Autoantibodies involved in primary and secondary adrenal insufficiency following treatment with immune checkpoint inhibitors

Abstract

Primary and secondary adrenal insufficiency (AI) are commonly known immune-related adverse events following treatment with immune checkpoint inhibitors (ICIs), and are clinically relevant due to their morbidity and potential mortality. For this reason, upfront identification of patients susceptible for ICI-induced AI could be a step in improving patient’s safety. Multiple studies have focused on the identification of novel biomarkers for ICI-induced AI, including autoantibodies, which may be involved in ICI-induced AI as a result of the T-cell-mediated activation of autoreactive B cells. This review highlights the currently described autoantibodies that may be involved in either primary [e.g. anti-21-hydroxylase, anti-17α-hydroxylase, anti-P450scc, anti-aromatic L-amino acid decarboxylase (AADC), anti-interferon (IFN)α and anti-IFNΩ] or secondary AI [e.g. anti-guanine nucleotide-binding protein G(olf) subunit alpha (GNAL), anti-integral membrane protein 2B (ITM2B), anti-zinc finger CCHC-type containing 8 (ZCCHC8), anti-pro-opiomelanocortin (POMC), anti-TPIT (corticotroph-specific transcription factor), anti-pituitary-specific transcriptional factor-1 (PIT-1) and others], and discusses the current evidence concerning their role as biomarker for ICI-induced AI. Standardized autoantibody measurements in patients (to be) treated with ICIs would be a clinically accessible and patient-friendly screening method to identify the patients at risk, and could change the management of ICI-induced AI.

Graphical abstract

Highlights

• •ICIs may adversely cause AI.
• •Autoantibodies are promising candidate biomarkers for ICI-induced AI.
• •Upfront risk identification for ICI-induced AI may improve clinical management.

Introduction

Immune checkpoint inhibitors (ICIs) have become a cornerstone for the treatment of multiple types of cancer.^1,2 By targeting programmed cell death-1 (PD-1), programmed death-ligand 1 (PD-L1) or cytotoxic T-lymphocyte antigen-4 receptor (CTLA-4), ICIs potentiate the immune response against tumors and may provide cancer patients with substantial clinical benefit including cure.³ However, as immune checkpoints physiologically prevent autoimmune responses, their blockade frequently induces immune-related adverse events (irAEs) as well, with grade ≥3 irAEs occurring in 6%-55% of the patients.^4,5 While many irAEs are rapidly reversed upon immunosuppression (which distinguishes them from autoimmune diseases), ICI-induced endocrine dysfunctions are predominantly irreversible and require lifelong hormone replacement therapy.^6,7

One of the most frequent endocrine irAEs is adrenal insufficiency (AI), characterized by a malfunctioning hypothalamic–pituitary–adrenal axis and a subsequent deficiency of adrenal cortisol production.^8,9 Clinical hallmarks of AI include unintentional weight loss, fatigue, headache, postural hypotension, muscle pain and hyponatremia.^10,11 Diagnosis involves assessment of hormones of the hypothalamic–pituitary–adrenal axis (Figure 1A). With regard to ICI treatment, AI either arises from a primary adrenal or a secondary pituitary disorder.¹¹ Primary AI encompasses low levels of serum cortisol and elevated levels of serum adrenocorticotropic hormone (ACTH), with the latter resulting from the absence of negative feedback by cortisol on the pituitary and hypothalamus (Figure 1B). In contrast, secondary AI is reflected by low serum cortisol and low serum ACTH levels (Figure 1C).10, 11, 12 An additional distinction involves the frequent presence of mineralocorticoid dysfunction in primary AI patients, while this is typically preserved in secondary AI patients. Standard proof of diagnosis is the ACTH stimulation test, or Synacthen test, which also discriminates between primary and secondary AI.⁹ Management of AI comprises corticosteroid replacement and prevention of life-threatening adrenal insufficiencies, also known as Addison’s crises, by providing patient education and glucocorticoid injection kits.⁸

Figure 1.

