When Checkpoint Inhibitors Break Barriers: Mechanisms and Challenges of irAEs of the Skin, Gastrointestinal Tract, and Lung.

Introduction
Immune checkpoint inhibitors (ICIs) are first-line therapies for an expanding set of contexts in cancer treatment.1 The rapid expansion of ICI use has similarly increased the incidence of immune-related adverse events (irAEs). Consequences of irAEs can include ICI treatment interruption or discontinuation in up to 43% of patients,2 inhibition of anti-tumor immunity by immunosuppressive treatments,3 permanent organ dysfunction, and/or death. The mechanisms driving irAEs are incompletely understood; however, research has begun to elucidate the immune perturbations that underlie these organ-specific toxicities. An improved understanding of irAEs will allow clinicians to better predict which patients are at risk and effectively diagnose and treat toxicities without abrogating anti-tumor responses.
Although irAEs affect every organ system, epithelial barrier organs are among the most frequently impacted and include the skin (impacting up to 60% of patients),4 the gastrointestinal tract (up to 40%),5 and the lungs (up to 9.5%).6 These organ systems maintain tissue homeostasis and barrier function via rapid epithelial turnover, sense environmental cues, harbor microbiota, and have tissue-resident immune populations that promote tolerance while clearing pathogenic organisms.7–9 In contrast, non-barrier organs such as the heart,10 pituitary gland,11 and kidneys12 are less frequently impacted by irAEs and have no commensal microbiota, fewer antigenic exposures, and negligible resident immune populations under physiologic conditions.13 Understanding shared irAE mechanisms in barrier organs would improve the early diagnosis and treatment of these complications and potentially reduce compounded morbidity.14 Combination therapies including multiple ICIs or ICIs with other agents are commonplace and becoming more so; while this can complicate the attribution of toxicity and generalizability of ICI-specific observations, this fact serves to underscore the urgency of a thorough knowledge of the scope and pathogenesis of irAEs.

