Immune check point inhibitor-induced inflammatory arthritis as a model of autoimmune arthritis

Summary

The development of inflammatory arthritis in patients receiving immune checkpoint inhibitor therapy is increasingly recognized due to the growing use of these drugs for the treatment of cancer. This represents an important opportunity not only to define the mechanisms responsible for the development of this immune-related adverse event to ultimately predict or prevent its development, but also to provide a unique window into early events in the development of inflammatory arthritis. Knowledge gained through the study of this patient population, for which the inciting event is known, could shed light into the pathogenesis of autoimmune arthritis. This review will highlight the clinical and immunologic features of these entities to define common elements for future study.

Introduction

The decision for a T cell to mount a response to a target antigen is dependent on the integration of specificity and contextual signals, which can be manipulated pharmacologically for the treatment of cancer and autoimmune diseases. Recognition of an antigenic peptide in the groove of a major histocompatibility complex (MHC) molecule by the T-cell receptor (TCR) provides the specificity and is often referred to as signal 1. However, engagement of signal 1 alone is insufficient to mount a T cell response and can lead to a hyporesponsive state referred to as anergy. Therefore, a contextual signal provided by co-stimulatory molecules present on the antigen presenting cell (APC), referred to as signal 2, is required for full T cell activation. Signal 2 includes the classic engagement of CD28 on the T cell with co-stimulatory molecules B7.1 and B7.2 on the APC, also called CD80 and CD86, respectively (Figure 1). The list of co-stimulatory interactions that function to regulate T cell activation is growing and has been extensively reviewed elsewhere.¹ Following the effector phase of T cell activation and expansion, it is important for the T cell response to downregulate to prevent unchecked inflammation. Inhibitory interactions between the T cell and APC or target cell are critical in this process and protect against autoimmunity by preventing autoreactive T cell activation. A classic example of this is the upregulation of the inhibitory molecule cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) on activated T cells that outcompetes CD28 for binding to CD80/CD86, leading to attenuation of the T cell response. Together, these co-stimulatory and inhibitory interactions are called “immune checkpoints” owing to their important role in modulating immune responses.

Figure 1: Inhibitory checkpoints as regulators of peripheral tolerance and drug targets.

A. Current ICI therapies are monoclonal antibodies (green text) that bind to either CTLA-4, PD-1, or PD-L1 to block inhibitory signaling through these checkpoints, leading to increased T cell activation and proliferation. While this is beneficial for fighting cancer, it can also lead to the development of irAEs. B. Agonists of inhibitory checkpoints in the form of fusion molecules (i.e. PD-L1/Ig and CTLA-4/Ig; red text) can be used as treatments for IA by promoting inhibition of T cell responses. The extracellular domain of CTLA-4 fused to an immunoglobulin (Ig) domain acts to suppress T cell activation by binding to B7.1/B7.2 molecules, sequestering them away from interacting with CD28 on the T cell, mimicking the natural function of CTLA-4. The PD-L1/Ig fusion molecule binds to PD-1 on the T cell, promoting signaling through this inhibitory checkpoint, mimicking the natural function of PD-L1. The effect of these molecules is suppression of T cells as well as reduced monocyte/macrophage effector functions, resulting in overall suppression of pathways that promote IA.

Tumors take advantage of these checkpoints by using inhibitory signals to evade immune surveillance and induce T cell anergy and exhaustion. Many solid tumors can express programmed death-ligand 1 or 2 (PD-L1 or PD-L2), which act as a shield to suppress infiltrating activated T cells that have upregulated the inhibitory receptor PD1. Immune checkpoint inhibitor (ICI) therapies for the treatment of cancer aim to block such inhibitory signals on T cells or tumor cells through the use of antibodies (Figure 1A).² Anti-PD1, anti-PD-L1, and anti-CTLA-4 antibodies have been approved as treatments for many types of cancers, most notably melanoma and non-small cell lung carcinoma.^3,4 However, as these inhibitory signals play a major role in preventing autoimmunity, adverse effects seen in patients treated with ICI therapy include inflammatory responses against self. Pneumonitis, colitis, rashes, myositis, and arthritis are among some of the immune related adverse events (irAE) seen in response to ICI therapy.^3,5

In this review, we will compare the clinical features and mechanisms of ICI-induced inflammatory arthritis (ICI-IA) to canonical forms of inflammatory arthritis (IA) found in patients with rheumatoid arthritis (RA) and spondyloarthritis (SpA). Future research directions integrating the study of ICI-IA and canonical forms of IA will also be discussed. By studying the similarities and differences between ICI-IA and classic forms of IA, we may gain insights into how ICI-IA develops and how immune checkpoint pathways function in RA and SpA pathogenesis. Importantly, the study of ICI-IA offers the unique opportunity to examine the earliest events in the initiation of IA since the inciting event is known, as well as factors that participate in the reestablishment of immune tolerance. In canonical rheumatic diseases, including IA, the inciting events leading to the development of autoimmunity occur years to decades before the onset of clinical symptoms, creating a significant barrier to studying disease initiation and designing preventative strategies. Understanding the clinical and mechanistic parallels between ICI-IA and canonical forms of IA can potentially inform treatment strategies and models of pathogenesis for IA more generally.

Immune checkpoints in IA

Even before the appearance of ICI-IA as a clinical entity, the importance of inhibitory immune checkpoints in the pathogenesis IA was appreciated. This stemmed from mechanistic insights into the function of inhibitory checkpoint molecules in experimental models and the therapeutic efficacy of a CTLA-4/Ig and PD-1/Ig fusion molecules in treating IA in clinical and pre-clinical studies, respectively. These fusion molecules are comprised of the extracellular domains of CTLA-4 or PD-1 fused to an immunoglobulin domain, which have the opposite effect of ICI therapy by promoting engagement of these checkpoints and inhibitory signaling to T cells (Figure 1B). The evidence supporting a role for inhibitory check points in the pathogenesis of IA and hypotheses for how blockade of the CTLA-4 or PD-1 checkpoint could promote the development of IA will be discussed.

CTLA-4 checkpoint

The importance of CTLA-4 as a potent inhibitor of early T cell responses via interaction with B7.1 and B7.2 on antigen presenting cells (APCs) was determined through a series of genetic and molecular experiments. CTLA-4-deficient mice were developed and demonstrated a fatal lymphoproliferative phenotype resulting in multiorgan immune cell infiltrate and autoimmune-mediated death within weeks.⁶ Although T cells were the primary immune targets of genetic perturbation, the inflammatory infiltrate was rich in macrophages and neutrophils, as well as T cells. Prominent infiltration of the heart and pancreas with significant tissue damage was observed, as was non-destructive inflammation in other organs, including the joints of some mice. In contrast to the congenital knock out, systemic deletion of CTLA-4 later in adulthood using the Cre-Lox system, resulted in a non-fatal autoimmune phenotype with a different pattern of organ infiltration that spared the heart.⁷ Inducible deletion of CTLA-4 exacerbated the severity and incidence of arthritis in a mouse model of collagen-induced arthritis (CIA), and robust arthritis development was also observed in CTLA-4 deficient mice who received a suboptimal induction of CIA that resulted in little to no disease in wild-type littermates.⁷ Together, these data provided important evidence for the role of CTLA-4 in preventing IA and suggests a mechanistic foundation for the development of IA in some patients receiving ICI therapy.

