The joint involvement in adult onset Still's disease is characterised by a peculiar magnetic resonance imaging and a specific transcriptomic profile

Adult onset Still's disease (AOSD) is a rare systemic autoinflammatory disease, characterised by fever, arthritis, and skin rash, and joint involvement is one of its clinical manifestations. The aims of this work were to assess joint involvement, to describe main patterns of involvement, and associated clinical characteristics. In this work, we aimed at assessing the joint involvement in AOSD by using MRI, to describe main patterns and associated clinical characteristics. In addition, we aimed at assessing the global transcriptomic profile of synovial tissues in AOSD to elucidate possible pathogenic pathways involved. We also evaluated the global transcriptomic profile of synovial tissues to elucidate possible pathogenic pathways involved in the disease. Thus, AOSD patients, who underwent to MRI exam on joints, were assessed to describe patterns of joint involvement and associated clinical characteristics. Some synovial tissues were collected for RNA-sequencing purposes. The most common MRI finding was the presence of synovitis on 60.5%, mainly in peripheral affected joints, with low to intermediate signal intensity on T1-weighted images and intermediate to high signal intensity on T2-fat-saturated weighted and STIR images. Bone oedema and MRI-bone erosions were reported on 34.9% and 25.6% MRI exams, respectively. Patients with MRI-bone erosions showed a higher prevalence of splenomegaly, a more frequent chronic disease course, lower levels of erythrocyte sedimentation rate, and ferritin. In AOSD synovial tissues, a hyper-expression of interleukin (IL)-1, IL-6, and TNF pathways was shown together with ferritin genes. In conclusion, in AOSD patients, the most common MRI-finding was the presence of synovitis, characterised by intermediate to high signal intensity on T2-fat-saturated weighted and STIR images. MRI-bone erosions and bone oedema were also observed. In AOSD synovial tissues, IL-1, IL-6, and TNF pathways together with ferritin genes resulted to be hyper-expressed.

MRI findings in AOSD. The present evaluation included 44 MRI exams (13 exams of hand/wrists, 12 of knee, 5 of pelvis, 3 of shoulder, 3 of ankle, 3 of feet, 2 of hip, 2 of lumbosacral spine and 1 of cervical spine). The most common MRI finding was the presence of synovitis on 26 (60.5%) exams, mainly in peripheral affected joints, as shown in Fig. 1. In these exams, MRI revealed a mild to moderate proliferative synovitis, as thickening of the synovial membrane, suggesting the presence of a hyperplastic than of a hypertrophied synovial tissue. In these patients, we observed a low to intermediate signal intensity on T1-weighted images and an intermediate to high signal intensity on T2-fat saturated weighted and STIR images. Contrast enhancement (CE)-MRI was performed in our series in 20 patients with severe synovial involvement, allowing to better differentiate active hyperplastic synovitis than fluid effusion. CE-MRI confirmed the findings on T2-fat saturated weighted and STIR images without CE, as shown in Supplementary Figure 1. The presence of bone oedema, characterised by high signal intensity on T2-fat saturated weighted and STIR images, was reported on 15 (34.9%) MRI exams. In 11 (25.6%) patients, MRI-bone erosions were observed, codified as on T1-weigheted images as discontinuity of signal void of cortical bone and loss of normal high signal intensity of bone marrow fat. In all patients but one, MRI-bone erosions were synchronous with bone oedema, overlapping completely the locations, as reported in Fig. 1. Less frequently, tenosynovitis, defined as peritendinous effusion, and a focal joint space narrowing were reported on 8 (18.6%) and 7 (16.3%) patients MRI exams, respectively. Concerning MRI on spine, although the low number, no inflammatory lesions were detected. No ankylosis was observed in any MRI, even if this technique could not be the first choice in assessing this feature.