Dermatologic immunity and irAEs
The skin is in continuous contact with the external environment and represents an immune environment in constant flux. The skin barrier is comprised of layers of keratinocytes joined by tight and adherens junctions, forming a surface exposed to diverse environmental antigens and both commensal and pathogenic microbiota. These exposures contribute to the maintenance of skin as a barrier but complicate the preservation of self-tolerance.15 The skin microbiome is fluid, varies by anatomic location, and reflects the product of skin-microbe and microbe-microbe interactions influenced by multidirectional signaling between microbes, keratinocytes, immune cells.16 Keratinocytes are essential to maintaining immune homeostasis in this setting, as they maintain the physical properties of the skin surface, modulate microbial populations through the secretion of defensins, and influence immune cell activity through homeostatic chemokines and cytokines including TGF-β.17 Beneath the keratinocyte barrier is a basement membrane composed of glycoproteins and proteoglycans on a structural framework of collagen IV, which provides an anchor for the epidermis and is supported by dermal fibroblasts.18 Besides their contributions to physical structure, fibroblasts shape immune responses via cytokine secretion and direct fibroblast-immune cell interactions.19–22
In healthy skin (Figure 1A), resident immune cells are recruited and maintained by signals from keratinocytes and fibroblasts, establishing a tolerogenic milieu.15 Myeloid populations including Langerhans cells, macrophages, and dermal dendritic cells survey antigens and suppress inflammation through PD-L1 signaling,23 cytokine secretion, and FOXP3+ regulatory T cell (Treg) support.15 Keratinocytes and myeloid cells recognize pathogen- or damage-associated molecular patterns through toll-like or NOD-like receptor activation and activate tissue-resident memory T cells (TRM), innate lymphoid cells (ILCs), natural killer (NK) cells, macrophages, and mast cells.15,18,24
Although skin lacks the permanent organized secondary lymphoid structures provided by mucosa-associated lymphoid tissue (MALT), chronic inflammation can induce the formation of tertiary lymphoid structures (TLSs).25 TLS-promoting chemokines like CXCL13 and CCL21 are upregulated in response to checkpoint inhibition,26–28 though their contributions to organ irAEs, including cutaneous irAEs are incompletely understood.
Dermatologic irAEs encompass at least ten separate subtypes that share clinical, histologic, and molecular features of recognized inflammatory dermatologic diseases.29 Some speculate that skin irAEs are an unmasking of subclinical disease,30 while others have identified features unique to irAEs.31–33 For example, ICI-lichen planus, occurring in up to 17% of anti-PD-(L)1 treated patients,34 is similar to spontaneous lichen planus in terms of its band-like lymphocytic infiltrate at the dermal-epidermal junction (DEJ) and marked expression of IFN-γ.33,35 However, ICI-lichen planus lesions contain increased myeloid cells including eosinophils, lymphocyte exocytosis, spongiosis, and a distinct transcriptional signature.23,33,35,36
ICI-bullous pemphigoid (ICI-BP) frequently occurs in patients anti-PD-(L)137 and anti-CTLA-438 therapies and can involve ocular and upper respiratory mucosae.39–42 Like spontaneous bullous pemphigoid (BP), it is associated with autoantibodies to basement membrane antigens deposited in a characteristic linear arrangement43 and infiltrating lymphocytes and eosinophils (Figure 1B). Autoantibodies may develop in response to DEJ antigen expression by tumors, particularly those occurring in barrier organs such as melanoma, non-small cell lung cancer (NSCLC), and urothelial carcinoma.44 In support of this hypothesis, skin irAEs are more frequent in patients with melanoma than non-skin cancers.45 Additionally, it has been shown that cancer patients exhibit elevated levels of autoantibodies compared to healthy controls, although causal relationships between autoantibody levels, ICI treatment, and irAE development remain uncertain.46 Alternatively, early inflammation at the DEJ (as seen in ICI-lichen planus) may expose self-antigens that generate a secondary antibody response.47 Tellingly, ICI-BP can respond to B cell depletion with rituximab,43 supporting the necessity of autoantibodies for ongoing toxicity. Further, ICI-BP and several additional cutaneous irAEs (eczematous, pruritus, maculopapular) exhibiting elevated Th2 cytokines IL-4 and IL-13, have been successfully treated with the anti-IL4Rα dupilumab, echoing its efficacy in BP and atopic dermatitis.35,48,49
Although rare, ICI-associated Stevens-Johnson Syndrome and toxic epidermal necrolysis (SJS/TEN) (or the related progressive immunotherapy-related mucocutaneous eruption)50 often require treatment with systemic corticosteroids and have mortality rates up to 48%.51 While non-ICI SJS/TEN is histologically pauci-inflammatory, ICI-SJS/TEN exhibits a lymphocytic infiltrate,52 which may represent a common feature across cutaneous ICI subtypes.29,35
The mechanisms by which checkpoint inhibition generate a broad array of anatomically distinct cutaneous toxicities are an active area of current investigation. Profiling studies show that adaptive immune cellular and molecular are a defining feature of cutaneous irAEs.4,35,53–55 Recent work has shown that PD-1 is required for CD8+ T cell tolerance towards cutaneous neoantigens, suggesting a potential direct role for PD1 inhibition in the genesis of skin irAEs.23 Studies leveraging single-cell RNA sequencing (scRNA-seq) have specifically implicated expanded cytotoxic TRM T cells (particularly Th1 cells) in cutaneous irAEs.56,57 These studies highlight the broad diversity of cellular, molecular, and environmental features characteristic of cutaneous irAEs and will enable future studies that define common immune pathways implicated in other barrier organ irAEs.

