Airway abnormalities are the most prevalent HRCT findings in ueRA

Demographic and clinical data are summarized in Table 1. The median age of the RA patients was 58 years, 83% were female, and the median patient-reported symptom duration was 6 months. Four patients reported history of lung disease (all cases being asthma). Among the RA patients, 60% were never smokers and 20% were currently smoking. The mean disease activity was moderate by DAS28 (4.9) but high by CDAI (23.6) composite scores. Seventy-three percent of the patients were positive for RF (IgM or IgA), 77% were positive for ACPA, and 53% of the patients were double-positive.

The frequency of having any RA-associated PA by HRCT was 60% (18 out of 30 patients) (Table 2). Airway abnormality was the most prevalent HRCT finding in this ueRA cohort (53%). Two patients had bronchiectasis and 14 patients (47%) showed signs of bronchiolitis (bronchial wall thickening and/or air trapping). Parenchymal nodules were found in 13 patients (43%). The mean number of nodules was 1 (range 0–9), and the mean total nodule volume per patient was 48 (range 26–480) mm3. Radiological signs of ILA were uncommon (10%), and all these patients had mild findings of reticulations. No patients showed radiological signs of pleural involvement. Representative images of the findings are included in Additional file 1, figure S3.

Table 2 RA-associated pulmonary abnormalities by HRCT at diagnosisPulmonary abnormalities are associated with a peripheral activated neutrophil phenotype

It is unknown whether peripheral neutrophil phenotypes are related to pulmonary pathology in patients with ueRA. First, unsupervised PCA was used to make unbiased cluster analysis of collected data including; presence of any PA by HRCT, neutrophil activation status (CD11b and CD62L expression), neutrophil subtypes (% LDG and CD177 expression) and neutrophil-derived proteolytic enzyme (calprotectin), as well as disease activity measures and demographic data (Fig. 1). Any PA was clustered together with high expression of CD11b (indicating activation) on NDGs, high frequency of LDGs and calprotectin indicating positive association between these variables. This cluster also included RF IgM levels and the clinical parameters of inflammation (CRP, ESR, number of swollen joints (SJC)) and the PhGA. Importantly, history of lung disease or symptom duration did not associate with the PA cluster suggesting that the PA of this study population and neutrophil activation are early events in RA. ACPA levels and current smoker were not clustered with Any PA in this analysis.

Fig. 1

Principal component analysis (PCA) plot showing an overview of the relationship between the presence of PA by HRCT (Any pulmonary abnormalities), neutrophil activation status (CD11b and CD62L expression) and subtypes (% LDG and CD177 expression) on LDGs and NDGs, calprotectin, as well as disease activity measures and demographic data in ueRA patients(n = 30). Closely clustered variables to Any pulmonary abnormalities (black circle) indicate positive association (grey area), while diagonally opposite variables indicate inverse relationship. Flow cytometry variables are represented as MFI

Next, an OPLS-DA analysis was performed to identify the variables discriminating the group Any PA from the group No PA. Any PA demonstrated a positive association with RF IgM, high expression of CD11b on NDGs, and CRP, and negative association with high expression of CD62L on NDGs (Fig. 2A). Subsequent univariate analysis confirmed that RA patients with Any PA had significantly increased CD11b expression on NDGs (MFI 6455 vs 5317, p = 0.014) and significantly lower expression of CD62L on NDGs (MFI 10,065 vs 12,676, p = 0.012), indicating neutrophil activation, as compared to No PA (Fig. 2B, C). The LDG proportions were higher in Any PA compared to No PA, but the proportions were low in both groups (< 4%) except for one outlier in Any PA (Additional file 1, figure S4A).

Fig. 2

OPLS-DA column loading plot (VIP > 0.95) demonstrating the difference between the two patient groups Any pulmonary abnormalities (Any PA) and No pulmonary abnormalities (No PA) with respect to neutrophil activation status (CD11b and CD62L expression) and subtypes (% LDG and CD177 expression) on LDGs and NDGs, calprotectin, as well as disease activity measures and demographic data (A). Univariate analysis of NDG CD11b (B), NDG CD62L (C), RF (D), ACPA (E), CRP (F) or ESR (G) in patients with Any PA compared to No PA. Open circles in E–F indicate current smokers. Bars show median. ns, non-significant, *p ≤ 0.05, **p ≤ 0.01 (Mann–Whitney U test)

Pulmonary abnormalities are associated with levels of RF but not levels of ACPA

A significant increase in levels of RF (75.4 vs 8.6, p = 0.0025), but not ACPA (336.1 vs 162.5 p = 0.17), was found in patients with Any PA as compared to No PA (Fig. 2D, E). When excluding current smokers (open circles in Fig. 2D, E) in the analyses, RF levels were still higher but not ACPA (p = 0.034 and p = 0.62), which indicates that smoking status could not explain the differences in RF levels. In addition, patients with Any PA had higher levels of CRP (11.5 vs 5.0, p = 0.0035) compared to No PA patients, while there was no significant difference in ESR between groups (Fig. 2F, G). The Any PA and No PA groups did not differ in disease activity measures CDAI, SDAI, SCJ28, TCJ28 and DAS28, nor with respect to smoking status: never, former or current (Additional file 2, table S2). Furthermore, there were no significant differences in calprotectin serum levels in patients with versus without PA nor in patients with versus without joint erosions (Additional file 1, figure S4 B-C). Thus, presence of PA in this study population was associated with CRP, high RF IgM titres and neutrophil activation.

