Hypophysitis induced by immune checkpoint inhibitors: a 10-year assessment

Abstract

Introduction:

Hypophysitis caused by immune checkpoint inhibitors (ICIs) has risen to the medical attention during the past decade. ICIs are monoclonal antibodies that block the interaction between molecules that normally inhibit the function of effector T cells, ultimately increasing their ability to destroy cancer cells but also causing immune-related adverse events, such as hypophysitis. Ipilimumab, a CTLA-4 blocker, was the first ICI approved from the Food and Drug Administration for advanced melanoma patients in 2011. Several additional ICIs targeting CTLA-4, PD-1, or PD-L1 are now used in many clinical trials, making it important for physicians to recognize and treat hypophysitis adequately.

Areas covered:

This review will provide insights into the mechanisms of pituitary toxicity, highlight the complexity of clinical phenotypes of ICI hypophysitis, and offer practical recommendations.

Expert opinion:

ICI hypophysitis differs in many respects from primary hypophysitis, and also according to the type of ICI that caused it. Its pathogenesis remains unknown, although the expression of CTLA-4 and PD-1 on pituitary cells could play a role. The diagnosis is mainly clinical since there are no specific serological markers and MRI findings are subtle. The treatment is based on long-term hormone replacement and does not typically require discontinuation of immunotherapy

1. Introduction: definitions and historical overview

Primary autoimmune hypophysitis can be defined as a chronic inflammation of the pituitary gland caused by overly active lymphocytes that ultimately mediate functional and pathological damage [1,2]. Similarly to other, more common, autoimmune diseases, this definition emphasizes the pathogenic role of autoreactive T and/or B lymphocytes, which are capable of recognizing self-antigens in the absence of infection and damaging the tissue [3]. Our initial 2005 review of primary hypophysitis identified 379 published cases (Table 2 in [1]), which grew to 1,005 in our latest 2016 analysis (Table 1 in [4]). During the past three years, 337 additional patients have been published, bringing the total primary hypophysitis patients summarized in this article to 1,342.

The definition of hypophysitis secondary to immune checkpoint inhibitors (ICIs) is much laxer because pathological data about the pituitary gland are missing. Of all the published ICI hypophysitis cases, in fact, the pituitary was analyzed pathologically only in a patient who died and consented autopsy [4]. In addition, ICI hypophysitis does not cause alterations that can be readily detected by MRI, making its diagnosis mainly a clinical one. The definition of hypophysitis secondary to ICI, therefore, is that of a functional deficit in one or more pituitary axes, possibly accompanied by subtle MRI abnormalities, that develops in a cancer patient treated with ICI. To aid the definition and diagnosis of ICI hypophysitis, a recent imaging study used machine learning to distinguish ICI hypophysitis (No. = 60 published cases) from pituitary metastases (No. = 62 published cases) based on a set of three clinical features (headache, hypopituitarism, and diabetes insipidus) and five MRI features (pituitary size, type of enhancement after gadolinium injection, stalk thickness, suprasellar extension, and cavernous sinus invasion) [5]. The study reported that diabetes insipidus was significantly more common in pituitary metastasis, pituitary height never surpassed 2 cm in ICI hypophysitis, the enhancement was homogenous in hypophysitis but heterogenous in metastases, the stalk was more commonly thickened in hypophysitis, and the extension above the sellar diaphragm or into the cavernous sinuses was more characteristic of pituitary metastases. But in general, an accurate diagnosis of ICI hypophysitis remains a challenge, also considering that serological markers specific for this condition are lacking.

ICI hypophysitis was first reported in 2003 at the National Cancer Institute in one of 14 patients with metastatic melanoma treated with ipilimumab, a monoclonal antibody blocking CTLA-4 [6], expanded by the same group in 2005 to include 7 additional patients [7], and first reviewed in this journal in 2009 [8]. An additional antibody targeting CTLA-4 (tremelimumab), three antibodies targeting PD-1 (nivolumab, pembrolizumab, and cemiplimab), and three targeting PD-L1 (atezolizumab, avelumab, and durvalumab) are now used in many clinical trials. The present review gives us the opportunity to offer our assessment of the field 10 years after the initial review.