Gastrointestinal immunity and irAEs
The gastrointestinal (GI) epithelium (Figure 2A) is the largest mucosal surface in the body. The first line of mucosal defense in the GI tract is a stratified layer of mucus and the epithelial glycocalyx. Underneath these layers is a monolayer of intestinal epithelial cells (IECs) connected via tight junctions.58 While most epithelial cells are absorptive, this monolayer contains subsets of differentiated secretory cells including goblet cells, which secrete antimicrobial peptides and highly glycosylated mucins that make up intestinal mucus. The colon, which has higher concentrations of luminal bacteria compared with other areas of the GI tract, has two layers of mucus, whereas the small intestine has a single, loosely attached layer of mucus that provides some barrier protection while allowing for nutrient absorption.59 For this reason, the small intestine has additional secretory cells (Paneth cells), which secrete antimicrobial peptides that limit the number of bacteria contacting intestinal epithelium. Barrier breakdown and the loss of goblet cells are common features of immune-mediated GI diseases such as inflammatory bowel disease (IBD).60
Gastrointestinal immune cells in the mucosa are found intercalated in the epithelium and throughout the underlying lamina propria tissue layer. They also exist in gut-associated lymphoid tissue (GALT), which is a subtype of MALT comprised of Peyer’s patches, lymphoid follicles, and mesenteric lymph nodes (MLNs). DCs and other antigen-presenting cells (APCs) take up luminal antigens from both M-cells, which are specialized IECs that sample antigens from the intestinal lumen, and by extending dendrites through tight junctions.61 APCs then activate naïve and memory B and T cells in GALT structures. After an acute immune response, subsets of expanded effector lymphocytes contract into long-lived TRM lymphocytes. In most epithelial organs, TRM cells express CD69 and/or CD103, which mediates homing to the epithelial adhesion molecule E-cadherin.62 Intestinal transplant studies have identified an additional population of intestinal CD8+ TRM cells that lack CD103 and instead express ITGB2, which mediates homing to inflammation-associated intercellular cell-adhesion molecules.63,64
Gastrointestinal tolerance, like skin tolerance, is an active immune state that helps prevent pathologic inflammation to commensal microorganisms and food antigens and is maintained by a variety of immune cells, including CD103+ DCs interacting with Tregs65 and tolerogenic FOXP3+ CD8+ T cells66 in the intestine. B cells additionally contribute to gastrointestinal tolerance. Patients with IBD have significant skewing of the plasma cell repertoire from IgA to IgG,67 which may directly increase myeloid cytotoxicity,68 and patients treated with B cell-depleting anti-CD20 are at risk of developing de novo IBD.69
The loss of GI tolerance is a common side effect of ICI therapy. New-onset diarrhea occurs in patients receiving anti-CTLA-4 (up to 34%), anti-PD-1 (21%), and dual anti-PD-1/CTLA-4 (45%) therapies.70 As observed in cutaneous irAEs, patients who develop colitis have increased overall survival.71 GI irAEs range in severity from mild diarrhea or nausea/vomiting to severe diarrhea or bowel perforation.72 Biopsy-proven colitis is observed in 3 – 45% of ICI-treated patients, enteritis in 1 – 15% of patients, and gastritis in less than 5% of patients.73,74 The incidence of ICI-induced colitis is not higher in colorectal cancer (CRC) patients treated with immunotherapy. In fact, trials of nivolumab plus ipilimumab in microsatellite instability-high metastatic CRC showed higher incidence of cutaneous, thyroid, and liver toxicity than colitis, despite a high rate of diarrhea in these patients.75,76 Esophageal involvement is rare, but a frequent co-occurring feature of ICI-related myositis and/or myocarditis where skeletal muscle inflammation in the upper esophagus can present with dysphagia.77 Definitive diagnosis of intestinal inflammation requires endoscopic mucosal biopsies, which is essential to high-quality patient care as up to 35% of patients with new-onset adverse GI symptoms, like diarrhea, lack histologic evidence of intestinal inflammation.78 Most cases of ICI-associated colitis present with neutrophilic infiltrates, crypt atrophy, and epithelial apoptosis on biopsy, whereas 35% of patients on anti-PD-1 therapy and less than 10% of patients on anti-CTLA-4 therapy present with microscopic lymphocytic colitis marked by increased intraepithelial lymphocytes with minimal epithelial surface injury (Figure 2B).79,80 Approximately 10% of patients with new-onset GI symptoms have isolated upper GI tract involvement including a gluten-induced enteropathy akin to celiac disease.78 Similar to cutaneous irAEs, GI irAEs have shared and distinct immunologic features compared with spontaneous autoimmune intestinal diseases.