The hypothalamic–pituitary–adrenal axis and its malfunctioning during primary and secondary AI. (A) Under physiological conditions, the hypothalamus releases CRH, which stimulates the release of ACTH from cells of the pituitary. ACTH, in turn, induces the secretion of cortisol from the adrenal gland into the circulation, which subsequently exerts multiple functions on peripheral tissues. In addition, cortisol provides negative feedback on both the hypothalamus and pituitary for the release of CRH and ACTH, respectively. (B) During primary AI, the adrenal gland dysfunctions. As such, circulating levels of cortisol will be lower, leading to less negative feedback on the hypothalamus and pituitary. Consequently, the levels of CRH and ACTH will be increased. (C) During secondary AI, the pituitary gland dysfunctions, leading to lower levels of circulating ACTH. As a result, the pituitary will secrete less cortisol. Low cortisol levels will induce high CRH levels due to the absence of negative feedback from cortisol on the hypothalamus.

ACTH, adrenocorticotropic hormone; AI, adrenal insufficiency; CRH; corticotrophin-releasing hormone.

Despite optimal current treatment, patients with AI have a reduced quality of life as well as an increased mortality when compared to the general population.^10,13, 14, 15, 16 In order to minimize the adverse outcomes and prevalence of ICI-induced AI, there is a major need for novel biomarkers to upfront identify patients at risk. For irAEs in general, several potential biomarkers have already been evaluated, including autoantibodies,17, 18, 19 and cytokine levels.^20,21 In addition, tumor mutational burden,²² clonal expansion of CD8 T cells²³ and, in case of colitis, the gastrointestinal microbiome diversity were each associated with the development of irAEs.²⁴

Of these potential baseline biomarkers, autoantibodies are especially interesting, as they can be easily detected from minimally invasive blood testing, have a high specificity for the individual irAE and are likely involved in the mechanism of action of its irAE.¹⁸ Growing evidence supports the role of autoantibodies in irAEs. For example, Da Gama Duarte et al.¹⁷ described two patients treated with ipilimumab that exhibited profound increases in the autoantibody repertoire directed against self-antigens, which preceded the development of multiple high-grade irAEs, including rash, hepatitis and colitis. In line with this, Das et al.¹⁹ detected increased levels of antibody-producing plasmablasts in patients experiencing higher rates of grade 3 irAEs (colitis, hepatitis, rash, pancreatitis) 6 months after ICI treatment, which correlated with both frequency and timing of irAEs. Moreover, Gowen et al.²⁵ identified a baseline autoantibody profile that was able to identify patients who eventually developed any high-grade irAEs, including AI. On the other hand, in a study carried out by de Moel et al.,⁶ 19.2% of patients treated with ICIs developed common clinical autoantibodies, which were non-significantly associated with any irAE.

As the field of predictive autoantibodies in irAEs is still being explored, a comprehensive overview of candidate autoantibodies in ICI-induced AI is currently lacking. Interestingly, multiple organ-specific autoantibodies have already been associated with the development of either primary or secondary AI, with a subset of them being detected in relation to ICIs. In this review, we therefore evaluate available literature on autoantibodies from autoimmune AI and discuss their possible role as baseline biomarkers for the development of ICI-induced AI.

Methods

To identify all autoantibodies associated with either primary or secondary AI, literature search was carried out in the National Center for Biotechnology Information PubMed using the following search strategy: (”Autoantibodies”[Majr] OR “Autoantibodies”[Ti] OR “Autoantibody”[Ti] OR “Auto-antibodies”[Ti] OR “Auto-antibody”[Ti] OR “Autoimmune antibody”[Ti] OR “Autoimmune-antibody”[Ti]) AND (”Adrenal Insufficiency”[Majr] OR “Adrenal insufficiency”[Ti] OR “Adrenal-insufficiency”[Ti] OR “Adrenal insufficiencies”[Ti] OR “Adrenal-insufficiencies”[Ti] OR “Addison Disease”[Ti] OR “Addison's disease”[Ti] OR “Addisons disease”[Ti] OR “Hypoaldosteronism”[Ti] OR “Hypo-aldosteronism”[Ti] OR “Hypocortisolism”[Ti] OR “Hypo-cortisolism”[Ti] OR “Adrenal crisis”[Ti] OR “Adrenal crise”[Ti] OR “Hypophysitis”[Ti] OR “Hypopituitarism”[Ti] OR “Adrenal dysfunction”[Ti] OR “Pituitary dysfunction”[Ti]). On 3 November 2021, this search strategy yielded a total of 168 studies. Of these, all studies (i) written in English that were (ii) discussing autoantibodies (iii) in the context of AI were considered.