Follow-up study. During the follow-up, in 20 patients, 20 MRI exams were repeated (6 of hand/wrist, 5 of knee, 4 of pelvis, 2 of ankle, 1 of hip, 1 of cervical spine and 1of lumbar spine), and evaluated for possible changes, without any inferential analysis due to the lack of power. Following the therapies, an improvement of Histological findings. During the follow-up, in two patients, replacements of hip were performed and the synovial tissues were collected. Based on the peculiar MRI signal intensity of synovitis, we conducted a pathogenic study on these synovial samples. In this evaluation, we performed a histological study to evaluate the pro-inflammatory infiltrate and the presence of some cytokines involved in AOSD pathogenesis. Two patients (female 25 years-old, male 54 years-old, respectively) underwent joint surgery since a repeated and uncontrolled inflammatory involvement of hips because of the disease. They reported a long history of AOSD, treated with glucocorticoids, 0.2-0.3 mg/kg/day of prednisone equivalent, and non-steroidal anti-inflammatory drugs, ondemand, during the acute flares. In these patients, at the time of synovial tissues collection, no biologic DMARD was administered. www.nature.com/scientificreports/ As shown in Supplementary Figure 2, a moderate perivascular mononuclear infiltrate in the sub-lining stroma of hip synovial tissues was observed, associated with a slight increase in the cellular density of the stroma and a synovial lining of 2-3 layers. These features could suggest a low-grade synovitis. Furthermore, an increase number of vessels number could be suggested, but without a haematic extravasation. In addition, interleukin (IL)-1β, IL-6, TNF, and heavy ferritin subunit (FeH) were found on AOSD synovial tissues, as shown in Supplementary

RNA sequencing analysis.
After the histological assessment, we performed in the same patients a bulky RNA-sequencing to provide a transcriptomic synovial tissue profile of main pathogenic pathways involved in AOSD. This, RNA-sequencing analysis assessed the global transcriptomic profile of synovial tissues on AOSD patients and gender matched-controls (female 70 years-old, male 65 years-old, respectively). The comparison of the gene expressions profile between AOSD patients and controls revealed 1123 genes downregulated (log2FC ≤ − 1 and FDR < 0.05) and 2716 genes upregulated (log2FC ≥ 1 and FDR < 0.05) in AOSD synovial tissues (Fig. 3). Among upregulated ones, plxnc1, clstn3, c3, gbp1, irf1, gch1, cyr61, ccl8, and chi3l2 showed the highest differences between the 2 groups (Fig. 3). After that, we explored genes implicated in IL-1, IL-6, TNF, IFN-γ, and iron uptake and transport pathways. Assessing IL-1 pathway, we found an increased expression of il1a, il1b, il1rap, il1r1, il18r1, and il18bp on AOSD tissues when compared with controls. On the contrary, il18 resulted to be more expressed in the controls (Fig. 4A). In IL-6 pathway, we found an increased expression of il6 and il6st/ gp130 on AOSD synovial tissues, whereas an increased expression of il6r was shown in the controls (Fig. 4B). Among genes involved in TNF pathway, tnf, traf1, traf2, tnfaip3, and tnfrsf1a resulted to be more expressed in AOSD synovial tissues than controls (Fig. 5A). Furthermore, no significant difference was retrieved analysing the interferon (IFN)-γ pathway (Supplementary Figure 5). Finally, fth1 and ftl were more expressed in AOSD patients than controls (Fig. 5B), when we explored the iron uptake and transport pathway.