Lung immunity and irAEs
The pulmonary immune system monitors environmental antigens from thousands of liters of air daily and clears pathogens while leaving gas exchange intact (Figure 3A).81 Respiratory epithelial cells reside beneath a protective mucous layer enriched in mucin glycoproteins, immunoglobulins, and antimicrobial peptides, which trap potentially harmful microbes. Larger particles are trapped in the trachea, bronchi, and bronchioles and are cleared by a ‘mucociliary escalator’ while particles < 2 μm are deposited in alveoli.81,82 Alveoli are lined by specialized epithelial cells that contribute to host defense through the secretion of cytokines and chemokines and by directly interacting with alveolar macrophages.
Like the skin and intestine, the respiratory barrier is surveyed by resident immune cells including tissue-resident macrophages, γδ T cells, ILCs, and DCs.81,83 Lower respiratory macrophages include alveolar and interstitial subsets.84 In inflammatory conditions, airway epithelial cells both initiate and sustain immune responses by secreting cytokines and chemokines and function as APCs through the upregulation of MHC molecules.81,85 Like the skin and intestine, the lungs contain commensal bacteria that interact with lung-resident immune cells and contribute to respiratory immunity.82,83
Pneumonitis occurs in up to 9.5% of ICI-treated patients6 with an associated mortality rate of 12 – 23%,86,87 making it one of the most lethal irAEs.88 ICI-induced pneumonitis is more common with immunotherapy regimens targeting PD-(L)1 than anti-CTLA-4 monotherapy.89 Furthermore, patients with NSCLC had a higher incidence of pneumonitis (4.1%) versus those with melanoma (1.6%).90 A retrospective study of NSCLC patients showed a higher incidence (19%) of pneumonitis than the rate observed in clinical trials, suggesting that the real world incidence may be higher.91 Several patient-specific factors are linked to ICI-pneumonitis including age, male gender, preceding thoracic radiation, pre-existing interstitial lung disease, and smoking history.92–94
Unlike cutaneous and intestinal irAEs, pulmonary irAEs are not diagnosed by direct tissue observation or histology, but rather by imaging characteristics and the exclusion of other pulmonary diagnoses including alveolar hemorrhage, infection, and/or tumor progression.88 Radiographic patterns associated with ICI-pneumonitis include cryptogenic organizing pneumonia (most commonly observed), hypersensitivity pneumonitis, acute interstitial pneumonia, and nonspecific interstitial pneumonia patterns.95In contrast to radiation therapy-related pneumonitis, ICI-pneumonitis is typically bilateral.96 Furthermore, ICI-related lung toxicity can present as pleural effusions, exacerbations of obstructive lung disease, and the development of sarcoid-like granulomatosis.88
The histopathologic findings in ICI-induced pneumonitis are nonspecific and defined by combinations of interstitial thickening, lymphocytic inflammation, pneumocyte desquamation, intra-alveolar fibrin, and foamy macrophages.97 As lung biopsies are not typically performed to work-up ICI-pneumonitis, little is known about the lung parenchymal cell types associated with ICI-mediated lung toxicity. However, several recent scRNA-seq studies highlight key immune features of bronchoalveolar lavage (BAL) fluid characteristic of ICI-pneumonitis. These include expended Tregs, CXCL13-expressing T cells, proinflammatory monocytes, and depletion of alveolar macrophages (Figure 3B).98–101
The spectrum of severity of ICI-pneumonitis ranges from asymptomatic cases diagnosed incidentally on imaging studies to fatal respiratory failure.102 Typical symptoms of ICI-pneumonitis include cough, dyspnea, and decreased exercise tolerance.88,102 Unlike skin and intestinal ICI toxicities, ICI-induced pneumonitis is not clearly associated with patient survival. Furthermore, in patients with NSCLC, ICI-pneumonitis is associated with decreased overall survival, which is reflected in high ICI-pneumonitis mortality rates estimated at 15 – 20%.86,103 A higher grade of ICI-pneumonitis, supplemental oxygen requirement, and ICI discontinuation are all associated with an increased risk of death in patients with ICI-pneumonitis.104 Unlike skin and intestinal ICI toxicities, clinical guidelines for ICI-pneumonitis recommend early treatment with high dose corticosteroids without the prioritization of steroid-sparing or -minimizing therapy.105,106 Further work will need to clarify if the high mortality associated with ICI-pneumonitis is directly related to end organ damage, high immunosuppressant exposure, or some combination of these factors.

Epithelial barrier organs rarely impacted by irAEs
Like the skin, GI tract, and lungs, the genitourinary, nasal, and ocular mucosae have significant numbers of resident immune cells and commensal microbiota, yet are rarely impacted by irAEs. ICI ureteritis and cystitis occur in less than 1% of patients and can cause urinary urgency, sterile pyuria, and hematuria.107 Ophthalmic irAEs occur in 1% of ICI-treated patients and can present with decreased visual acuity, diplopia, and tearing.108 Why some barrier organs are less frequently impacted by irAEs, despite sharing anatomic and immunologic features of skin, intestine, and lung, is an important area of study that will shed light on irAE pathogenesis.

Immune mechanisms
Although ICI-associated skin, intestinal, and lung toxicities vary in their presentation, the immunologic perturbations defining these barrier organ toxicities show similarities suggesting shared mechanisms. These include the activation and expansion of tissue-specific TRM cells, upregulation of inflammatory cytokines, generation of autoantibodies, dysbiosis, and toxicity modulation by androgen activity and oxygen tension (Table 1).