Primary AI

Prevalence and pathogenesis

Primary AI involves direct adrenal failure. Autoimmune adrenalitis, also referred to as Addison’s disease, is the most common cause of primary AI with an estimated prevalence of 1 in 10 000 persons.26, 27, 28 Addison’s disease usually presents between the ages of 20 and 50 years, is most prominent in women, and in most cases co-occurs with other autoimmune diseases, such as thyroid diseases, type 1 diabetes mellitus, premature ovarian failure and manifestations of autoimmune polyendocrine syndrome (APS) type 1 or 2.^8,11 Other general, non-autoimmune causes of primary AI include 21-hydroxylase deficiency, P450 oxidoreductase deficiency, cortisone reductase deficiency, metabolic disorders, infectious diseases, adrenal dysgenesis, adrenal hemorrhage and malignancies.^27,29

Multiple cases of patients with primary ICI-induced AI, sometimes presenting as part of APS, have been described in the literature.30, 31, 32, 33 A recent meta-analysis of multiple clinical trials carried out by de Filette et al.³⁴ showed that 5.2% [95% confidence interval (CI) 2.9% to 9.2%] to 7.6% (95% CI 1.2% to 36.8%) of the patients receiving ipilimumab with nivolumab or pembrolizumab, respectively, developed primary AI. On ICI monotherapy, primary AI appeared to occur less frequently, with predicted incidences of 1.4% (95% CI 0.9% to 2.2%) on ipilimumab, 2.0% (95% CI 0.9% to 4.3%) on nivolumab and 0.8% (95% CI 0.3% to 2.0%) on pembrolizumab monotherapy. Another meta-analysis reported a slightly lower incidence of 4.2% in patients treated with anti-PD-(L)1 plus anti-CTLA-4 and 0.7% in the overall study population.³⁵ However, the proportion of primary ICI-induced AI might be overestimated in both meta-analyses, as in clinical trials adrenal insufficiencies may have been mentioned as a toxicity without the distinction between a primary and secondary origin. This is also reflected by Grouthier et al.,³⁶ reporting on 451 VigiBase records of primary ICI-induced AI of which only 45 patients were classified as ‘definite primary AI’.

It is noteworthy that no consensus has been reached yet considering the mechanism via which primary ICI-induced AI may develop (Figure 2). Most evidence, however, supports the idea that the non-specific mode of action of ICIs lead to the (re)activation of otherwise dormant autoreactive T cells, thereby introducing a break in peripheral T-cell tolerance.^7,11 The (re)-activated autoreactive helper CD4+ and cytotoxic CD8+ T cells are thought to subsequently infiltrate and destroy the adrenal cortex.¹¹ These autoreactive T cells may be pre-existent before the onset of cancer (Figure 2A), but could also have been induced by similar antigens on both the tumor and the affected organ, resulting in cross-reaction (Figure 2B).⁷ Examples of the latter have already been reported for other irAEs.^37,38 Within both mechanisms, there is also a role for autoantibodies, as the break in T-cell tolerance may result in the (re)activation of autoreactive B cells, which in turn could produce autoantibodies.^7,11 Interestingly, since autoreactive T and B cells are frequently present already before the start of ICI treatment, autoantibodies associated with AI may be detectable before the start of ICI treatment as well and could possibly serve as baseline biomarkers. Our search identified six specific autoantibodies that have been associated to varying evidence with primary AI and will be discussed first (Table 1). Subsequently, we will discuss autoantibodies that have been associated with secondary AI, and their potential clinical implementation as biomarkers.

Figure 2.