Discussion
The present work showed MRI patterns of joint involvement in AOSD, mainly reporting the presence of synovitis, characterised by low to intermediate signal intensity on T1-weighted images and intermediate to high signal intensity on T2-fat saturated weighted and STIR images. MRI-bone erosions and bone oedema in peripheral joints were also observed, which overlapped the locations. Assessing these findings with clinical features, the  www.nature.com/scientificreports/ The most common MRI finding was the presence of synovitis in peripheral joints, including wrists, shoulders, and knees, which showed a low to intermediate signal intensity on T1-weighted images and an intermediate to high signal intensity on T2-fat saturated weighted and STIR images. The latter feature appears to differentiate AOSD synovitis from haemorrhagic or post-traumatic ones, which are typically characterised by low or high intensity on T2-fat saturated weighted and STIR images, respectively 14,15 . This feature could be associated with the presence of FeH on AOSD synovial tissues, which could represent a specific marker of the disease, as observed in other affected tissues 16,17 . Despite the lack of power, the results obtained from the follow-up, suggested that MRI-synovitis could also follow the activity of the disease, reducing or enhancing in case of remission or flare. Thus, MRI-synovitis could be considered a possible surrogate of disease activity in AOSD to be assessed during the disease course, at least regarding the joint involvement.
In addition, a percentage of patients showed the presence of MRI-bone erosions and bone oedema. These findings seemed to be synchronous between them, overlapping the locations, and it could suggesting another typical feature of AOSD joint involvement. Differently from what previously reported [6][7][8][9] , no MRI findings to be correlated with ankylosis were observed on MRI-exams. In fact, although considered a typical feature of AOSD, the pericapitate ankylosis on conventional radiograph may be considered a finding of a long-standing and insufficiently treated disease. Thus, our MRI exams could have not revealed this aspect since the characteristics of assessed patients, with a relatively low disease duration and all variably treated with glucocorticoids (GCs), synthetic-and/or biologic DMARDs.
Analysing comparatively clinical characteristics and MRI-findings, the results suggested some clinical features related to MRI-bone erosions. In fact, these patients with MRI-bone erosions were characterised by a higher prevalence of splenomegaly, lower levels of both ESR and ferritin, and experienced more commonly a chronic disease pattern. In this context, it has been reported that splenomegaly and chronic disease pattern could predict the need of prolonged therapies 18 , differently from what observed in patients achieving a monocyclic pattern 19 . In fact, these patients were more frequently treated with synthetic-and biologic-DMARDs, which are administered in non-responder patients to first-line therapeutic strategies 20,21 .
However, analysing predictive factors of MRI-bone erosions, the use of synthetic-DMARDs was predictive of this finding, whereas the use of biologic-DMARDs did not. In fact, During AOSD, the administration of biologic-DMARDs is related to additional prescribing motivations to joint involvement in AOSD, considering the heterogeneous clinical course of these patients the disease 22 . The association among MRI-bone erosions and www.nature.com/scientificreports/ lower levels of both ESR and ferritin could reflect a more intensive therapeutic strategy in these patients, paralleling what observed in rheumatoid arthritis as a confounding by indication bias 23,24 . Another possible explanation could be provided by a recent published multidimensional characterisation of the disease AOSD patients 25 . In this work, four different clusters of AOSD patients were identified by using values of systemic score, ferritin, CRP, ESR, and age. Patients codified in cluster 4 were characterised by lower levels of ferritin and ESR than others. Thus, it could be possible that patients with MRI-erosions in the present evaluation could be codified in this cluster 4. These observations may suggest the need of further studies to fully validate the proposed clinical clusters of AOSD and to evaluate associated clinical characteristics. Interestingly, the global transcriptomic profile of AOSD synovial tissues suggested the hyper-expression of IL-1α, IL-1β, IL-6, and TNF as well as of their respective receptors, furtherly reinforcing the rationale of inhibiting these cytokines during the disease 26 . Thus, it could be possible to speculate the existence of a specific pro-inflammatory signature on AOSD joint involvement, characterised by the hyper-activation of these pathways. In addition, the hyper-expression of ferritin genes on AOSD synovial tissues could highlight the idea of a possible pathogenic role of ferritin in enhancing the pro-inflammatory burden, as proposed by the concept of the "hyper-ferritinaemic syndrome" [27][28][29] .