Early molecular studies demonstrating that CTLA-4 was a ligand for B7.1/B7.2 and a potent inhibitor of T cell responses, used a CTLA-4/Ig fusion protein to dissect the function of this molecule.^8,9 The CTLA-4/Ig fusion protein was later found to be an effective treatment for IA in the collagen-induced arthritis (CIA) rodent model (Figure 1B).^10,11 The success of this therapy in pre-clinical models led to the approval of CTLA-4/Ig therapy (abatacept) for the treatment of RA in 2005 and later for psoriatic arthritis (PsA) and juvenile idiopathic arthritis (JIA), as well the closely related molecule belatacept for transplant. Additional strategies aimed at modulating the CTLA-4 axis have also shown success in preclinical models, including viral transduction of a CTLA-4/Fas ligand fusion protein in the joints of mice with CIA to simultaneously inhibit T cell co-stimulation and induce T cell apoptosis.¹²

In addition to its effect on T cells, recent evidence suggests that modulation of the CTLA-4 checkpoint has effects on other immune cell types that may be relevant to the development of IA. CTLA-4/Ig was shown to bind to monocytes via engagement of B7.1 and B7.2 on these cells, leading to the production of indoleamine-pyrrole 2,3-dioxygenase, and subsequent apoptosis.¹³ This cascade blocked their differentiation into bone eroding osteoclasts in vitro and in vivo, even in the presence of inflammatory cytokines¹⁴ known to promote osteoclastogenesis.¹⁵ This observation is significant since osteoclasts are known to play a role in bone resorption and the formation of bone erosions in the setting of IA.¹⁶ Blocking this normal inhibitory function of CTLA-4 on osteoclast development in patients receiving anti-CTLA-4 therapy may promote the formation of osteoclasts capable of damaging bone even in the absence of inflammation or activation of joint-specific T cells. In support of this, increased levels of osteoclast precursors were found in untreated patients with RA compared to healthy donors and these levels normalized following treatment with CTLA-4/Ig therapy.¹³ The migratory capacity of monocytes and ability of synovial macrophages to produce inflammatory cytokines in vitro was also attenuated by CTLA-4/Ig.^17–19 A direct role in for anti-CTLA-4 therapy in promoting joint damage¹⁷ is further strengthened by the finding that the levels of circulating osteoclast precursors and potential to form osteoclasts increased in patients receiving anti-CTLA-4 treatment (ipilimumab) for the treatment of melanoma. Together, these data support a mechanism where CTLA-4 blockade may promote the development of ICI-IA through two mechanisms: 1) the reactivation of autoreactive joint-specific T cells; and 2) direct effects on monocytes that promote the production of inflammatory cytokines, extravasation into tissues, and differentiation into osteoclasts.

PD-1 checkpoint

Unlike the CTLA-4 checkpoint, whose predominant role is downregulating T cell responses in the lymph node during interactions with APCs, the PD-1 checkpoint predominantly regulates T cell activity in peripheral tissues upon engagement of PD-L1 and PD-L2.²⁰ In addition, while CTLA-4 is predominantly expressed on T cells, PD-1 is also expressed on natural killer cells, B cells, and activated monocytes.²¹ Compared to mice deficient in CTLA-4, genetic deletion of PD-1 or PD-L1 resulted in a delayed onset autoimmune phenotype, with variable target organ involvement depending on the mouse strain. For example, deletion of PD-1 in B6 mice resulted in the development of spontaneous lupus-like disease by 14 months of age, with proliferative glomerulonephritis and joint inflammation with features of inflammatory infiltrate, synovial hyperplasia, and bone erosion.²² In contrast, PD-1 deficiency on the BALB/c background resulted in premature death at 5–30 weeks from immune mediated damage to the heart resulting in dilated cardiomyopathy.²³ This suggests that background genetics are important influencers of the end-organ damage resulting from disruption of the PD-1 checkpoint, suggesting that this may play a role in the susceptibility to IA in some patients receiving anti-PD-1 or PD-L1 therapy.

The ligands for PD-1 have distinct patterns of tissue distribution, with PD-L2 being primarily expressed on hematopoietic cells, and PD-L1 having broader expression, which includes non-hematopoietic cells to protect self-tissue from autoreactive cells. Mice deficient in PD-L1, PD-L2, or both are phenotypically normal but exhibit immunologic abnormalities upon challenge. This includes increased CD8+ T cell responses following immunization and more severe neurological disease in a mouse model of multiple sclerosis in PD-L1-deficient mice,²⁴ as well as rapid development of diabetes in PD-L1/PD-L2-deficient non-obese diabetic mice.²⁵ Interestingly, osteoarthritis (OA) was also significantly augmented by antibody blockade of PD-L1 in mouse models of osteoarthritis, with more cartilage damage and increased levels of proinflammatory cytokines.²⁶ PD-L1 levels were increased in the joint tissue of mice with OA,²⁶ and analysis of human synovial tissue revealed that both PD-1 and its ligands are elevated in the joints of patients with RA compared to those with OA.²⁷ While PD-1 was only elevated in the joints of RA patients with early and established RA, PD-L1 and PD-L2 were significantly elevated even in patients with pre-RA including in those with arthralgias and undifferentiated IA.²⁷ Transcriptional analysis of synovial biopsies revealed that patients with early and established RA had upregulation of the same set of genes as patients given anti-PD1 therapy (nivolumab) for the treatment of cancer. This suggests a disruption of the PD-1 checkpoint in patients with RA that mimics the effects of pharmacologic PD-1 checkpoint blockade.

The role of the PD-1 checkpoint in maintaining peripheral tolerance to synovial antigens suggests this pathway as a viable therapeutic target in patients with IA. In support of this, a PD-L1/Ig fusion protein was developed and demonstrated therapeutic efficacy in the CIA model of IA (Figure 1B).²⁸ PD-L1/Ig bound to splenocytes from these mice, including activated T cells, and reduced cellular proliferation and production of IL-17 and IL-23 in vitro and in vivo. Like CTLA-4/Ig, the PAD-L1/Ig fusion protein may also affect the function of non-lymphoid cells important for the development of IA, such as macrophages. Lack of PD-1 signaling on tumor-associated macrophages increases phagocytosis, polarization to the inflammatory M1 phenotype, as well as the production of IL-12 and upregulation of co-stimulatory molecules upon stimulation.²⁹ Together, these data suggest that disruption of the PD-1/PD-L1 axis with ICI therapy may promote autoreactive adaptive and pro-inflammatory innate immune responses that may drive the development of IA in genetically susceptible individuals.

Clinical features and Epidemiology of RA, SpA, and ICI-IA

ICI-IA is clinically heterogeneous with some features similar to RA and others reminiscent of SpA.^30–36 Understanding how clinical features and response to treatment of ICI-IA compare to those of known autoimmune diseases can potentially inform our models of pathogenesis for all the disease entities. By studying similarities and differences between ICI-IA and classic forms of IA, we may gain additional insight both for how ICI-IA develops and for how immune checkpoint pathways function in RA and SpA pathogenesis. In this section, we will review the pertinent clinical features of RA and SpA, then compare these entities to ICI-IA (Table 1 and Figure 2).

