Storm warning: lung disease in systemic juvenile idiopathic arthritis

In few rheumatic diseases has the pace of therapeutic progress equaled that in systemic juvenile idiopathic arthritis (sJIA). Previously the childhood arthritis with the most dismal prognosis in terms of disability and even death, the identification of IL-1β and IL-6 as key disease mediators has led to therapeutic interventions of remarkable efficacy (1–4). Observational studies defining the utility of cytokine blockade early in disease have enabled treatment regimens that avoid corticosteroids altogether, with previously unheard-of levels of disability-free and even drug-free remission (5, 6). Yet while the landscape of sJIA was being thus transformed, clouds appeared on the horizon. In 2013, Kimura and colleagues observed that some patients with sJIA developed life-threatening pulmonary complications of a kind unfamiliar to practitioners who had treated this disease for decades (7). A case series available in abstract form identified 61 additional cases (8). In this issue of Arthritis & Rheumatology, Schulert, Grom and co-workers report yet more patients and provide the first detailed mechanistic look at sJIA lung disease (sJIA-LD) (Schulert et al.). The storm is upon us. What does it mean, and how should it affect our management of sJIA and its adult counterpart, adult-onset Still’s disease?

Let’s look at what we know. Kimura et al. described 25 patients with sJIA and pulmonary complications identified through an international pediatric rheumatology listserv (7). Clinical manifestations included pulmonary hypertension (64%), interstitial lung disease (28%), and alveolar proteinosis and/or lipoid pneumonia (20%), reflecting the accumulation of lipids and proteins in the alveolar spaces (more on this later). Mortality was an appalling 68%, often less a year from discovery of lung involvement. Compared with 389 sJIA patients from the Childhood Arthritis & Rheumatology Research Alliance (CARRA) Registry, patients with lung disease were distinguished by a history of macrophage activation syndrome (MAS) in 80% and clinical features including lymphadenopathy, hepatosplenomegaly, and clubbing (in 40%). Patients with sJIA-LD had also been exposed to a wider range of therapies, including biologics in at least 80% (principally IL-1 blockade but also tocilizumab, TNF blockade, and abatacept) but also more cyclosporine, cyclophosphamide, IVIG and steroid pulses. Despite intensive treatment, 92% had active systemic disease at the time that lung involvement was discovered. Patients with sJIA-LD were thus the “sickest of the sick.” While they might therefore have been expected to have a worse prognosis than other sJIA patients, the unfamiliarity of the lung manifestations raised concerns that new trends in treatment might be contributing to the problem.

The new report adds importantly to this initial description (Schulert et al.). Schulert et al. report 18 patients from Cincinnati Children’s Hospital Medical Center, long a referral center for sJIA and now increasingly for sJIA-LD as well. The clinical features of their patients reflect the heightened awareness of sJIA lung complications among pediatric rheumatologists: many had relatively mild symptoms (cough, tachypnea, dyspnea, or even no respiratory complaints at all), and almost 80% had clubbing. Their course was less severe than observed by Kimura; over a period of observation averaging 1 year, 14 of 18 patients showed stable or improved disease while 4 exhibited a deteriorating course; none died. Although 16/18 had parenchymal disease by computed tomography, pulmonary hypertension was confirmed in one patient only. Compared to a local control cohort of sJIA without known lung disease (n=36), patients were more likely to have developed sJIA before the age of 2 years (50% vs. 14%, odds ratio [OR] 6.5) and to have had a history of MAS (~80% vs. ~20%, OR 14.5), concordant with the predilection of early-onset sJIA patients for more severe disease and MAS (9). Intriguingly, almost half of patients had experienced an adverse reaction to a biologic, most characteristically an allergic-type reaction to tocilizumab, an otherwise quite uncommon event. These finding are echoed in the still-unpublished larger series, including 7 patients shared with Schulert and colleagues, although that series observed an extrapolated mortality of over 50% at 5 years as well as features such as lymphopenia and a striking overrepresentation of patients with trisomy 21 (8). Considering only referrals within the Cincinnati catchment area, Schulert et al. provide the first estimate of the prevalence of lung disease in sJIA patients: 5/74, or almost 7%, making sJIA-LD a complication that we should have noticed years ago if present at this frequency.

Schulert and colleagues also provide the first biological insight into sJIA-LD. Patients with sJIA-LD had elevated levels of IL-18 compared with other sJIA patients. Detailed evaluation of lung imaging showed involvement of pleura, bronchi/perivascular areas, and alveolar spaces, along with swelling of thoracic lymph nodes. Lung biopsy tissue was available from 8 patients, revealing alveolar filling with protein and lipid, a robust CD4+ T cell-predominant inflammatory infiltrate, fibrosis, and vasculopathy, including arterial wall thickening. Bacterial, viral and fungal pathogens were not observed. Overall the histopathology was typically in the “endogenous lipid pneumonia/pulmonary alveolar proteinosis (ELP/PAP) spectrum.”

ELP and PAP refer to the accumulation of lipid and/or proteinaceous fluid in the alveoli (“endogenous” indicates that the lipid is host-derived rather than from inhalation). ELP is otherwise seen with bronchial obstruction and is marked by the presence of cholesterol crystals. PAP reflects failure of alveolar macrophages to clear the alveoli, often due to disrupted GM-CSF signaling. The histology of sJIA-LD thus suggests macrophage dysfunction, although in sJIA-LD there is also inflammation, vasculopathy, and fibrosis not typical for ELP/PAP.