Finally, AOSD is burdened by life-threatening complications, such as MAS mainly MAS, a hyper-inflammatory syndrome which leads to a high mortality rate, although the disease usually affects young people [30][31][32] . Interestingly, in our cohort, no episode of MAS was registered. In our study, the patients in our cohort did not appear to be characterised by a hyper-activation of IFN-γ pathway, which is considered of crucial importance on the occurrence of life-threatening complications during AOSD and its juvenile counterpart [33][34][35][36] . In fact, the activation of IFN-γ would prompt the occurrence of a systemic inflammation, since it has been implicated as a pivotal mediator of the pathogenesis in different hyper-inflammatory diseases 35,36 . Similarly, although elevated levels of IL-18 are were reported in AOSD 37 , this cytokine appeared downregulated in synovial tissues of our AOSD patients. This possible discrepancy could be related to the pathogenic role of IL-18 in development of both MAS and cytokine storm syndrome 38,39 . Thus, it could be possible that IL-18 could be also more involved in the systemic manifestations of the disease than the articular ones. In addition, considering that synovial tissues were collected in patients with a long history of AOSD, the assessed cells could be in a state in which the production of www.nature.com/scientificreports/ IL-18 could be reduced due to the drugs and/or the exhaustion of pro-inflammatory cells producing cytokines. Taking together these observations, further studies are needed to fully evaluate these issues. The present evaluation is affected by different limitations due to the observational study design and the lack of control group, reducing the validity of our findings and suggesting a cautious generalization of the results. Furthermore, the relative low number of assessed MRI-exams would recommend the necessity of specific designed studies to entirely clarify this issue. However, the rarity of AOSD makes very difficult to arrange prospective studies, suggesting the significance of cohort studies to identify relevant clinical features and to generate hypotheses to be subsequently investigated.
In conclusion, in this work, MRI patterns of AOSD joint involvement were reported suggesting possible peculiar features on these patients. The most common finding was the presence of synovitis, characterised by an intermediate to high signal intensity on T2-fat saturated weighted and STIR images. MRI-bone erosions and bone oedema in peripheral joints were also observed, which could be synchronous between them, overlapping the locations. Assessing these findings with clinical features, the subset of patients showing MRI-bone erosions was characterised by a chronic disease course, lower levels of both ESR and ferritin. In addition, a hyper-activation was observed on of IL-1, IL-6, TNF, and ferritin pathways were observed by a global transcriptomic profile of AOSD synovial tissues, suggesting pathogenic mechanisms and therapeutic targets. Taking together all these findings, although further studies are needed to entirely elucidate these issues, some specific aspects of AOSD could be observed on joint involvement, suggesting typical radiologic findings, associated clinical features, involved pathogenic mechanisms, and therapeutic targets. Settings. Patients with AOSD, who attended the Rheumatologic Clinics of L' Aquila, Siena, Naples, and Bologna, were considered. These Italian Rheumatology centres, involved in this project, were selected due to previous expertise on management of AOSD and in inception cohort studies.

Methods
Patients. Patients were considered whether fulfilling Yamaguchi's criteria 40 for diagnosis of AOSD and undergoing at least one MRI exam on joints during the follow-up, for a clinical suspicion of joint involvement by the disease. Since the observational nature of the study, the choice of performing or not the MRI of joint was left to the physicians in charge on the patient. After the execution of this radiologic exam, the data of these patients were registered.