2. Data sources and analysis of the published literature

At the Johns Hopkins Hypophysitis Research Center we maintain a database of articles published on all forms of hypophysitis, both primary and secondary. Articles are routinely found by searching for the word “hypophysitis” in Scopus (https://www2.scopus.com), Google Scholar (https://scholar.google.com), and PubMed (https://www.ncbi.nlm.nih.gov/pubmed), as well as by screening the references cited in the retrieved articles. For the purpose of this review, we excluded all cases of hypophysitis secondary to an infectious agent or a systemic disease affecting the sellar region.

We identified a total of 203 ICI hypophysitis articles, spanning the period from July 2003 to August 2019, and classified them into seven categories (Figure 1A): case reports (articles describing in details a single patient, No.= 80), cases series (those describing in details between 2 and 17 patients, No.= 37), cohort studies (those describing a group of patients with ICI hypophysitis whose clinical characteristics, however, are not individually identifiable, No.= 15), reviews (No.= 54), experimental studies (No.= 2), imaging studies (No.= 2). and editorials (No.= 4). Occasionally (No.= 9), articles reported both detailed patient features and a literature review. Articles were written mainly in English (192, 95%), with a few in Japanese (No. = 3), French (3), German (2), Spanish (1), Danish (1), or Chinese (1). The number of articles has increased steadily through the years, reaching a peak of 45 in 2019 (Figure 1B, circles). The interest in ICI hypophysitis has skyrocketed, as it can be gathered from the number of reviews published about it, a number that jumped to 16 during the first eight months of 2019 (Figure 1B, diamonds).

Figure 1.

Epidemiology of ICI hypophysitis. A) Distribution of 203 ICI hypophysitis papers according to the type of article. B) Yearly counts of ICI hypophysitis papers (circles for all article types, and diamonds for reviews), from July 2003 to August 2019. C) Geographical location and count of ICI hypophysitis patients published in USA from 2003 to 2019. D) Geographical location and count of ICI hypophysitis patients published in Japan from 2016 to 2019.

The 203 articles indicated above featured a total of 276 ICI hypophysitis patients (Table 1) (the corresponding citations are provided as Supplemental Table 1). Most of the patients developed hypophysitis after administration of a CTLA-4 blocking antibody (192 of 276, 69%); a fifth developed it after PD-1 blockade (64, 23%); a minority (5, 2%) after PD-L1 blockade; and the remaining few (15, 6%) after combination therapy or an unspecified ICI. As for primary hypophysitis (see Figure 2 in [8]), the countries reporting the greatest number of ICI hypophysitis patients were the United States of America (Figure 1C) and Japan (Figure 1D), with concentration in the regions hosting the major academic centers.

Table 1.

Classification of 276 published cancer patients (July 2003 – August 2019) who developed hypophysitis after immune checkpoint inhibitor (ICI) treatment, according to the type of ICI.

Immune Checkpoint Inhibitor	No.	(%)
Anti-CTLA4:	192	(70%)
Ipilimumab	190
Tremelimumab	2
Anti-PD1:	64	(23%)
Nivolumab	53
Pembrolizumab	11
Anti-PD-L1:	5	(2%)
Atezolizumab	3
Avelumab	1
Durvalumab	1
Anti-CTLA4 and anti-PD1 combination:	11	(3.9%)
Blinded ICI type:	4	(1.4%)
Total	276	100