Proposed mechanisms of ICI-induced AI. (A) Upon ICI treatment, pre-existing (dormant) autoreactive T cells may get activated and initiate an autoimmune reaction in the adrenal or pituitary gland. Additionally, they could activate pre-existing autoreactive B cells, thereby leading to the production of autoantibodies that could aid in the development of AI as well. (B) Cross-reactive T and B cells, which may result from the ectopic expression of tissue antigens by the tumor cells, are activated following ICI treatment and may both attack the tumor cells as well as the adrenal or pituitary gland. (C) Due to the ectopic, non-canonical expression of CTLA-4 on pituitary cells, these cells become the site of complement activation upon treatment with anti-CTLA-4 antibodies. This subsequently leads to the activation of the classical pathway of complement activation and antibody-dependent cellular cytotoxicity, which could result in secondary AI.

AI, adrenal insufficiency; CTLA-4, cytotoxic T-lymphocyte antigen-4 receptor; ICI, immune checkpoint inhibitor.

Table 1.

Autoantibodies associated with (ICI-induced) AI

Autoantibodies	Reported in patients with ICI-induced AI
Primary AI
Anti-21-hydroxylase	Yes32,33
Anti-17α-hydroxylase	No
Anti-P450scc	No
Anti-AADC	No
Anti-IFNα	No
Anti-IFNΩ	No
Secondary AI
Anti-GNAL	Yes18
Anti-ITM2B	Yes18
Anti-ZCCHC8	Yes39
Anti-POMC	Yes40
Anti-TPIT	No
Anti-PIT-1	No
Anti-Rabphilin-3A	No

AADC, aromatic L-amino acid decarboxylase; AI, adrenal insufficiency; GNAL, guanine nucleotide-binding protein G(olf) subunit alpha; ICI, immune checkpoint inhibitor; IFN, interferon; ITM2B, integral membrane protein 2B; P450scc, cholesterol side-chain cleavage enzyme; PIT-1, pituitary-specific transcriptional factor-1; POMC, pro-opiomelanocortin; TPIT, corticotroph-specific transcription factor; ZCCHC8, zinc finger CCHC-type containing 8.

Autoantibodies associated with primary AI

Anti-21-hydroxylase

The most frequently described autoantibodies in the context of primary AI involve those targeting the cytochrome P450 steroidogenic enzyme 21-hydroxylase (P450c21).⁴¹ 21-Hydroxylase converts 17-hydroxyprogesterone into 11-deoxycortisol, thereby being responsible for one of the vital steps of the mineralocorticoid pathway.⁴¹ Its actions are mainly restricted to the adrenal cortex, as extra-adrenal tissues are thought to depend on other hydroxylating systems.⁴¹ For this reason, anti-21-hydroxylase autoantibodies are highly specific for primary AI.

Anti-21-hydroxylase autoantibodies are present in 64%-90% of the patients with primary AI after excluding non-autoimmune causes of primary AI, and remain relatively stable after diagnosis.42, 43, 44, 45 For this reason, it is recommended to test for anti-21-hydroxylase autoantibodies in patients with acquired primary AI.^43,46, 47, 48 Remarkably, in a study carried out by Naletto et al.,⁴⁹ 28/114 (25%) of the patients with APS that were positive for anti-21-hydroxylase autoantibodies prospectively developed Addison’s disease, suggesting that anti-21-hydroxylase autoantibodies can be present long before the disease onset.

In the context of ICI treatment, two ICI-induced AI patients have been reported so far to have been tested positive for anti-21-hydroxylase autoantibodies.^31,32 The first patient was a 55-year-old female receiving pembrolizumab (anti-PD-1) monotherapy for a metastatic choroidal melanoma, who presented with an acute Addison’s crisis requiring hospitalization. The second patient was a 60-year-old man receiving atezolizumab (anti-PD-L1) who developed APS type 2, manifesting with Addison’s disease, type 1 diabetes mellitus and hypophysitis.^31,32 Although primary ICI-induced AI patients testing negative for anti-21-hydroxylase autoantibodies have also been infrequently described, these two positive patients highlight the potential of anti-21-hydroxylase autoantibodies as predictive biomarkers for primary ICI-induced AI.⁵⁰