Variables on MRI exams. MRI exam was included if performed as detailed in available imaging protocols in musculoskeletal system 41 . The strict adherence to these standardised protocols has been deemed as crucial to minimise the variability among different diagnostic centres and readers, involved in the present study. In cases with severe synovitis, CE-MRI was employed to a better characterisation of the findings, by a bolus of gadolinium (0.1 mmol/kg). MRI-findings, namely bone erosions, osteitis/bone marrow oedema, synovitis, joint space narrowing, and tenosynovitis, were scored, by experienced radiologists, according to OMERACT MRI definitions 42 . Clinical variables. In this work, we related clinical characteristics and predictive factors of joint involvement in AOSD. These findings were recorded at the time of MRI execution. Clinical features, systemic score, occurrence life-threatening complications, therapies, and patterns of disease were reported. The systemic score was derived, as previously detailed 6,43 . The presence of fever, arthritis, arthralgia, complications, and other clinical manifestations were also evaluated and defined as discretely distributed variables (yes/no). The occurrence of MAS and other AOSD-life threatening complications were evaluated as previously reported 32,44 . Patients in remission were defined as those achieving a complete disappearance of any clinical and laboratory feature of the disease as previously reported 43 . Inflammatory markers, ESR and CRP, and ferritin were registered, and expressed as continuous variables. At the end of follow-up, patients were codified into three different disease courses, monocyclic, polycyclic, and chronic patterns as previously performed 6,32,[43][44][45][46] . The administration of therapies for managing AOSD was also recorded, at the diagnosis and during the follow-up, GCs, synthetic-and biologic-DMARDs. The use of GCs was stratified into two categories, low dosage (≤ 0.5 mg/kg/day of prednisone-equivalent) and high dosage (> 0.5 mg/kg/day of prednisone-equivalent), as already done 43  www.nature.com/scientificreports/ Bias. Considering the observational design, our study could be subjected to a number of possible biases, which we tried to overcome by a relatively simple study design and a careful selection of involved centres, all to be considered tertiary referral centres for AOSD and with a high experience in inception cohort studies. We also tried to minimise the main methodological issues by a careful definition of each variable to be assessed. Furthermore, participants with significant missing data, which were considered to be meaningful for the analyses, were removed.
Study size. No sample size estimation was provided, since our evaluation would provide a "real-life" estimation of the joint patterns of involvement in AOSD. From January 1, 2015 to December 31, 2019, consecutive patients with AOSD, admitted to Italian Rheumatology Units, who underwent to at least 1 MRI exam on the joint, were assessed and followed-up.
Hematoxylin and eosin. During the follow-up, in two patients, replacements of hip were performed and the synovial tissues were collected for histological purposes. Paraffin-embedded hip synovial tissues were processed for haematoxylin and eosin (H&E) staining. Samples were analysed at a single cutting level according to standard of practice. Synovial sections were analysed according to a semi-quantitative assessment of synovitis according to a previously validated score 47 . Images were acquired with Olympus BX53 microscope.
The sections were dewaxed, and antigen retrieval was carried out using Dako Target  Statistical methods. Statistics firstly provided the descriptive analysis of all assessed variables, in included patients. The data distribution was tested by using Shapiro-Wilk test and, thus, collected results were presented as mean and standard deviation (SD) or median and interquartile range (IQR). To compare the clinical characteristics of patients with-and without MRI-erosions parametric or non-parametric t tests were used for all the continuous variables and Chi-squared test was used for the categorical variables, as appropriate. Furthermore, possible correlations among the presence of MRI-bone erosions and ESR, CRP, ferritin, and systemic score were estimated by using a point-biserial coefficient correlation. In addition, a regression analysis was built to evaluate possible predictive factors, assessing the significance of selected clinical and laboratory features, on the likelihood of MRI-erosions presence. The statistical analysis of this feature was planned considering that the variable was discretely distributed ("yes MRI-bone erosions"/"no MRI-bone erosions"). The selection process of covariates started by a univariate analysis of each variable, any variable having a significant univariate test was selected as candidate for the multivariate analysis, whereas removed if non-significant. At the end of this process, the multivariate model was provided, with OR estimations of independently predictors of MRI-bone erosions. Due to the relatively simple study design, a very few missing data were managed by exclusion of these from analyses. www.nature.com/scientificreports/ ber 0122353/17) approved the study, which was performed according to the Good Clinical Practice guidelines and the Declaration of Helsinki. Informed consent was obtained from each patient for the use of de-identified clinical and laboratory data and images for study purposes.