Clinical trials were used to obtain an overall estimate of the incidence of hypophysitis after administration of the eight ICIs, but not included in this analysis because the individual patient characteristics were not described with sufficient details and there was often uncertainty about the exact diagnosis of hypophysitis. Nevertheless, we analyzed the ICI clinical trials deposited at https://clinicaltrials.gov to offer a snapshot of the magnitude of this area of medicine, expanding the assessment previously made by Chen and Mellman [9]. We queried the clinical trial database using the names of the 8 ICIs (ipilimumab, tremelimumab, nivolumab, pembrolizumab, cemiplimab, atezolizumab, avelumab, and durvalumab) and retrieved a total of 3,845 clinical trials as of August 1, 2019 (Table 2). The majority of them (55%) used an antibody directed against PD-1; about a quarter of them (24%) an antibody against PD-L1; and the remaining 17% an antibody against CTLA-4. About half of the trials are classified as phase 2, and 555 as recruiting status. The trials involved 44 countries of the world, with the USA being the dominant one (2,752 of 3,845, 72%, Supplemental Figure 1). Overall, these data reveal the remarkable footprint clinical trials using ICIs have, and bode to an ever-increasing appearance in the clinics of immune-related adverse events such hypophysitis.

Table 2.

Clinical trials using humanized monoclonal antibodies directed against immune checkpoint PD-1, CTLA-4, or PD-L1. Data were extracted from the ClinicalTrials.gov database on August 1, 2019.

Antibody name (Brand)	Antibody isotype	Antibody target	FDA 1st Approval	No. (%) of trials
Pembrolizumab (Keytruda®)	IgG4	PD-1	2014	1,064 (28%)
Nivolumab (Opdivo®)	IgG4	PD-1	2014	1,000 (26%)
Ipilimumab (Yervoy®)	IgG1	CTLA-4	2011	632 (16%)
Durvalumab (Imfinzi®)	IgG1	PD-L1	2017	383 (10%)
Atezolizumab (Tecentriq®)	IgG1	PD-L1	2016	360 (9%)
Avelumab (Bavencio®)	IgG1	PD-L1	2017	187 (5%)
Tremelimumab (No Brand)	IgG2	CTLA-4	N.A.	183 (5%)
Cemiplimab (Libtayo®)	IgG4	PD-1	2018	36 (1%)
Total				3,845 (100%)

3. Physiopathology of hypophysitis secondary to immune checkpoint inhibitors

3.1. Normal functions of CTLA-4 and PD-1

Cytotoxic T lymphocyte-associated protein 4 (CTLA-4), also known as CD152, is a member of the immunoglobulin superfamily, discovered in 1987 while screening a cDNA library prepared from mouse CD8 T cells [10]. The 223 amino acid long protein contains an extracellular domain, a transmembrane domain, and a cytoplasmic tail, and is encoded by a 4-exon gene located on chromosome 2 in humans and 1 in mice [11]. Exons 1 and 2 code for the extracellular portion of CTLA-4, exon 3 for the transmembrane region, and exon 4 for the cytoplasmic tail. The full-length mRNA transcript, featuring exon 1 through 4, can undergo alternative splicing, with differences between humans and mice. Humans can make a transcript coding for soluble CTLA-4 that lacks exon 3, and a transcript made up only of exons 1 and 4. Mice can make those two variants, as well as a third one called ligand-independent CTLA-4 that contains exons 1, 3, and 4 [11]. The membrane-bound, full-length CTLA-4 functions as a homodimer held together by a disulfide bond, whereas soluble CTLA-4 as a monomer. The extracellular portion of CTLA-4 has a single IgV-like domain that contains the residues used to bind its ligands: B7-1 (CD80) and B7-2 (CD86), which are mainly expressed on professional antigen-presenting cells (dendritic cells, B cells, and macrophages).