Table 1:

Comparison of clinical features between ICI-IA, RA, and SpA

	RA	SpA*	ICI-IA
Epidemiology	2–3 times more common in women	AS more common in men	Even gender distribution
	Incidence peaks in 30s-50s	Axial SpA most commonly presents before age 40	Adults 30s-80s exposed to ICI therapy
			Time from exposure to development ranges from weeks-years
Risk Factors	Tobacco smoke	GI/GU infections (ReA)	Largely unclear
	Periodontal disease	Obesity (PsA)	Combination or long duration ICI therapy, multiple irAEs (persistent ICI-IA)
Joint involvement	Small and large joints; classically MCPs, PIPs, wrists	Some subtypes have oligoarthritis, lower extremity predominant	Various subtypes: polyarthritis, large joint oligoarthritis
	Axial disease rare except for C1–2	Axial disease can be seen in all subtypes	Rare reports of axial disease (sacroiliitis)
Imaging findings	Ultrasound: synovitis (Doppler positive); small and large joint effusions; tenosynovitis	MRI: Sacroiliitis with bone marrow edema, erosions, and sclerosis	Ultrasound: synovitis (Doppler positive); small and large joint effusions; enthesitis; enthesophytes
	Radiographs: erosions and joint space narrowing	Radiographs: facet or SI joint ankylosis; syndesmophytes; juxta-articular bone formation (PsA)	Radiographs: erosions can be observed within months of symptoms (rare)
Extra-articular disease	Interstitial lung disease common, can be subclinical	Some subtypes associated with GI inflammation (IBD associated, ReA)	Concurrent colitis reported in those on combination ICI therapy
	Increased risk for lymphoma, lung cancer	Skin psoriasis can be seen in all subtypes	Multiple irAEs common
	Ocular: scleritis	Ocular: uveitis	Ocular: uveitis, may not be concurrent with IA
Prognosis	Chronic; requires long term treatment with DMARDs	Most subtypes are chronic	Varies: some self-limited but others persist
	Common to have disability in 2–3 years if untreated	ReA self-limited in > 90%	Bone damage within months of symptom onset reported

^*Includes AS (ankylosing spondylitis) as well as axial, IBD (inflammatory bowel disease)-associated, and undifferentiated SpA (spondyloarthritis), ReA (reactive arthritis), and PsA (psoriatic arthritis); RA (rheumatoid arthritis); ICI-IA (immune checkpoint inhibitor-induced inflammatory arthritis); irAE (immune related adverse event); DMARD (disease modifying anti-rheumatic drug); MRI (magnetic resonance imaging)

Figure 2: Immunologic and Clinical Features of IA.

A summary of immune infiltration, inflammation, and pathology seen in healthy (A.), RA (B.), SpA (C.), and ICI-IA (D.) joints is shown. A. Healthy joint anatomy is shown. B. The synovial fluid of patients with RA is enriched in neutrophils and autoantibodies, while other immune cells (i.e. T cells B cells, NK cells, dendritic cells, and macrophages) and tertiary lymphoid structures can be found in the synovial tissue. Bone erosion mediated by osteoclasts, synovial hyperplasia, and pannus formation are also seen. The strongest genetic risk factors for RA development are a group of HLA-DRB1 alleles, referred to as the “shared epitope alleles”. RF and ACPAs are hallmark serological findings. C. Inflammatory infiltrate comprising neutrophils and lymphocytes is observed in the joints of patients with SpA. In addition, enthesitis is common, as are paradoxical findings of bone formation and erosion. Susceptibility to spondyloarthritis is conferred by HLA-B27 alleles. Patients are mostly seronegative, although in some cases, autoantibodies can be found. D. Although features of RA and SpA are seen in ICI-IA (neutrophil infiltration, enthesitis, synovitis, and sometimes erosion), patients are largely seronegative, and not much is known about the nuances of the inflammatory infiltrate, genetic susceptibility, or antigenic targets.

Rheumatoid arthritis

Epidemiology

RA is a systemic autoimmune disease most notable for chronic inflammatory arthritis that can lead to erosive joint disease and permanent joint damage. Though joints are most predominantly affected, patients may also experience systemic symptoms and extra-articular disease. The prevalence of RA varies depending on country and study, but is estimated to affect 0.5–1% across different populations worldwide.³⁷ In one systematic review, Northern European and North American countries had higher incidence rates of RA than Southern European countries and developing nations. Women are affected more often than men, making up about two-thirds of the population with RA.³⁸ Family history is a risk factor for developing RA, with heritability about 40% higher in seropositive patients than in seronegative patients.³⁹ Age of onset can vary, but peaks around 50 years of age in studies from North American and European countries.⁴⁰ A variety of environmental risk factors for the development of RA have been identified including exposure to tobacco smoke, silica dust, air pollution, obesity, and low vitamin D levels.³⁸

Clinical features

Classically, RA is described as symmetric polyarthritis predominantly affecting the hands and wrists. With efforts to identify disease earlier and facilitate intervention, including changing classification criteria,⁴¹ fewer patients present with symmetric polyarthritis, though this certainly evolves over time. Patients may present differently in terms of number and specific joints involved. RA can affect almost any peripheral joint, while axial disease is typically limited to the cervical spine, particularly the C1–2 articulation. Of the peripheral joints, the most commonly affected include the small joints of the hands and wrists. Patients may experience pain, swelling, or stiffness of their joints. Longstanding or undertreated disease can lead to structural deformities such as flexion contractures in the elbows, ulnar deviation of the phalanges at the metacarpal phalangeal (MCP) joints, and swan neck or boutonniere deformities of the fingers. Tenosynovitis, or inflammation of the tendon sheath, has become increasingly recognized in early RA using ultrasound and magnetic resonance imaging (MRI).⁴² Evaluation with plain radiographs can show joint space narrowing, erosions, and subchondral cysts, but MRI and ultrasound are more sensitive in detecting erosions and can define synovitis which is not apparent on plain radiography. Ultrasound features of RA include synovial hypertrophy, power Doppler signal in the synovium indicating synovitis, joint effusions, erosions, and tenosynovitis. Before the availability of biologics and the treat to target approach, bone erosions developed in many patients in the first two years of disease.⁴³

A variety of extra-articular features are seen in people with RA. Interstitial lung disease (ILD), rheumatoid nodules, rheumatoid vasculitis, and scleritis are some of the conditions that can accompany joint disease. ILD varies in severity and is clinically apparent in 6 to 30% of RA patients depending on how it is diagnosed.⁴⁴ However, asymptomatic subclinical ILD has been reported in up to 60% of patients. Patients with RA also have an increased risk of lymphoma, lung cancer, and cardiovascular disease.^45,46

Disease course and prognosis

The advent of more aggressive management with conventional synthetic disease-modifying anti-rheumatic drugs (DMARDs), biological DMARDs, and targeted synthetic DMARDs combined with earlier disease recognition, initiation of definitive DMARD therapy, and treat to target approaches have drastically changed the prognosis of RA. Disability occurred early in older studies of patients with RA. In a study of two cohorts of RA patients from the 1980s and 1990s, 20–30% of patients became unable to work within three to four years of diagnosis.⁴⁷ Disease activity levels and disability have fallen in the last 15 years with more intensive treatment.^48,49 Despite the advances in therapy, RA is a chronic disease where most patients need some form of DMARD therapy indefinitely.