T-cell populations
TRM cells make up a majority of T cells in most epithelial organs and play crucial roles in tissue homeostasis, recall responses to pathogens, and autoinflammatory conditions.109,110 They are abundant in the skin, intestine, and lung, reflecting diverse antigen exposures in these organs.111 TRM T cells are identified by the canonical markers CD69 +/− ITGAE/CD103 and can express high levels of the T cell co-inhibitor receptors PD-1, TIM-3, LAG-3, and CTLA-4, suggesting that immune checkpoints may constrain inflammatory TRM cell responses.62,112 In cancer, TRM cells participate in anti-tumor responses to immunotherapy112–114 and their migration into the periphery is a positive prognostic indicator for ICI efficacy.114,115 TRM cell expansion is also emerging as a characteristic feature of barrier organ irAEs. This expansion has been observed in multiple subtypes of ICI-dermatitis.57 In the intestine, several recent scRNA-seq studies show that CD8+ TRM expressing cytotoxic genes (e.g. IFNG, GZMB) are expanded in the colonic mucosa of patients with ICI-colitis and may directly contribute to epithelial barrier breakdown and absorptive defects.116–119 Treg immunosuppressive activity depends, in part, on immune checkpoint pathways. For example, CTLA-4 is constitutively expressed in most FOXP3+ Tregs55 and is required for Treg immunosuppressive functions in specific contexts.120,121 Furthermore, patients with CTLA4 haploinsufficiency can develop variable, tissue-specific immune infiltrates leading to increased risk of infections and autoinflammatory thyroiditis, psoriasis, autoimmune hemolytic anemia, thrombocytopenia, and/or enteropathy.122
Treg transcriptional signatures in irAEs vary considerably by organ system. Compared to lichenoid keratosis, ICI-lichen planus lesions are relatively depleted of Tregs.123 In the lungs, diminished Treg immunosuppressive activity is implicated in the pathogenesis of ICI-pneumonitis.124 Analysis of BAL fluid from ICI-pneumonitis showed reduced PD-1 and CTLA-4 expression in Tregs, suggesting impaired immunoregulation.125,126 Treg inhibition is also implicated in ICI-colitis, which demonstrates expanded colon mucosal Tregs that upregulate Th1 genes including TBX21, CXCR3, and IL12RB1/2.116,118 Th1-like Tregs have also been observed in cancer where they suppress anti-tumor immunity.127 Further mechanistic work is required to define the specific contribution of Th1-like Tregs to intestinal irAEs.
T cell receptor (TCR) analysis showing shared TCRs in tumor and irAE tissue has hinted at the possibility of shared antigens driving anti-tumor immune responses and barrier tissue irAEs. The strongest data supporting a role for shared T cell antigens is the observation of shared CD8 T cell clones in melanoma and vitiligo, which is an autoimmune skin depigmentation condition associated with improved melanoma responses and exacerbated by ICI therapy.128 Mechanistic studies have specifically identified several melanoma/melanocyte shared antigens recognized by CD8 T cells including Melan-A, premelanosome, and tyrosinase.129 Patients with NSCLC had shared CD4 and CD8 TCRβ clones in lung tumor tissue and irAE skin lesions,130 raising the possibility of a shared antigen driving T cell responses in both organs. However, the clinical significance of shared TCRs is unclear, especially as this study lacked controls from non-irAE tissues to assess for irAE-specific enrichment of tumor TCRs. In irAE colitis, intestinal CD8 TRM T cells show significantly increased TCR diversity compared to healthy controls,116 suggesting that intestinal irAEs are unlikely to be driven by a dominant CD8 T cell antigen. In contrast, BAL CD4 T cells from patients with ICI-pneumonitis show clonal expansion, which may reflect antigen-drive T cell activation.100

Myeloid cell populations
Several studies implicate myeloid cells in irAE pathogenesis. For example, eosinophils and eosinophil homeostatic cytokine IL-5 are elevated in some skin irAEs.35 Myeloid expansion is also present in murine checkpoint inhibitor-induced dermatitis and in ICI-lichen planus.23 Granulocytes are particularly enriched in ICI-urticaria,35 ICI-lichen planus,33 ICI-BP, and ICI-psoriasis.35 In ICI-colitis and pneumonitis, there is marked expansion of activated macrophages and monocytes that express the interferon-stimulated chemokines CXCL9/10/11.98,99,116,118 As in tumors, these chemokines likely contribute to effector T cell activation and recruitment through binding CXCR3.
The efficacy of the integrin inhibitor vedolizumab in treating irAE colitis supports a causative role for myeloid cells in intestinal irAEs. In patients with IBD, vedolizumab functions primarily by limiting the recruitment of blood mononuclear phagocytes including CD1c+ DCs into intestinal tissue (bioRxiv 2023.01.21.525036; ref131). Vedolizumab is highly effective in irAE colitis, leading to clinical remission in 90% of patients132,133 and has lower colitis recurrence rates than anti-TNF therapy. If vedolizumab has a myeloid-directed effect in ICI-colitis, it has yet to be established.
In tumor immunology, myeloid-derived suppressor cells (MDSCs) mediate resistance to anti-tumor immunity, including anti-PD-1 immunotherapy by inhibiting T effector activation.134 To date, it is unclear if MSDCs also constrain organ-specific irAEs. Newer immunotherapies such as those targeting CSF1R specifically re-program macrophages to enhanced anti-PD-1 tumor immunity.135 The use of these and other combination treatments incorporating myeloid therapies promise to elucidate how myeloid cells contribute to irAEs.