Anti-17α-hydroxylase

Anti-17α-hydroxylase (P450c17A1) autoantibodies target another key enzyme in the mineralocorticoid pathway, but are found less frequently in primary autoimmune-mediated AI patients than anti-21-hydroxylase autoantibodies.⁵¹ For example, Nigam et al.⁵² detected anti-17α-hydroxylase autoantibodies in 4/19 (21%) patients with idiopathic Addison’s disease. In line with these data, Chen et al.⁴² and Seissler et al.⁴⁴ reported a prevalence of anti-17α-hydroxylase autoantibodies in patients with Addison’s disease as low as 3/64 (5%) and 3/25 (12%), respectively. Interestingly, the latter two studies additionally showed that the majority of patients with anti-17α-hydroxylase autoantibodies were also positive for anti-21-hydroxylase autoantibodies, suggesting that screening for anti-17α-hydroxylase autoantibodies may be redundant to screening for anti-21-hydroxylase autoantibodies.^42,44 Considering this together with the absence of case reports related to ICIs, the value of anti-17α-hydroxylase autoantibody as a biomarker for ICI-induced AI seems to be limited.

Anti-P450scc

Cholesterol side-chain cleavage enzyme (P450scc) converts cholesterol intro pregnenolone, the first reaction in steroidogenesis.^53,54 Multiple studies have reported anti-P450scc autoantibodies in patients with primary AI.^42,44,52 None of these studies, however, included patients with primary ICI-induced AI. The general prevalence of anti-P450scc autoantibodies in isolated Addison’s disease is comparable to that of anti-17α-hydroxylase autoantibodies, ranging between 4% and 16%.^42,44,52 Of note, the majority of patients with anti-P450scc autoantibodies were also positive for anti-21-hydroxylase autoantibodies, indicating that the use of anti-P450scc autoantibodies as biomarkers for ICI-induced AI should also not have priority in future studies.^42,44

Anti-AADC, anti-IFN-α and anti-IFN-Ω

Several other autoantibodies have been associated less frequently to primary AI, and have to our knowledge not yet been reported in ICI-induced primary AI cases. For example, autoantibodies against aromatic L-amino acid decarboxylase (AADC), which are present in 51%-61% of the patients with APS type 1, are believed to be able to identify a subgroup of patients with Addison’s disease with milder disease manifestations.55, 56, 57 Furthermore, both Meager et al.⁵⁸ and Meloni et al.⁵⁹ detected autoantibodies targeting interferon-α (IFN-α) and IFN-Ω subtypes with a strikingly high prevalence of 100% (76/76 and 39/39 of the patients, respectively) in patients with APS type 1, suggesting that they may be useful as an additional diagnostic criterion for this disease. How they relate specifically to the mechanism of action in development of primary AI, however, remains unknown. Considering the broad expression of these targets, e.g. also being described in patients with life-threatening coronavirus disease 2019 infection, a high sensitivity but low specificity can be envisioned.^60,61

Secondary AI

Prevalence and pathogenesis

Secondary AI involves disorders of the pituitary and occurs more frequently than primary AI, with an estimated prevalence of 42 per 100 000 persons.^8,62,63 Causes for secondary AI include malignancies, trauma, infarction, radiotherapy, genetic conditions and, most relevant in the context of ICI treatment, autoimmune reactions.^8,11 Secondary AI may present as isolated ACTH deficiency or as a deficiency of ACTH and other pituitary hormones, generally referred to as hypophysitis.^8,11,64 Of note, there is no general consensus on the diagnostic criteria of hypophysitis and several different criteria are used in the literature. For example, hypophysitis as proposed by Nguyen et al.⁶⁵ includes (i) ACTH or thyroid-stimulating hormone (TSH) deficiency in combination with magnetic resonance (MRI) changes suggestive for hypophysitis, or (ii) ACTH and TSH deficiencies in combination with headache or fatigue in the absence of MRI findings/evaluation, while Percik et al.⁶⁶ defines hypophysitis as abnormalities in at least three pituitary axes.