Spondyloarthritis

Epidemiology

There are several forms of chronic inflammatory arthritis that are considered within the family of spondyloarthritis (SpA) due to their similar clinical characteristics and shared pathogenic features. Five main subgroups include axial SpA, psoriatic arthritis (PsA), reactive arthritis, inflammatory bowel disease (IBD)-associated arthritis, and undifferentiated SpA.^50,51 Axial SpA is often divided into ankylosing spondylitis and non-radiographic axial SpA. The prevalence of SpA is less well defined than RA due to changes in classification criteria over the last several decades, ranging from 0.01 to 2.5% depending on the study.⁵² Some forms of SpA, like ankylosing spondylitis, occur more commonly in men. HLA-B27 is a risk factor for SpA, particularly in axial SpA (see “Genetics” section).

Clinical features

The arthritis in SpA may involve the axial skeleton, peripheral joints, or both.⁵¹ Inflammatory back pain is a major feature of SpA and is not always associated with radiographically visible disease. Inflammatory changes associated with sacroiliitis (inflammation of the joints connecting the sacral vertebrae to the ilium of the pelvis) and spondylitis (inflammation of the vertebrae) can be seen with MRI, but become apparent as erosions, sclerosis, or fusions only later in the disease process with plain radiographs. Structural changes in the spine like syndesmophytes and ankylosis of the facet or sacroiliac joints are classic findings of ankylosing spondylitis and can also be seen on radiographs. Dactylitis or swelling of an entire digit along the tendon sheath, is unique to SpA. Enthesitis, inflammation at tendon insertion sites, can also occur in all SpA subsets. There are several unifying clinical features amongst different forms of SpA as well as other features which distinguish subtypes.

Extra-articular manifestations include inflammatory bowel disease, skin and nail psoriasis, uveitis, cardiovascular disease, and lung disease. Uveitis is common, occurring in 32.7% of patients with SpA in one systematic review.⁵³ Reactive arthritis, which often follows a gastrointestinal or genitourinary infection, is associated with conjunctivitis and a sterile urethritis.

Disease course and prognosis

Most subsets of SpA are chronic diseases, with the exception of reactive arthritis, which is self-limited in the majority of patients.⁵⁴ PsA is clinically heterogeneous; some patients have mild symptoms that may be limited to dactylitis alone or oligoarthritis, while others can have arthritis mutilans, a destructive form of arthritis leading to significant joint deformities.⁵⁵ Despite recent advances in treatment, disability is still common, and half of patients with SpA reported disability in one recent survey of Italian patients.⁵⁶

ICI-induced IA

Epidemiology

There is a paucity of epidemiologic data for ICI-IA. The prevalence of IA may range from 3–7.5% in those treated with ICIs,^57–59 but there is a lack of prospective data, and clinical trial data is limited. Unlike RA or some forms of SpA, which are more prevalent in women or men, respectively, there does not seem to be a relationship between sex and developing ICI-IA, as most cohorts are 50–65% male.^{30,33,36,57,60} The age of development can vary greatly, as there are many indications for ICI treatment that may be prevalent in older or younger patients.^30,33,57,60 Tumor type does not appear to be related to developing IA, as it has been seen in patients with a diversity of tumor types.³⁰ Without good estimates of prevalence of IA in patients treated with different classes of ICIs, it is hard to determine how specific agents may influence epidemiology. Some clinical features of ICI-IA do vary based on combination CTLA-4/PD-1 blockade versus monotherapy with PD-1 or PD-L1 blockade,³⁰ so it is conceivable that rates of development may also differ by treatment regimen.

Clinical features

Joint disease in patients with ICI-IA can look like RA or SpA in terms of number and size of joints involved and accompanying features. Some patients have large joint involvement first and only develop an oligoarthritis, more consistent with SpA. Others have the small joints of the fingers or wrists affected first, in a more RA-like presentation. Tendon involvement is common; tenosynovitis, enthesitis, enthesophytes, and dactylitis are all appreciated.^{30–33,35,61,62} However, inflammatory axial disease has been less commonly reported.⁶³ Rapid bone resorption near a joint that would be atypical for both SpA and RA has also been seen with ICIs.⁶⁴

Imaging has been characterized for a limited number of patients with ICI-IA, but there are some notable findings. One striking feature is the accelerated time course for the development of structural damage. Erosions similar to those seen in RA or PsA can occur within months of symptom onset.^31,62 Similarly, enthesophytes, which are also representative of structural change, were present in some patients at first ultrasound evaluation.⁶²

Extra-articular disease is common for patients with ICI-IA since many can have other irAEs during their course of ICI treatment. Colitis, pneumonitis, thyroid disease, and rash are well-known irAEs that can occur in patients who also develop ICI-IA. In one study of patients with ICI-IA, all those who had a reactive arthritis phenotype also had colitis and had been treated with combination ipilimumab/nivolumab therapy.³⁰ There are no clear patterns described, however, of which types of irAEs are more likely to cluster in the same patient.

Disease course and prognosis

Some irAEs affecting other organ systems, such as the gastrointestinal tract, are monophasic and self-limited, resolving after ICI cessation and/or corticosteroid or brief courses of immunomodulatory therapy.⁶⁵ In ICI-IA, however, there is a large subgroup of patients with persistent disease up to several years after the ICI is stopped.⁶⁶ In one study, 49% of patients with follow up at 6 months post-ICI cessation still had active IA. Interestingly, as in traditional reactive arthritis, patients with a reactive arthritis phenotype after ICI have had a self-limited course.

Overall, the observation that ICI-IA is a heterogeneous entity, with some clinical features of RA or SpA and other unique features (Table 1), has potential significance for understanding pathogenesis and treatment decision making. Further study of whether patients with differing clinical presentations of ICI-IA (i.e. more or less similar to RA or SpA) have different risk factors for developing IA, respond differently to immunosuppressive therapies, or are more or less likely to develop persistent IA will clarify the significance of the clinical heterogeneity.

Serology

Antibodies targeting self-proteins are hallmark serological findings in patients with autoimmune rheumatic diseases. They often appear in the pre-clinical phase of disease, which can persist for months or years before the onset of clinically apparent symptoms. In addition, autoantibodies targeting specific self-proteins are strongly associated with the resulting disease phenotype. This makes autoantibodies sensitive and specific diagnostic biomarkers, as well as immunologic records of important pathways that may be involved in disease pathogenesis. As such, the search for autoantibodies in RA has yielded important insights into disease pathogenesis, and similar studies have been undertaken in SpA and more recently, ICI-IA.