Interferons
Interferons (IFNs) represent a broad class of cytokines with varied functions in diverse inflammatory processes including antiviral responses, autoimmunity, and tumor immunity136 and are divided into three types: type I (IFN-α/β), type II (IFN-γ), and type III (IFN-λ). IFNs mediate their effects canonically via the JAK/STAT signaling pathway,136 and both type I and type II IFNs have been implicated in irAE pathogenesis.
Elevated type I interferons in the periphery are associated with irAE colitis, dermatitis, and pneumonitis.137,138 Elevated serum IFN-γ after ICI is associated with irAE development and improved anti-tumor responses.139 High IFN-γ levels were found to be expressed by CD8+ TRM and effector CD4+ T cells in ICI-colitis.116,117 Multiple murine models of irAEs, including ICI-like pneumonitis, colitis, and ICI-lichen planus, also show higher IFN-γ levels, supporting its potential role in barrier organ irAE pathogenesis.23,140 This evidence suggests a putative role for JAK inhibitors in irAE treatment; small case series support using the JAK inhibitors tofacitinib and ruxolitinib in ICI-colitis and dermatitis respectively.117,141,142

Interleukins

IL-6
IL-6 is a pro-inflammatory cytokine that promotes CD8+ T cell cytotoxicity and CD4+ T cell differentiation into Th17 cells, inhibits Treg differentiation, and induces B cell differentiation into immunoglobulin-secreting plasma cells.143 While IL-6 contributes to the pathogenesis of multiple autoimmune diseases like rheumatoid arthritis, it has mixed roles in the tumor microenvironment where it promotes tumor growth, but has variable effects on cancer immunotherapy responses.144 IL-6 is actively upregulated in multiple irAEs including colitis,145 cutaneous psoriasiform reactions,146 and pneumonitis, where IL-6 levels correlate with irAE severity.147,148 Mice treated with anti-PD-1 had higher IL-6 levels and accelerated development of psoriasiform dermatitis, which improved with an anti-IL-6R antibody.32 Systemic IL-6 release was noted following anti-CTLA-4 treatment in mice, and IL-6 blockade reduced intestinal damage while enhancing antitumor immunity.149 The upregulation of IL-6 in ICI-associated skin, GI tract, and lung toxicities suggests a role for anti-IL-6R therapy (e.g. tocilizumab) in irAE treatment.32,148 Trials of tocilizumab have demonstrated efficacy in treating ICI-associated pneumonitis, hypophysitis, colitis, pancreatitis, hepatitis, and dermatomyositis.150–152

IL-17
Multiple lines of evidence support a role for IL-17 in irAE pathogenesis. Several diseases impacting barrier organs are thought to arise, in part, from aberrant IL-17-driven epithelial tissue repair.153 IL-17 signaling has a well-established role in psoriasis and psoriasiform reactions to ICI.154,155 Multiple studies show that serum IL-17 increases after ICI administration.156,157 In a PDCD1–/– murine model of psoriasis, IL-17A production by skin T cells increased proportionally with epidermal hyperplasia and neutrophil infiltration.154,158 IL-17 inhibitors are effective for both psoriasis and psoriasiform reactions to ICIs.159,160 A cautionary note on anti-IL-17 therapy was raised by a case in which a patient whose irAEs responded to anti-IL-17 simultaneously exhibited tumor progression.160
IL-17 levels are increased in the intestine and blood of patients with ICI-colitis, though unlike in cutaneous irAEs, it is unclear if IL-17 promotes or restrains pathologic inflammation. Mucosal IL-17A expression is elevated in patients with ICI-colitis and peripheral IL-17 levels mirror colonic mucosal inflammatory activity in both ICI-colitis and IBD.161,162 IL-17A upregulation in ICI-colitis is detected in activated CD4+ T effector cells and a subset of CD103/ITGAE CD8+ TRM, akin to Tc17 cells described in other tissues.116 Unlike in psoriasis, therapeutic IL-17 inhibition can exacerbate and cause de novo IBD.163 To date, it is unclear if the role of IL-17 in ICI-colitis and IBD pathogenesis is the same. Importantly, one study reported that two patients with myocarditis, colitis, and skin rash had resolution of these adverse events after anti-IL-17 treatment.164 Both patients, however, had been given high-dose steroids and thus, it was unclear to what extent irAE improvement was related to corticosteroids and/or IL-17 blockade. In humans, the immunoregulatory IL-10 family member, IL-26, is strongly associated with the Th17 transcriptional program and is frequently co-expressed with IL17A. A recent publication showed that IL-26 is upregulated in IL17A-producing CD8 TRM cells in ICI-colitis116 (Figure 2B, Table 1). IL-26 is also strongly upregulated in patients with ulcerative colitis where it likely plays an immunoregulatory role akin to IL-10.165
BAL fluid from patients with ICI-induced pneumonitis showed significant Th17 cell expansion,166 and single-cell transcriptomics identified increased T cells with features of both Th1 and Th17 cells in the BAL fluid of patients with ICI-pneumonitis.99 A separate group found that IL-17A levels correlated with ICI-pneumonitis severity,167 suggesting (as in ICI-colitis) that IL-17A could serve as a biomarker of irAE severity.