As stated before, the proportion of secondary ICI-induced AI compared to primary could be a large underestimation due to imprecise definition and adverse event documentation in ICI trials. Currently, hypophysitis is believed to occur in 6.4%-10.5% of patients that receive nivolumab plus ipilimumab, as reported by meta-analyses of de Filette et al.³⁴ and Barroso-Sousa et al.³⁵ However, depending on which diagnostic criteria of hypophysitis are used, hypophysitis may not always involve ACTH deficiency, and the incidence of secondary ICI-induced AI is therefore expected to be lower than that of ICI-induced hypophysitis. Male sex and older age are proposed risk factors for the development of ICI-induced hypophysitis, although confirmation studies are needed.⁶⁷ Isolated ACTH deficiency is believed to occur in 3.2% of the patients treated with anti-PD-(L)1 plus anti-CTLA-4, as reported by Percik et al.⁶⁶

Interestingly, for both studies on hypophysitis a higher prevalence was demonstrated for anti-CTLA-4 monotherapy as compared to anti-PD-(L)1 monotherapy.^34,35 A plausible theory to explain this discrepancy may rely on recent findings by Iwama et al.,⁶⁸ who identified the non-canonical expression of CTLA-4 in TSH-, ACTH- and prolactin (PRL)-secreting cells of the pituitary. Upon the binding of anti-CTLA-4 antibodies to CTLA-4 on these cells, the classical pathway of complement activation was initiated, eventually leading to pituitary cell death via antibody-dependent cellular cytotoxicity.⁶⁸ This concept has not been described for PD-(L)1 and is supported by the fact that ipilimumab (anti-CTLA-4) is an immunoglobulin G (IgG)-1 antibody and therefore a potent activator of the classical complement pathway, while nivolumab and pembrolizumab (both anti-PD-1) are IgG-4 antibodies and less strong activators of complement.⁶⁹ Collectively, these findings suggest that the expression of CTLA-4 makes the pituitary prone to inflammatory reactions upon anti-CTLA-4 treatment, and indicates the possible existence of a novel mechanism of secondary ICI-induced AI (Figure 2C). Another theory that could hypothetically explain the difference in the occurrence of hypophysitis upon anti-CTLA-4 versus anti-PD-(L)1 involves the different effects of both therapies on T cells.⁷⁰ In general, anti-PD-(L)1 is believed to exert its functions mainly by boosting pre-existing T-cell immune responses, while anti-CTLA-4 treatment is believed to diversify the immune response by activating recent thymic emigrants.^71,72 This would mean that anti-CTLA-4 treatment could induce the generation of new pituitary-reactive T cells, while anti-PD-(L)1 may only cause secondary AI by boosting already pre-existing pituitary-reactive T cells.⁷⁰

It is noteworthy that, extensive evidence indicates the involvement of autoantibodies in the development of secondary ICI-induced AI as well.⁷³ For example, in a study carried out by Kobayashi et al.,⁷⁴ 11/17 (64.7%) of the patients with ICI-induced ACTH deficiency were positive for anti-pituitary antibodies (APAs) at baseline, whereas this was only the case for 1/40 (2.5%) of the patients without pituitary irAEs, thereby supporting the mechanisms explained in Figure 2A and B. In addition, 3/4 (75%) patients with ICI-induced hypophysitis developed APAs under ICI treatment while being negative at baseline.⁷⁴ In line with this, Iwama et al.⁶⁸ reported APAs to be present in 7/7 (100%) of the patients developing hypophysitis following treatment with ipilimumab. Moreover, APAs were found in patients who developed pituitary dysfunction following atezolizumab and ipilimumab plus nivolumab treatment, and were additionally found to predict the presence of pituitary-directed autoimmune reactions.^31,75,76 In total, seven specific APAs have been associated with the development of secondary AI, of which four have been described in the context of ICI treatment (Table 1).

Autoantibodies associated with secondary AI

Anti-GNAL

Perhaps the currently most relevant autoantibodies associated with secondary AI following ICI treatment involve those targeting guanine nucleotide-binding protein G(olf) subunit alpha (GNAL). Anti-GNAL autoantibodies are hypothesized to interfere with the GNAL-induced cyclic adenosine monophosphate-signaling pathway, which is, for example, involved in the production of TSH.77, 78, 79, 80, 81, 82 In a recent study carried out by Tahir et al.,¹⁸ anti-GNAL autoantibodies were detected at significantly higher levels in pre- and post-treatment plasma of eight ICI-induced AI cases as compared with 21 controls who did not develop hypophysitis. The autoantibody levels were reported to increase 1.7-fold after treatment as compared to pre-treatment samples. These findings highlight the potential use of anti-GNAL autoantibodies as both baseline as well as on-treatment biomarker for ICI-induced hypophysitis, but need to be confirmed in larger cohorts.¹⁸