RA autoantibodies

Seropositivity in RA is characterized clinically by the presence of anti-citrullinated protein antibodies (ACPAs) or rheumatoid factor (RF) and is included in the 2010 ACR classification criteria for RA.⁴¹ RF are IgM autoantibodies against the Fc portion of IgG antibodies and are present in 60–80% of RA population.^67–71 ACPAs are also in 60–80% of patients with RA, depending on the study, and are highly specific for the disease.^72–77 These antibodies recognize specific self-proteins in which arginine residues have been deiminated by peptidylarginine deiminase (PAD) enzymes, generating the non-classical amino acid citrulline.⁷⁸ ACPAs are present in RA patients years before they show symptoms,^79,80 and epitope spreading with recognition of an increased number of specific citrullinated proteins occurs within a 2-year window of disease diagnosis.⁷⁹ ACPAs are detected clinically by the anti-cyclic-citrullinated peptide (anti-CCP) assay. It is important to highlight that the peptides used in the anti-CCP assay are short synthetic citrullinated peptides, originally based on peptides from citrullinated filaggrin, but are artificial antigens that are able to detect the pool of ACPAs present in most patients with RA.^81,82 However, studies have shown that a small fraction of patients who are negative by the anti-CCP assay, have detectable ACPAs when specific citrullinated proteins are used as antigens.⁸³

Other autoantibodies are well described in patients with RA but are present less frequently and are currently not measured in clinical settings. These include members of the PAD enzyme family itself. Although there are five PAD isoenzymes in humans, PAD2 and PAD4 are implicated as the dominant producers of citrullinated autoantigens in RA. Both enzymes have also been shown to be recognized by autoantibodies in different subsets of patients with RA.^84–86 While anti-PAD4 antibodies are associated with severe erosive joint disease and are strongly associated with classic RA genetic and serologic risk factors, anti-PAD2 antibodies appear to identify a serologically and genetically distinct group of patients with milder disease.^85,87–89 Another well-described target in RA is a nuclear antigen known as RA33. Initially defined by western blot as a 33kDa protein in nuclear extracts from HeLa cells that was recognized by sera from RA patients, RA33 was later identified as hnRNP A2/B1, a protein of the spliceosome that can be targeted by autoantibodies in its citrullinated or non-citrullinated form.^90–92

A growing number of other anti-modified protein antibodies (AMPAs) have been identified in patients with RA, including those to carbamylated (anti-CarP) and acetylated antigens as well as malondialdehyde-acetaldehyde (MAA) adducts.^93,94 Formed as a result of the lipid peroxidation product malondialdehyde (MDA) chemically reacting with acetaldehyde and proteins, MAA adducts have been seen to co-localize with citrullinated proteins by confocal microscopy and are recognized by autoantibodies in 38% of ACPA-positive and 26% of ACPA-negative patients with RA.⁹⁵ Carbamylation, the conversion of lysine residues to homocitrulline through a reaction with cyanate, is also recognized by autoantibodies in a subset of patients with RA. Anti-CarP antibodies are present in 49–73% of ACPA-positive and 8–16% of ACPA-negative patients, and correlate with worse joint damage in ACPA-negative patients.^96,97 Anti-acetylated vimentin antibodies have also been found in an average of 52.1% of ACPA-positive and 21.7% of ACPA-negative patients with RA.⁹⁴ Despite the growing study of these new antibody specificities in RA, their diagnostic and prognostic value in the clinical setting has yet to be evaluated.

Although seropositive RA patients are the majority, seronegative RA does exist. Clinically, seronegativity in RA is defined as being negative for RF and ACPA. However, since other autoantibodies have been identified, as described above, it is important to understand if truly seronegative patients exist or if additional biomarkers may help to close the serological gap in patients with RA (reviewed in Trouw and Mahler 2012⁹⁸). This is important since seronegative patients are less likely to fulfill disease classification criteria and may be more difficult to diagnosis, possibly leading to delays in treatment.⁹⁹ Once thought to represent patients with a milder form of RA, recent studies suggest that seronegative patients can have severe treatment-refractory disease, but that remission can be achieved if treatment is initiated early in the disease course.¹⁰⁰

SpA autoantibodies

SpA is generally considered clinically to be a seronegative disease, due to the paucity of RA-associated autoantibodies and lack of success identifying prevalent and specific autoantibodies in this patient population. While this may suggest that SpA patients do not mount efficient autoreactive humoral responses or that this disease is autoinflammatory in nature, autoantibodies to β2 microglobulin, protein phosphatase magnesium-dependent 1A (PPMA1), CD74, LL37 and others have been found in SpA, most commonly in single-cohort studies (reviewed in Liu et al 2019¹⁰¹).^102–104 Antibodies to β2 microglobulin were among the first to be described and were found in 68% of ankylosing spondylitis patients in a single study, but were not specific for the disease when compared to patients with lupus or RA.¹⁰³ Anti-PPMA1 antibodies have also only been described in a single study, but were found to be 67% sensitive and 73% specific for SpA, compared to other forms of IA, and a decrease in the level of these antibodies was associated with a positive clinical response to therapy.¹⁰⁴ CD74, which is the class II-associated invariant chain peptide (CLIP) that participates in peptide binding and correct folding of class II HLA molecules, is one of the more widely studied autoantigens in SpA.¹⁰⁵ Anti-CD74 was originally found to be an autoantigen in 56–67% of European patients with SpA and was 95% specific for the disease.¹⁰⁶ These autoantibodies were more prevalent in patients with axial disease (range 69–85%) or disease duration less than one year.^106,107 Although the prevalence of anti-CD74 antibodies has been found to be lower in Asian and Middle Eastern cohorts, a study of Lebanese patients with SpA found that analysis of the IgG4 isotype of anti-CD74 provided robust positive and negative predictive value for the disease.^101,108 Together, this suggests that further study is need on anti-CD74 antibodies and other serologies in SpA to translate these findings into the clinic.

ICI-IA autoantibodies

Similar to SpA, patients with ICI-IA have been primarily found to be seronegative for RF and ACPA,^{30,31,36,57–59,109–112} and an in-depth analysis of other serologies has not been performed. Thus far, most reports are from small case series or single center cohort studies and the assays used to detect seropositivity are not uniform between studies. Despite these limitations, aggregation of the data from multiple studies reveals a low rate of seropositivity among patients developing IA following treatment with ICI therapy with 5.5% (12/217; range 0–36%) positive for RF and 5% (11/217; range 0–11%) positive for ACPAs.^{30,31,33,36,57–59,66,109–115} In a study comparing ICI-IA patients to ethnically matched counterparts with RA, 7.7% of ICI-IA patients were positive for RF or ACPAs; whereas, 64.6% of RA patients were positive for ACPAs and 56.7% positive for RF, with 49.3% being positive for both.¹¹⁶ Although these results indicate that the prevalence of seropositivity in ICI-IA is much lower than in RA, it is unknown if seropositivity is increased in patients who develop RA-like disease following ICI therapy.

Only a few studies have looked at the prevalence of seropositivity in patients with ICI-IA whose clinical features met classification criteria for RA or SpA.^41,117–119 A large French study on irAEs from anti-PD1, anti-PD-L1 and/or anti-CTLA-4 therapy identified 30 individuals who developed systemic irAEs after treatment.¹¹⁷ Three patients fulfilled the classification criteria for RA and all three developed positivity for RF and ACPAs. Two other patients developed seronegative psoriatic arthritis which met classification criteria for SpA.^117,119 A different study by the same group analyzed seropositivity in six patients who developed classifiable RA following ICI therapy. All six patients were positive for ACPAs, and four were also positive for RF.¹¹⁸ The increased prevalence of seropositivity in patients who fulfill classification criteria for RA, compared to the larger group of ICI-IA patients, is not surprising given the weight ascribed to the presence of ACPAs or RF in making a diagnosis of RA. This highlights that clinical features of RA with classic serologies can develop following treatment with ICI therapy.