If this pathophysiologic clue is correct, how might macrophage dysfunction arise? Schulert and colleagues show that sJIA-LD patients do not have anti-GM-CSF antibodies or grossly defective GM-CSF signaling, ruling out usual PAP mechanisms. Instead, their data point in a different direction. Most patients with sJIA-LD have had MAS, and many have had it several times, as characteristic of sJIA patients with high levels of circulating IL-18 (10). Increasingly, interferon gamma (IFNγ) has become implicated in the biology of MAS, likely as part of a positive feedback loop whereby lymphocytes activate macrophages that in turn further activate lymphocytes. Challenging to measure directly, IFNγ activity can be assessed via “proxy” molecules that it induces, including CXCL9 and CXCL10. Studies now suggest that the explosive onset of MAS often reflects the appearance of IFNγ, otherwise not considered a major player in sJIA (11). IL-18 potentiates IFNγ generation in cells exposed to cytokines such as IL-2, IL-12, or IL-15 (12). The fact that most sJIA-LD patients have a history of MAS and high IL-18 suggests that they are particularly prone to IFNγ excess, and thus potentially to IFNγ-mediated activation of alveolar macrophages. Evidence for such activation is provided by the authors’ gene expression studies in sJIA-LD lung tissue, where both CXCL9 and CXCL10 were substantially upregulated, with corresponding proteins detected in 2 of 6 bronchoalveolar lavage samples. Intriguingly, overexpression of IFNγ and other Th1 cytokines by T lymphocytes can cause PAP-like disease in mice (13). There is even a report in the lay press of a patient with MAS and sJIA-LD who improved in response to treatment with the IFNγ antagonist emapalumab (https://systemicjia.org/lilys-fight-with-macrophage-activation-syndrome-mas-new-drug/). Together these data suggest, but do not yet prove, that IFNγ represents a key pathophysiological link in sJIA-LD.

Why are we seeing sJIA-LD now and not in years gone by? In some patients, lung disease might have remained hidden, either because children died of MAS or were so disabled that lung impairment remained occult. Yet even in the bad old days such cases were uncommon, and the distinctive erythematous clubbing seen in many sJIA-LD patients should still have been obvious. SJIA-LD really does look like a new problem, so it is important to consider whether it is of our own making. In particular, is the increased use of IL-1 and/or IL-6 blockade contributing to sJIA-LD? This is a difficult question, especially now that pediatric rheumatologists turn to biologic agents in many or even most sJIA patients, as first-line treatment or after the failure of NSAIDs, methotrexate, and corticosteroids. All children reported by Schulert and colleagues had received IL-1 blockade, and many also IL-6 blockade, because that is how we treat sJIA nowadays, especially in the most severely affected children. There is thus inescapable confounding by indication. Indeed, not all observations fit comfortably with a simple causal model. Some patients in the Kimura series never received biologics (7). Patients in the Schulert series who developed lung involvement on biologics often remained on treatment with stabilization or even improvement, especially if systemic disease could be controlled. As the authors note, sJIA-LD is not as common in parts of Europe where biologic use in sJIA is comparable, including the recent Utrecht series of 42 sJIA patients treated with first-line anakinra (recombinant IL-1 receptor antagonist), of whom only one developed pulmonary manifestations in the course of dying from refractory MAS (6). Finally, PAP is a complication not only of sJIA/AOSD but also of hematologic malignancies, especially of myeloid origin, without any exposure to IL-1 or IL-6 antagonists (14). Thus the relationship between biologic therapy and sJIA-LD remains to be established.

Despite these caveats, there remains undeniable cause for concern. In the only randomized controlled trial of anakinra in sJIA, Quartier, Pascual and colleagues found that patients treated with anakinra developed a type I IFN gene expression signature, even (especially?) if they did not respond to treatment (15). In another series, 6 of 13 sJIA patients receiving IL-1 blockade exhibited an IFN signature (16). IFNγ is a type II rather than type I interferon, and whole-blood mRNA expression data can reflect shifting cell populations in addition to changes in transcription by individual cells, yet these findings heighten the plausibility of a reciprocal relationship between IL-1 and IFN (17). Whether IL-1 blockade augments IFNγ signaling, and how IL-6 blockade could act similarly, remains to be clarified. Cytokine antagonism could also contribute to lung disease in other ways. For example, it could predispose to an infection that triggers sJIA-LD, although to date no organism has emerged. In mice, IL-1α promotes maturation of alveolar macrophages, and deficiency of this cytokine (but not of IL-1β) results in PAP-like failure to clear the alveoli (18). Alternately, the “success” of biologic treatment could render patients susceptible to sJIA-LD by reducing exposure to steroids and/or methotrexate. We simply do not know.

The senior author of the present report, Dr. Alexei Grom, discussed another possibility in a recent lecture about sJIA-LD. For decades, amyloidosis was a scourge of sJIA patients in Europe but curiously remained rare in North America, and it faded away on the Continent even before the advent of biologics. It is possible that sJIA-LD is just such a secular trend. Yet enough data have accumulated that we must not assume so, but rather seek with urgency to understand this dark cloud over the otherwise dramatically successful advent of biologic treatment in sJIA. At present, my own practice in new-onset sJIA remains first-line biologic therapy, typically with IL-1 blockade. But I now have new doubts, especially for children with severe sJIA presenting early in life, accompanied by MAS, or with trisomy 21. The storm is here, and navigating sJIA treatment has become even more challenging than before.