Anti-ITM2B

Akin to GNAL, integral membrane protein 2B (ITM2B) is involved in pituitary hormone production.¹⁸ By eliminating the inhibitory effect of guanylate cyclase, ITM2B is believed to indirectly stimulate the release of ACTH.^77,83,84 However, in contrast to anti-GNAL autoantibodies, anti-ITM2B autoantibodies were not significantly enriched in pre- and post-treatment plasma of patients with hypophysitis compared to controls, as reported by Tahir et al.¹⁸ Still, in patients with hypophysitis, Tahir et al.¹⁸ found a 2.5-fold median increase in anti-ITM2B autoantibodies in post-treatment samples as compared to pre-treatment samples, which was significantly higher than in patients without hypophysitis. These findings suggest that anti-ITM2B autoantibodies could serve as an on-treatment biomarker for the development of ICI-induced hypophysitis, while the predictive baseline value of this biomarker is still unclear.

Anti-ZCCHC8

Zinc finger CCHC-type containing 8 (ZCCHC8) plays a role in multiple cellular processes, including telomerase RNA maturation and RNA processing/degradation, as well as in several diseases, including familial pulmonary fibrosis and congenital glioblastoma multiforme.85, 86, 87 Interestingly, autoantibodies targeting ZCCHC8 were recently found in two ICI-induced hypophysitis cases.³⁹ One of these patients received ipilimumab, while the other patient received pembrolizumab. In both patients, anti-ZCCHC8 antibodies increased more than threefold after hypophysitis. As ZCCHC8 was shown to be expressed in pituitary cells, these findings suggest that anti-ZCCHC8 autoantibodies may mediate ICI-induced hypophysitis and could serve as a possible biomarker; however, their increased expression could have also been a bystander effect.

Anti-POMC

Pro-opiomelanocortin (POMC) is the precursor of ACTH as well as of several other pituitary-secreted proteins and is therefore expressed in ACTH-producing cells.⁸⁸ Interestingly, anti-POMC autoantibodies have been described in numerous patients with pituitary dysfunction and are believed to originate at least in part as a result of paraneoplastic syndrome, with, for example, large cell neuroendocrine carcinomas ectopically expressing POMC.⁸⁸ In a recent study by Kanie et al.,⁴⁰ anti-POMC autoantibodies were found in 2 out of 18 (11%) analyzed patients with isolated ACTH deficiency following treatment with anti-PD-1. Remarkably, tumors of these patients were found to ectopically express ACTH, while this was not the case in tumors of patients without anti-POMC autoantibodies. This supports the mechanism explained in Figure 2B, and illustrates the potential role of anti-POMC autoantibodies as biomarker for ICI-induced AI.

Anti-TPIT

Corticotroph-specific transcription factor (TPIT, also referred to as TBX19) plays an essential role in the terminal differentiation of all pituitary cells expressing POMC.^89,90 Consequently, germline mutations in TPIT are the major cause of congenital isolated ACTH deficiency.⁹¹ Interestingly, in a study focusing on 86 hypophysitis patients, Smith et al.⁹² detected the presence of anti-TPIT autoantibodies in 9/86 (10%) patients compared to 1/90 (1%) controls. To our knowledge, the presence of anti-TPIT autoantibodies has not yet been detected in patients with ICI-induced hypophysitis. Although the proportion of anti-TPIT-positive hypophysitis patients as reported by Smith et al.⁹² is rather small, anti-TPIT autoantibodies may serve as interesting candidate biomarkers, either baseline or on-treatment, which needs to be elucidated in larger ICI patient cohorts.