An important caveat in the study of autoantibodies post-ICI therapy, is whether these antibodies were present in individuals prior to treatment. Since autoantibodies can develop years or decades prior to the onset of RA,⁷⁹ analysis of pre-treatment serum is critical for addressing this question. While not uniformly tested, some studies have analyzed banked serum from patients prior to ICI exposure.^{58,59,117,118} For example, in the previously mentioned study by Belkhir et al, out of 3 patients who were APCA-positive positive post-ICI treatment and had pre-treatment serum available, two patients were already positive for ACPAs pre-treatment.¹¹⁸ Flares of existing IA that had been controlled by medication have also been reported following ICI therapy, indicating that these treatments can exacerbate existing disease.^57,110,120 Thus, it is possible that those who develop seropositive RA-like inflammatory arthritis after ICI treatment were already immunologically pre-disposed to developing the disease.

While the majority of patients who develop IA after ICI therapy are seronegative, as defined by the absence of RF or ACPA, further study is needed to determine whether these patients are positive for other RA-or SpA-associated autoantibodies (as described in “Serology” section) or possess a unique set of serologies that suggest a distinct mechanism. This is important to define, since the true absence of autoantibodies can also be mechanistically informative. Although enhanced T cell activity is the primary goal of ICI therapy, these therapies may affect B cells directly (discussed in “Immune checkpoints in IA” section) or indirectly by modulating the role of CD4+ T helper cells. Since CD4+ T helper cells provide help to B cells to undergo somatic hypermutation and class switch recombination in order to produce high-affinity antibodies, T cell targeted therapies may be expected to impact the production of autoantibodies. In support of this, the presence of anti-acetylcholine receptor or anti-thyroid autoantibodies has been shown to be associated with the development of myasthenia gravis or thyroid dysfunction, respectively, after ICI treatment.^121,122 Furthermore, production of de novo autoantibodies associated with arthritis or specific endocrine, dermatologic, or gastrointestinal disorders was associated with developing any irAE following anti-CTLA4 therapy, but was not linked to the organ-specificity of the irAE.¹²² While this suggests that the development of ICI-IA may be associated with other serologies not traditionally linked to arthritis, there are other reasons why autoantibodies may exist but have not yet been discovered: 1) the appropriate detection method, source of target antigens, or modified antigens have not yet been studied; 2) a longer follow-up period is needed to capture development of autoantibodies; or 3) ICI-IA is a heterogenous group of conditions and comprehensive serological testing may reveal distinct subsets that mirror known types of IA.

The possibility that ICI-IA may encompass distinct subgroups of inflammatory arthropathies is reminiscent of JIA. There are currently seven subgroups of JIA that have been defined, distinguished by clinical course and serological features (reviewed in Mahmud and Binstadt 2019¹²³). Like patients with ICI-IA, the RF-positive subgroup of JIA that is similar to adult RA comprises only a small fraction of the total JIA population, less than 5%, and many of the RF-positive pediatric patients also have ACPAs.^124,125 While the other JIA subgroups are largely seronegative, with similar clinical features to seronegative RA or SpA, 37–73% of patients with JIA are positive for anti-nuclear antibodies (ANA), depending on the study.^126,127 While ANA is not specific for a particular subgroup, it does correlate with an increased incidence of JIA-associated uveitis, particularly iridocyclitis.¹²⁸ Aggregate analysis of ANA positivity, as defined by a titer greater than 1:40, from several published studies revealed that 15.3% (32/209; range 0–89%) of patients who developed ICI-IA were positive for ANA.^{30,31,36,57–59,66,109–115} In addition, antibodies against specific nuclear antigens, dsDNA, Ro, and La (usually seen in patients with systemic lupus erythematosus or Sjögren’s syndrome), have been detected in ICI-IA patients.^31,36,112 Although ICI-IA develops in a subset of adult patients treated with ICI therapy, it is possible that the pathogenesis and development of this disease may more closely parallel the development of JIA rather than adult RA or SpA. As ICI-IA becomes an increasingly recognized entity, a deeper dive into the serologic profile of these patients is warranted and may reveal novel insights into the mechanism of ICI-IA and IA subtypes more broadly.

Genetics

The single most important genetic risk factor for the development of autoimmune disease lies within chromosome 6 of the human genome, in the Human Leukocyte Antigen (HLA) locus. This locus houses the highly polymorphic genes that encode the alpha and beta chains of MHC class II (HLA-DR, -DP, and -DQ) and alpha chains of MHC class I (HLA-A, B, and C) molecules, as well as several other less variable immune genes.¹²⁹ Genetic studies have revealed a strong association of RA and SpA with specific MHC class II and class I alleles, respectively, as well as single nucleotide polymorphisms (SNPs) in other non-HLA genes located throughout the genome, but the genetic contribution to IA development following ICI therapy is only beginning to be addressed.

RA genetic associations

The genetic susceptibility to RA was initially attributed to the MHC class II molecule, HLA-DR4.^130,131 However, it was hypothesized and later demonstrated that this risk was conferred by the presence of a 6 amino acid long sequence at residues 70–74 of the HLA-DRβ chain (i.e. QKRAA) and that other alleles associated with RA contained similar motifs (i.e. QRRAA or RRRAA).^132–134 Since the original method used for HLA-typing relied on recognition of different HLA molecules by monoclonal antibodies,¹³⁵ the common epitope recognized in this group of RA-associated HLA variants was termed the “shared epitope (SE)”. SE alleles include: HLA-DRB1*04:01, *04:04, *01:01, *01:02, *04:05, and *10:01.¹³⁴ Studies have since shown that, together, SE alleles are associated with more severe RA as well as developing ACPAs.^136,137 On the other hand, HLA-DRB1*03, *08, and other alleles encoding for a DERAA motif in this same position on the HLA-DRβ chain tend to be associated with seronegative RA.^138,139 In addition to the MHC locus, SNPs in other genes have also been linked to an increased risk for developing RA in certain ethnic populations. These include PTPN22, IL23R, STAT4, PADI2, and PADI4, among others (reviewed elsewhere¹⁴⁰). SNPs in the CTLA4 gene itself have also been shown to confer risk for the development RA in some studies.^140,141

SpA genetic associations

The majority of the genetic susceptibility to SpA is attributed to the HLA class I family of HLA-B27 alleles, especially the HLA-B*27:02, *27:04, and *27:05 variants.^{138,142–144} Studies have also found associations with other HLA-B alleles such as HLA-B*40:01, HLA-B*47:01, and with the HLA-A, HLA-DPB1, and HLA-DRB1 loci (reviewed elsewhere¹³⁸).¹⁴⁴ SNPs in genes outside of the MHC locus have also been linked to an increased risk for developing SpA, including those in ERAP1 and IL23R (reviewed elsewhere¹⁴⁵).¹⁴⁴