IL-23
IL-23 is an IL-12 cytokine family member that promotes the differentiation of IL-17-producing T cells; it also drives distinct features beyond the scope of IL-17-specific inflammation.168 Upon epithelial barrier injury, IL-23 is upregulated in tissue-resident myeloid cells and induces the expansion of Tc17 cells, γδ T cells, NK cells, mucosal-associated invariant T cells, and ILC3s (IL-17 secreting innate lymphoid cells).168 IL-23 plays a pathologic role in psoriasis and IBD, and can be therapeutically inhibited to treat these diseases.169,170 Pathologic IL-23-related inflammation is likely constrained, in part, by checkpoint inhibitor receptors like PD-1.171 Supporting this idea, a patient with genetic PDCD1 (PD-1 deficiency) developed IL-23-driven thyroiditis, autoimmune diabetes, arthritis, and pneumonitis, phenocopying features of irAEs occurring after pharmacologic suppression of this axis.171 Monocytes expressing IL-23, IL-6, and TNF, were enriched in the BAL fluid of patients with ICI-pneumonitis.99 Similarly, a mouse model of ICI-colitis demonstrated increased IL-23 pathway enrichment in CD4+ T cells, and blockade of IL-23 suppressed IFN-γ-producing colon CD4+ T cells and attenuated ICI-colitis.172 Both IL-12/23 inhibition (ustekinumab)173 and IL-23-specific inhibition (risankizumab) have been used to successfully treat ICI-colitis.174 Ustekinumab and the specific IL-23 inhibitors guselkumab and risankizumab have been effective in the treatment of ICI-psoriasis.175–177

Microbiome
The human microbiome engages in reciprocal regulation of diverse disease processes including autoimmunity, oncogenesis, response to cancer immunotherapies and irAEs.178 Increased abundance of Streptococcus, Fusicatenibacter, and Stenotrophomonas species in the stool microbiome was associated with severe irAEs in anti-PD-1 treated patients.179 In contrast, patients with increased Faecalibacterium and Lachnospiraceae developed mild or no irAEs.179 In other studies, patients with enriched fecal Bacteroidetes phyla at baseline were relatively protected against ICI-colitis180 while increased Faecalibacterium was associated with increased ICI-colitis risk, improved ICI response, and better overall survival.181 Further, fecal microbiota transplant has been used to successfully treat refractory ICI-colitis,182 suggesting that dysbiosis may play an important role in intestinal irAEs. However, since treated patients in this study were also exposed to immunosuppressive biologics, their clinical improvement could have represented a delayed anti-TNF or anti-integrin therapy response.
Although the majority of irAE microbiome studies have focused on the gut microbiome, the skin microbiome may also contribute to irAE development.183,184 Anti-CTLA-4 treated mice colonized with S. epidermidis developed dermatitis characterized by increased IL-17 and IFN-γ, mediators implicated in the development of psoriasisiform154 and lichen planus185,186 irAEs respectively. Given the potential role of microbiota in irAE pathogenesis, work is underway to determine if skin and gastrointestinal microbiota could be modulated to treat irAEs.181,187
In the lung, elevated relative abundance of the Proteobacteria and Firmicutes phyla in the lower respiratory tract was associated with an increased risk of ICI-pneumonitis.188 This observation is consistent with findings that the lung microbiome is associated with overall prognosis of ICI-treated patients with lung cancer.189

Autoantibodies
While ICI therapy works primarily through its effect on T cells, disruption of the PD-(L)1 or CTLA-4 pathways can lead to autoantibody-mediated damage of healthy tissue.190 In the skin, autoantibodies targeting the hemidesmosomal glycoproteins BP180 and BP230 and epidermal basement membrane protein LAD-1 are implicated in ICI-associated bullous pemphigoid (BP).191 Immunofluorescence studies demonstrate linear IgG and C3 deposition at the dermal-epidermal junction in ICI-associated BP, similar to non-ICI-associated BP, suggesting that autoantibodies against basement membrane proteins mediate this toxicity.191 In support of this idea, B cell-depleting anti-CD20 antibodies have been used to treat ICI-associated BP.192 In the intestine, ICI-celiac disease is a rare cause of ICI-related enteropathy defined by the presence of duodenal intraepithelial lymphocytes, villous atrophy, and elevated titers of IgA autoantibodies to tissue transglutaminase (TTG-IgA), which is the dominant autoantigen in spontaneous celiac disease.78 In a study of patients with Hodgkin Lymphoma who were treated with chemotherapy and ICIs, anti-PD-1 was associated with a decrease in several autoantibodies while anti-CTLA-4 was associated with a decrease in several autoantibodies while anti-CTLA-4 was associated with increased autoantibodies46 including those targeting matrix components like collagen and TTG. This suggests that CTLA-4 may have a greater impact than PD-1 on constraining autoantibody production. The role of autoantibodies in ICI-driven pulmonary toxicities is less well-defined. In one study, compared to patients without ICI-pneumonitis, those who developed pneumonitis had higher baseline levels of anti-CD74 which increased further after ICI treatment.193 Further work is required to determine if ICI-driven expansion of autoantibodies is a primary driver versus the result of ICI-mediated barrier damage.

