isolated nature of our model. The TNF-Tg mouse model expresses human TNF in a dysregulated manner, leading to the subsequent systemic inflammatory effects mimicking inflammatory arthritis. Anti-TNF therapy should therefore alleviate any acute inflammation, as shown in prior studies looking at joint inflammation in these mice (49, 58, 59). Given the pervasive nature of TNF in RA, it would follow that TNF would play an integral role in the predisposition toward pulmonary involvement as part of the systemic manifestations of disease. In fact, recent clinical data from Sparks et al. (78) found an association between RA disease activity and increased risk of RA-ILD development. However, the pulmonary pathology of the TNF-Tg model has not been expansively characterized in the context of the systemic extra-articular inflammation. Our laboratory has previously described the inflammatory nature of this model, with little-to-no fibrotic involvement of the pulmonary interstitium (50, 54). Our model provides a focused perspective on an isolated component of the complex manifestation of RA-ILD. Indeed, our results showed both cessation of disease progression as well as reversal of pre-existing infiltration in the pulmonary tissue (Figs. 2, 3).

This brings us to an interesting point with regards to our model and its possible implications regarding the development of fibrotic disease. Multiple publications have described the induction of pulmonary fibrosis in the surfactant protein-C/TNF-Tg (Sp-C/TNF) murine model, a similar yet more localized form of the TNF-Tg model in which TNF-α expression is controlled by the surfactant protein-C promoter, resulting in TNF overexpression in the pulmonary tissue (69, 77, 79–81). The Sp-C/TNF-Tg model is documented as having a lymphocytic infiltrate, which progresses to a variable fibrosing alveolitis with significant inflammation, diffuse interstitial collagen deposition, and emphysematous changes. This finding stands in stark contrast to our systemic TNF-Tg model, which exhibits a solely inflammatory phenotype with diffuse cellular accumulation in the interstitium (54, 62). Furthermore, a third model, with inducible TNF expression in the lung, develops emphysematous change with lymphoid follicles and no signs of augmented collagen deposition in the pulmonary tissue (77). Each model outlined above demonstrates significant and varied pathologic change on the pulmonary physiology, complicating our understanding of the effect of TNF in lung disease. The discordance of the phenotypes between the models raises important questions regarding the contribution of TNF in human RA-ILD and could provide clues to the transition between inflammatory and fibrotic disease.

One important point to note is that fibrosis is inherently a chronic and prolonged process, taking months to years to develop in clinical RA-ILD. Systemic TNF-Tg females have previously been found to have lifespans of 5.5 months. Compounded with the constitutive inhibition of TGF-β signaling by TNF (69–71), these factors provide the most likely explanation for the lack of fibrosis in the TNF-Tg model. Further studies regarding the role of TNF in RA-ILD development may benefit from a direct comparison of all the models’ local cytokine expression profiles. This approach could provide evidence toward the cause of the different terminal disease phenotypes.

In conclusion, we have shown that anti-TNF therapy alleviates pre-existing inflammatory pulmonary pathologic conditions and identified two distinctly upregulated myeloid populations of interest in the pulmonary cellular accumulation. Evaluation via μCT and histology both confirmed that anti-TNF therapy effectively reversed pre-existing inflammation in the lung, reducing cellular infiltration back to WT levels. Concurrent flow cytometric analysis confirmed the posttreatment lung composition as comparable to WT cellular profiles, indicating the removal of selective pathogenic cells. These included the disproportionately upregulated monocytes/macrophages and the cDC2-like populations, along with focal B cell accumulations in the lung parenchyma, which were all attenuated with anti-TNF treatment. Further studies demand the rigorous characterization of our populations of interest and further exploration of more nuanced clinical guidelines regarding anti-TNF therapies in the treatment of RA-ILD.

We thank Mike Thullen (Biomechanics, Biomaterials, and Multimodal Tissue Imaging Core, University of Rochester Medical Center) for technical expertise in performing the μCT scans. We would also like to acknowledge Kathy Maltby and Jeff Fox (Histology, Biochemistry and Molecular Imaging Core, University of Rochester Medical Center) for technical expertise with the histology performed in this study. The anti-TNF and placebo mAb used in this study were provided by Janssen Pharmaceuticals.

The authors have no financial conflicts of interest.