ICI-IA genetic associations

While studies about the genetic risk for RA and SpA are vast, genetic studies on ICI-IA are limited. Given that the mechanism of action of ICI therapy is focused on the interaction between T cells and their targets, an interaction requiring recognition of a specific peptide bound to an MHC molecule, it may be expected that individuals who possess RA-or SpA-associated HLA alleles may be at increased risk for developing IA post-treatment with ICI therapy. A 2019 study by Cappelli et al analyzed the HLA allele frequencies in 26 patients with ICI-IA.¹¹⁶ They found that having at least one RA-associated SE allele increased the probability of developing ICI-IA compared to the control population (p = 0.04, OR 2.3, 95% CI = 1.0–5.1). The SE allele HLA DRB1*04:05 was most strongly associated with developing IA in this study (p = 0.04, OR 8.6, 95% CI = 1.7–43.4). Other alleles such as HLA A*03: 01, HLA B*52:01, HLA C*12:02 were also enriched in patients with ICI-IA, while the frequency of HLA DQB1*03:01 was decreased in frequency compared to healthy controls, but these results were not statistically significant. In another study, the SE allele HLA DRB1*01:01, was found in two out of nine patients who had inflammatory arthritis after treatment with an ICI, however, comparison to a control population was not reported.⁵⁷ Interestingly, several studies have shown that SpA-associated HLA-B27 alleles do not appear to confer increased risk for ICI-IA, as almost all patients who developed ICI-IA were HLA-B27 negative.^{30,58,111,114,116} The genetic studies performed thus far show that SE, but not HLA-B27 alleles, associate with ICI-IA. However, it is important to note that these studies included all comers with ICI-IA, so the association of specific HLA alleles with the development of SpA-like or RA-like disease is unknown.

Although SNPs in other genes have been found that confer increased susceptibility to RA or SpA, such genetic studies have not yet been performed for ICI-IA, likely owing to the relatively small number of patients with this irAE. A single study has evaluated the effect of SNPs in genes associated with PD-1-dependent T cell responses (i.e. PDCD1, PTPN11, ZAP70 and IFNG) on the development of irAEs following treatment with nivolumab (anti-PD-1 therapy).¹⁴⁶ Although the development of ICI-IA was not specifically reported, a SNP in IFNG previously found to be associated with systemic lupus erythematosus¹⁴⁷ was shown to increase the risk for developing a rheumatologic irAE in a discovery cohort.¹⁴⁶ A variant of PDCD1 was also associated with developing any irAE in this cohort, but neither of these findings were replicated in a validation cohort.¹⁴⁶ Another study investigated the link between SNPs associated with traditional forms of autoimmunity and response to ICI-treatment in melanoma patients and found that a SNP in PTPN2 and a SNP in the locus containing IL2, ADAD1, and IL21 genes was associated with better response to anti-CTLA-4 and anti-PD1 therapies, respectively.¹⁴⁸ While this suggests a connection between background genetics and treatment response or developing an irAE, it will be important to examine the genetic underpinnings of ICI-IA and other irAEs more broadly in larger cohort studies.

Immune Dysregulation

Persistent or chronic inflammation of the joints is a hallmark feature of RA and SpA that is unlikely to resolve without immunomodulatory treatment. Infiltrating immune cells mediate the remodeling of the bone in the joint space which contributes to the development of joint pathology and symptoms (Figure 2). In RA, bone resorption is the most prominent feature, while in SpA, both bone formation and resorption are seen. In addition to immune cells, joint-resident stromal cells such as fibroblast-like synoviocytes and mesenchymal cells add to the inflammatory milieu leading to a positive feedback cycle of unresolved inflammation. The roles of these cells in the pathogenesis of ICI-IA is less well understood, but there are emerging studies that demonstrate parallels with traditional forms of IA.

Immune dysregulation in RA

Neutrophil infiltration is a key characteristic of inflammation seen in RA, with neutrophils comprising the majority of the innate immune cells found in the synovial fluid of RA patients.¹⁴⁹ Neutrophils are important producers of TNF-α, a central cytokine in RA pathogenesis and target of therapy for the treatment of RA (Table 2). A recent study showed that treatment of RA patients with anti-TNF-α antibodies reduced TNF-α production by neutrophils to a level similar to that of healthy controls.¹⁵⁰ In addition to inflicting damage at sites of inflammation through proteases and secreting cytokines which recruit additional immune cells, they also release neutrophil extracellular traps (NETs) comprised of intracellular proteins bound to DNA.¹⁴⁹ This process may redistribute autoantigens, such as citrullinated proteins, to the extracellular environment where they may be bound by autoantibodies and drive downstream inflammatory processes such as macrophage or complement activation.^151,152 These other innate effector mechanisms and cells, including dendritic and natural killer (NK) cells, are critical to RA pathogenesis as they sustain inflammation and contribute to adaptive immune cell activation.¹⁵³

Table 2:

DMARDs used for the treatment of IA

Drug/Drug class	Efficacy in RA?	Efficacy in SpA?	Use in ICI-IA?
Hydroxychloroquine	Yes	No	Yes
Sulfasalazine	Yes	Yes (peripheral arthritis only)	Yes
Methotrexate	Yes	Yes (peripheral arthritis only)	Yes
Leflunomide	Yes	Yes (PsA)	Yes
Apremilast (PDE4 inhibitor)	No	Yes (PsA)	Not reported (use in PsO only)
Abatacept (CTLA-4 Agonist, CTLA4-Ig)	Yes	Yes (PsA)	Not reported
TNF-inhibitors	Yes	Yes (axial and peripheral arthritis)	Yes
IL-6 inhibitors	Yes	No	Yes
IL-12/23 inhibitor	No	Yes (PsA)	Not reported (use in PsO only)
IL-17 Inhibitors	No	Yes (axial and peripheral)	Not reported
JAK-Inhibitors	Yes	Yes-PsA	Not reported
Rituximab (Anti-CD20 antibody)	Yes	No	Not reported

RA (rheumatoid arthritis); SpA (spondyloarthritis); ICI-IA (immune checkpoint inhibitor-induced inflammatory arthritis); PsA (psoriatic arthritis); PsO (psoriasis without arthritis); DMARD (disease modifying anti-rheumatic drug); PDE4 (phosphodiesterase 4); CTLA-4 (cytotoxic T-lymphocyte-associated protein 4); TNF (tumor necrosis factor); IL (interleukin); JAK (janus kinase)

In addition to the clear role for the innate immune system in RA pathogenesis, there is an important role of the adaptive immune system with the presence of antigen-specific T and B cells. Lymphoid structures found in inflamed joints of RA patients, referred to as tertiary or ectopic lymphoid structures, contain distinct T cell zones and germinal centers.^154,155 B cells present in these structures have been shown to secrete high affinity, class-switched ACPAs locally,¹⁵⁶ but in vitro studies suggest that B cells from synovial fluid and bone marrow of ACPA-positive RA patients can also secrete ACPAs even without T cell help.^71,157 While ACPA production may occur with or without T cell help, it is clear that T cells also play a major role in the pathogenesis of RA. CD4+ T cells have been identified that recognize self-peptides presented by HLA-DR molecules carrying the “shared epitope motif” (discussed in “RA genetic associations” section) linked to RA development, and can differentiate into Th1 and Th17 cells.^158,159

Though not considered to be conventional immune cells, fibroblast-like synoviocytes in the joint augment the immune response via hyperplasia and pannus formation that invades cartilage, production of degradative enzymes that contribute to bone and cartilage damage, and secretion of pro-inflammatory cytokines.¹⁶⁰ Thus, the inflammatory infiltration not only causes local tissue damage, but also promotes changes to resident cells in the synovium, which creates a feedback loop that propagates inflammation the joint.