Anti-PIT-1

Similar to anti-POMC autoantibodies, autoantibodies targeting the pituitary-specific transcriptional factor-1 (PIT-1, also referred to as POU1F1) are believed to have a paraneoplastic pathogenesis involving PIT-1 producing cancers.^88,93,94 However, in contrast to POMC, PIT-1 is essential for the function of growth hormone (GH)-, PRL- and TSH-producing cells and is not expressed in ACTH-producing cells.^95,96 As such, anti-PIT-1 antibody syndrome is mainly associated with deficiencies of GH, PRL and TSH, with the production of ACTH being preserved.^88,93,94 The potential role for anti-PIT-1 autoantibodies in ICI-induced AI therefore seems limited.

Other hypophysitis-associated antibodies of undefined relevance for ACTH deficiency

Various alternative autoantibodies have been correlated with hypophysitis as well. For example, in a study carried out by Iwama et al.⁹⁷ anti-rabphilin-3A autoantibodies were present in 22/29 (76%) of the patients with lymphocytic infundibulo-neurohypophysitis, 2/18 (11.1%) of the patients with lymphocytic adenohypophysitis and 0/34 of the patients without lymphocytic hypophysitis (Table 1). However, as ACTH is produced by the adenohypophysis only, the role of anti-rabphilin-3A may be limited. Another study carried out by Chiloiro et al.⁹⁸ showed that sera of patients with autoimmune hypophysitis react with GH, superoxide dismutase (SOD2), actin beta protein (ACTB) and fructase-biphosphate aldolase A (ALDOA), with a non-significantly higher reactivity against the latter antigen compared to controls [15/18 (83%) versus 10/18 (56%), respectively]. Moreover, Smith et al.⁹² identified autoantibodies against chromodomain-helicase-DNA-binding protein 8, presynaptic cytomatrix protein, Ca²⁺-dependent secretion activator, pituitary gland-specific factor 2 (PGSF2) and neuron-specific enolase (NSE) in serum samples from patients with lymphocytic hypophysitis, but found them at similar frequencies in controls. As such, the potential of these autoantibodies as biomarkers for ICI-induced AI seems to be limited.

Clinical implications

Primary and secondary AI are frequent endocrinopathies following treatment with ICIs and are especially clinically relevant causes of long-term morbidity and mortality.^8,9 For this reason, upfront identification of patients that are susceptible for AI upon ICI treatment is of great need, mainly in the light of increasing numbers of studies testing (neo)adjuvant combinations of ICIs. Multiple studies have focused on the identification of novel biomarkers for ICI-induced AI, including autoantibodies. This review highlighted the currently described autoantibodies that may be involved in either primary or secondary AI. Most of the analyses are small-cohort studies underpinning the need for analyses in larger and homogenous cohorts (e.g. neoadjuvant, adjuvant or lactate dehydrogenase-normal cohorts).^18,39,40

In the case of solid confirmation of one or more of the above discussed antibodies, we envision several applications. Firstly, the identification of patients at risk would allow clinicians to provide tailored patient information, to increase the frequency of ACTH and cortisol testing and to possibly carry out on-treatment biomarker testing.¹⁸ These would lead to earlier detection of AI and would increase the patient safety by preventing Addison’s crises.

Secondly, the risk for developing ICI-induced AI may influence the choice of treatment in shared decision making. For instance, the individual risk might in the future play an important role whether one would advise a patient for mono- or combination (neoadjuvant) ICI treatment, besides taking into account the individual response chance.

Finally, a more extensive research on autoantibodies upon ICI treatment might lead to a better understanding of the pathogenesis of ICI-induced AI and other irAEs and to a lesser extent of ICI-induced toxicity in general. In the light of the increasing numbers of approved indications for ICIs and the increasing cure rates seen in earlier-stage treatment with ICIs, the personalization of ICI treatment based on risk factors for severe irAEs will become increasingly relevant.

In conclusion, autoantibodies are highly potential candidate biomarkers for ICI-induced AI, with multiple specific autoantibodies being found in ICI-induced AI patients already. As pre- and post-treatment biomarkers could significantly improve the management of ICI-induced AI patients, future studies should elaborate on the currently described autoantibodies, while studies focusing on other types of biomarkers as mentioned briefly in the introduction are also of profound interest. In this way, we should aim to achieve maximal clinical benefits with minimal treatment-induced morbidity in cancer patients receiving ICIs.