Oxygen
Physiologic oxygen tension varies significantly across epithelial barrier organs and shapes immunity. In contrast to the lung, the skin and intestinal tract are relatively hypoxic.194,195 In tumors, hypoxia is likely a major driver of T cell exhaustion which is mediated, in part, through the upregulation of hypoxia-inducible factors (HIFs) that drive cellular adaptation to low oxygen environments.196 HIFs have a complex role in oncogenesis and tumor immunity, including ICI therapy responses.197 Further, HIFs are strongly upregulated in spontaneous autoimmunity. In IBD, hypoxia inducible factor (HIF)-1α stabilization is important for barrier function and adaptation to low oxygen tension.198 Similarly, ICI-colitis is marked by the epithelial upregulation of genes related to hypoxia and angiogenesis including HIF1A and VEGFA.116 In a human CTLA4-expressing murine model, targeting HIF-1α abrogated ipilimumab-mediated colitis.199 Hypoxemia is a known feature of ICI-pneumonitis,200 but it is unclear what effect tissue oxygenation has on pneumonitis development or treatment response. Similarly, the role of hypoxia is not known in cutaneous irAEs. Whether oxygen shapes tissue-resident memory immune populations in barrier organs and can blunt cytotoxic memory CD8+ T cell responses in irAEs as in the tumor microenvironment201 will be a rich area of future study.

Androgens
The basis of sex-specific immune differences is incompletely understood. Studies implicate the interplay of sex hormones, genetic risk factors, and the microbiome,202 which leads to higher rates of autoimmune disease and improved vaccine-induced humoral immunity in females.203,204 One explanation for this finding may be related to differing androgen levels in males and females. Androgens dampen CD8+ T cell responses to cancer205,206 and shape immune responses in the skin, gastrointestinal tract, and lung.207 Several meta-analyses show that males have increased clinical benefit from ICIs.208,209 In contrast, irAEs are more prevalent in females,210–212 which may lead to more frequent ICI disruptions and immunosuppression. While several studies support a role for sex differences in cutaneous and gastrointestinal irAEs,210–212 there are conflicting findings in ICI-pneumonitis.213,214 Sex-specific irAE burden may partially explain the apparent discrepancy between androgen-mediated immunosuppression and superior outcomes in ICI-treated male patients.

Conclusion
While barrier organ irAEs lead to significant patient morbidity, they also present an opportunity to better understand the broader immune impacts of checkpoint inhibition. Barrier organs give rise to the cancers most frequently treated with ICI therapy and the efficacy of ICIs in treating these cancers relies on tissue-intrinsic immune responses. These same tissue-specific immune pathways drive barrier irAEs potentially explaining why patients with irAEs generally have improved cancer prognosis. This intrinsic link between anti-tumor immunity and off-target toxicities raises the significance of identifying immune pathways and patient-specific factors that decouple ICI toxicity and efficacy in barrier organs.
Since ICI therapy was first FDA-approved in 2011, there have been significant advances in ameliorating ICI barrier toxicities, largely through empiric immune modulation. These therapies often target single organ toxicities despite growing awareness that over 70% of irAEs co-occur across multiple organs,215 almost always involving barrier tissues. To date, it is unclear if toxicities that impact multiple barrier organs are mechanistically distinct from single organ toxicities and may benefit from specifically tailored treatments. Notably, therapies that effectively treat multiple barrier irAEs, such as high-dose corticosteroid and anti-TNF, risk impacting anti-tumor efficacy. More specific immunomodulator therapy, such as IL-23 inhibition for ICI-psoriasis, IL-4/13 inhibition for ICI-eczema or ICI-BP, and anti-integrin therapy for colitis allows for highly effective toxicity management while potentially preserving antitumor immunity. Expanding this strategy across the spectrum of irAEs will require a more thorough understanding of pathogenesis. Barrier organ toxicities represent an opportunity to reconceptualize ICI therapy responses as systemic processes that simultaneously impact multiple tissues. Incorporating this holistic understanding of checkpoint inhibition will inform the development of safer and more effective cancer immunotherapies.