Immune dysregulation in SpA

Similar to RA, neutrophils, macrophages, dendritic cells and NK cells are all found in the inflamed synovium in SpA.¹⁶¹ However, in contrast to RA, the amount of neutrophil infiltrate is significantly lower and the amount of lymphocytic infiltrate is markedly higher.¹⁶² Infiltration both CD4+ and CD8+T cells can be seen in the joints of patients with ankylosing spondylitis with TNF-α mRNA seen in the cellular infiltrates.¹⁶³ IL-17 producing cells are also enriched at sites of joint inflammation in SpA compared to patients with OA and include IL-17 producing T cells as well as neutrophils and neutrophil precursors.¹⁶⁴ Increased numbers of IL-17 producing Th17 cells are also found in the peripheral blood of patients with SpA compared to healthy controls, with expression of IL-23 receptor on these cells.¹⁶⁵ Enthesitis, or inflammation at the sites where tendons or ligaments insert into the bone, is a hallmark characteristic of SpA, and IL-23 is implicated a key cytokine in this inflammatory process via its ability to support Th17 cell expansion and survival (reviewed in Schett et al¹⁶⁶).¹⁶⁶ Local production of prostaglandin E2 by joint-resident mesenchymal cells and IL-23 production by macrophages or dendritic cells has been shown to drive the recruitment and activation of IL-17-producing T cells in mouse models. In addition to classic Th17 cells, other IL-17 producing cells such as type 3 innate lymphoid cells and γδ-T cells may also play a role, leading to the secretion of IL-17 and TNF-α, which augment inflammation by recruiting neutrophils and macrophages. Remarkably, enthesitis in SpA results in new bone formation rather than destruction (e.g. in the spine in ankylosing spondylitis and in the hand in psoriatic arthritis). While not fully understood, mesenchymal cells are implicated in responding to the inflammatory mediators produced by entheseal inflammation by differentiating into chondroblasts and osteoblasts that contribute to aberrant bone formation, in a process reminiscent of bone fracture repair.¹⁶⁶ Bone growth can also be coupled to bone erosion mediated by osteoclasts that develop from precursors recruited to the bone surface.¹⁶⁷ As a result, SpA is characterized by seemingly contradictory mechanisms of bone growth and bone resorption resulting from joint inflammation.

Immune dysregulation in ICI-IA

Although the data is scare, cases studies have shown that the synovium of patients with ICI-IA can have features of RA and SpA. Similar to RA, neutrophils are the predominant cell type seen in the synovial fluid of patients with ICI-IA^31,34,168 Combination anti-CTLA-4 and anti-PD1 therapy has also been shown to increase circulating plasmablast numbers and CXCL13 levels compared to either treatment alone, suggesting augmented germinal center activation.¹⁶⁹ In addition, these patients had elevated levels of CD21-low expressing B cells, which expressed higher levels of PD-1 and were more clonal than CD21-high B cells. Although not specific to ICI-IA development, this study also found that changes in B cells induced by ICI therapy increased risk of developing multiple irAEs. As ICI-IA increases in incidence, it will be important to thoroughly examine the inflammatory infiltrate and histological changes occurring in the joint to identify pathogenic processes and determine if immunologically distinct subsets exist. Importantly, the early events driving inflammation in RA and SpA are unknown. Since ICI-IA is distinct in that the inciting event is known (i.e. treatment with ICI therapy), studying the joint histology of ICI-IA patients with RA-like and SpA-like arthritis could help identify the factors that initiate inflammation in classical RA and SpA.

Mechanism-guided treatment strategies

Due to the identification of common immunologic pathways contributing to the pathogenesis of RA and SpA these clinical entities also share several common treatments, most notably TNF-inhibition (Table 2).¹⁷⁰ However, there are also important differences in pathogenesis resulting in certain medications having efficacy in only one disease category.^170–172 Some of the therapies successful in one form of IA have been investigated in clinical trials in the other but with limited efficacy, while others may be in development in the other diseases. Given the lack of clinical trial data in ICI-IA and lack of a deep understanding of disease mechanism, treatment of ICI-IA has borrowed from what is known about RA and SpA.

Recommendations and Future Directions

One of the limitations to studying and analyzing arthritis as an irAE is the lack of consensus criteria for describing this entity. Different criteria are used in currently published studies and range from arthralgias and polyarthritis to RA-like and classically defined RA. While some studies report the 2010 ACR/EULAR RA classification criteria, the point system is heavily biased towards seropositivity. Since most ICI-IA patients are seronegative, using the 2010 ACR/EULAR criteria could overlook individuals with bone fide seronegative RA who would benefit from the initiation of DMARD therapy. Additionally, those patients with large joint oligoarthritis would not meet criteria, particularly if their symptoms had been going on less than 6 weeks. Rather than imposing classical RA or SpA criteria, it might be beneficial to identify RA-like and SpA-like inflammatory arthritis or adopt a similar classification strategy used in JIA. A refined method of classification could help define more homogeneous patient subgroups, which is essential for understanding underlying pathophysiologic mechanisms.

Systemic clinical changes could help to improve detection of IA as an irAE and identify people at risk. Though limited, preliminary evidence suggests that screening patients for RA-associated autoantibodies or genetic factors before starting ICI treatment may identify patients at risk for IA and facilitate closer clinical monitoring. The sizes of the cohorts of patients with ICI-IA thus far are small, limiting the ability to perform mechanistic studies.¹¹⁰ While this could be due to the low frequency of ICI-IA as an irAE, it has been suggested that underreporting of IA symptoms may be a factor for why there are so few cases of ICI-IA described.¹¹⁷ In practice, milder or non-life threatening irAEs, such as arthritis, may be overlooked or omitted from reports, and examining physicians may not refer such patients to rheumatologists.^109,110,115 Diagnoses in retrospective studies are also difficult, as there is often very little information about immunology, pathology, or radiology reports.^113,117 Thus, having rheumatologists involved in patient care can help in recognition and documentation of ICI-IA.¹⁷³

Finally, translational efforts are needed focusing on autoantibodies, genetic risk, and immune dysregulation in ICI-IA. Given the similarities of ICI-IA with RA and SpA, laboratory studies performed in parallel in all three groups may best characterize pathogenesis. Since RA and SpA have been studied for years, capitalizing on existing knowledge and techniques could speed progress in defining underlying disease mechanisms in ICI-IA. Reciprocally, defining the events leading to the breach of immune tolerance in patients with ICI-IA could shed light on the early immunologic events leading to the development of RA and SpA.

Conclusions

In this review, we compared the clinical manifestations and known pathophysiology of ICI-IA with that of RA and SpA (Table 1 and Figure 2). ICI-IA shares clinical features with both RA and SpA including distribution of joint involvement, radiographic findings, and treatment response to particular DMARDs. Although most ICI-IA patients are defined as seronegative, based on RF and ACPA testing, they may have other autoantibody specificities that have not yet been discovered or the prevalence of seropositivity may be diluted by the existence of distinct clinical and mechanistic subsets within the ICI-IA moniker. As more patients with ICI-IA are recognized, a detailed definition of the immunologic, radiologic, and pathologic features present in the circulation and joints of patients with this condition is needed to dissect primordial events leading to the development of IA. Systemic comparison with canonical forms of IA in adults as well as children may shed novel insights into the development of IA more broadly and identify common mechanisms and targets for therapy.