The fact that both RF IgM titres and markers of neutrophil activation in this ueRA cohort were associated with the presence of PA prompted us to investigate if there is a correlation between these two factors. Indeed, there was a significant positive correlation of RF IgM titres and CD11b expression on NDGs (r = 0.37, p = 0.05) (Fig. 3A). However, no significant correlation was found using loss of CD62L as a marker of activation (r =  − 0.19, p = 0.33) (Fig. 3B). No correlation was found between levels of ACPA IgG and mean fluorescence intensity of CD11b nor CD62L on NDGs (Additional file 1, figure S5).

Fig. 3

Correlation analysis between RF IgM titres and neutrophil activation status; NDG CD11b (A) and NDG CD62L (B) expression in RA patients (n = 30). ns, non-significant, *p ≤ 0.05 (Spearman’s rank correlation test). A linear regression line is presented in the respective plot

Airway abnormality is the main subtype behind the association of PA, neutrophil activation and RF titres

The overlap between the radiographic entities is illustrated by a Venn diagram in Fig. 4A. The vast majority (15/18) of patients with PA had airway abnormalities. Three patients with pulmonary nodules did not have any other PA. All patients with ILA also had airway abnormalities by HRCT. The small size of the ILA group precluded any further analysis with respect to this radiographic finding in this study.

Fig. 4

Venn diagram showing the overlap of the subtypes of PA (A). OPLS-DA score plot showing the separation of patients belonging to each PA subtype group; Airway abnormalities (including patients with coexisting nodules and/or ILA), Pulmonary nodules only (excluding patients with coexisting airway abnormalities) and No PA in the RA patients (n=30) (B). Closely clustered patients indicate similarities in neutrophil activation status (CD11b and CD62L expression), neutrophil subtypes (% LDG and CD177 expression), calprotectin, as well as disease activity measures and demographic data

To further elucidate which specific radiographic finding is the major driver for the association of Any PA with neutrophil activation, we performed an OPLS-DA analysis in which the three non-overlapping sub-groups; Airway abnormalities (including patients with coexisting nodules and/or ILA), Pulmonary nodules only (excluding patients with coexisting airway abnormalities) and No PA. The OPLS-DA score-plot demonstrated a separation between patients with Airway abnormalities and patients with No PA, indicating that these groups are different form each other with respect to neutrophil activation status, neutrophil subtypes, calprotectin, disease activity and demographic data. On the other hand, patients with Pulmonary nodules only clustered together with No PA indicating similarities between the groups (Fig. 4B).

Discriminant analysis of the two groups Airway abnormalities and No PA using OPLS revealed that Airway abnormalities was associated with the expression of CD11b on NDGs and RF IgM, and No PA with the expression of CD62L on NDGs, i.e. less neutrophil activation (Fig. 5A) (good quality (R2X > 0.4) and goodness of fit (R2Y > 0.5) of the model). Subsequent univariate analysis demonstrated a significant increase in CD11b (1.3-fold, p = 0.014) (Fig. 5B) and decrease in CD62L (0.6-fold, p = 0.003) (Fig. 5C) expression on NDGs in patients with airway abnormalities as compared to no pulmonary abnormalities. Furthermore, the group Airway abnormalities had significantly higher levels of RF IgM (8.8-fold, p = 0.0002) (Fig. 5D), but not ACPA, titres (p = 0.16) (Fig. 5E) compared to No PA. Again, excluding current smokers (open circles in Fig. 5D, E) from the analyses indicated that smoking status could not explain the differences in RF titres (RF p = 0.004 and ACPA p = 0.65). As in the non-stratified analysis, CRP levels were significantly higher (2.8-fold, p = 0.013), but not ESR (p = 0.14) (Fig. 5F, G), in Airway abnormalities as compared to No PA. Taken together, the stratified analysis indicates that airway abnormality is the major subtype behind the association of PA with high levels of RF IgM and neutrophil activation.

Fig. 5

OPLS-DA column loading plot (VIP > 1.0) demonstrating the difference between the two patient groups Airway abnormalities (including patients with coexisting nodules and/or ILA) and No PA, with respect to neutrophil activation status (CD11b and CD62L expression), neutrophil subtypes (% LDG and CD177 expression), calprotectin, as well as disease activity measures and demographic data (A). Subsequent univariate analysis of airway abnormalities and CD11b (B), CD62L (C), RF (D), ACPA (E), CRP (F) and ESR (G). Bars show median. ns, non-significant, *p ≤ 0.05, **p ≤ 0.01, *** p ≤ 0.001 (Mann–Whitney U test)