In mice, CTLA-4 is expressed predominantly in T cells and T cell rich organs (thymus and lymph nodes), with the highest levels found in the Foxp3+ CD4+ regulatory T cells subset (www.biogps.org). In regulatory T cells, CTLA-4 is expressed on the cell surface constitutively. Its engagement boosts the function of regulatory T cells by increasing their proliferation and the production of immunosuppressive cytokines, such as IL-10 and TGF-b, and molecules, such as indoleamine 2,3, -dioxygemase. In effector T cells that are resting, CTLA-4 is found intracellularly, in vesicles located close to the microtubule organizing center [12]. When effector T cells become activated (via engagement of the T cell receptor and CD28, a co-stimulatory molecule expressed on the T cell surface), a small pool of CTLA-4 molecules translocate to the cell surface, particularly to the uropod (a projection of the plasma membrane that forms the T-cell side of the immunological synapse) and undergo dimerization [13]. Surface CTLA-4 levels are highest 2-3 days after activation, and directly proportional to the strength of the T cell receptor signal. Following the initial discovery of CTLA-4, the laboratories of Jeffrey Bluestone [14] and James Allison [15] convincingly showed in mice that the appearance of CTLA-4 on the cell surface sends inhibitory signals into effector T cells that ultimately block their proliferation and activation. In other words, when CTLA-4 appears on the surface of effector T cells, it turns them off. CTLA-4 does so through two main mechanisms. The first one is by antagonizing the stimulatory signal induced by CD28: CTLA-4, in fact, has a 2-fold higher affinity and 100-fold higher avidity for B7.1 and B7.2 than CD28 [16]. During the initial phases of the T cell response, ligation of CD28 to B7 induces a stimulatory signal that is necessary for achieving optimal T cell activation. CTLA-4 competes with CD28 for binding to B7 and therefore, due to its higher affinity and avidity, reduces the CD28-dependent costimulation in T cells [17]. Considering that B7.1 and B7.2 are expressed on professional antigen-presenting cells, which are mainly found in secondary lymphoid organs, the inhibitory effect of CTLA-4 occurs in these locations, that is during the primary phase of T activation. The other mechanism by which CTLA-4 inhibits T cells is the delivery of inhibitory signals through its cytoplasmic tail, ultimately resulting in decreased nuclear accumulation of AP-1, NF-κB, and NFAT, although the downstream pathway remains to be defined [18]. Besides these immune-centric activities of CTLA-4, other cellular functions may be affected by CTLA-4 by its ability to modulate the small G protein Rap1 [19], and thus impact many cellular responses such as cell adhesion and mobility.

When both copies of the CTLA-4 gene are deleted globally (i.e., in all T cell subsets), mice are born healthy but after 5-6 days their T cells proliferate and infiltrate many non-lymphoid organs, resulting in death at about 1 month of age from myocardial infarcts [20]. These mice have enlarged spleen and lymph nodes (about 5-10 times the normal size) [21] and lymphocytic infiltration in many organs besides the heart, most prominently in thymus, lungs, bone marrow, liver, and pancreas [20], overall indicating that the absence of CTLA-4 leads to an unregulated expansion of effector T cells capable of mediating autoimmune responses that ultimately kill the host. These mice also have hypergammaglobulinemia, owing to boosted B cell activation. When CTLA-4 is deleted specifically in regulatory T cells, the phenotype is different from the global deletion and dependent upon the timing of deletion. Deletion during embryonic life causes autoimmune responses of lower severity, where mice die at about 6 months of age with modest and less prevalent multi-organ lymphocytic infiltration [22]; deletion during adulthood leads, on the contrary, to resistance to autoimmunity, where mice have greater numbers of regulator T cells that retain their immunosuppressive activity and induce the upregulation of inhibitory molecules (such as IL-10, PD-1, and LAG-3) on effector T cells [23]. These findings reveal a remarkable plasticity of CTLA-4, with different effects according to timing, conditions of engagement, and intracellular signaling.

In humans, CTLA-4 is also expressed in lymphoid tissues, with enhanced levels in tonsils, appendix, and lymph nodes, but it is also found in many other tissues, such as lungs, gallbladder, urinary bladder, testis, breast, placenta, and skin (see human protein atlas RNA-seq dataset at https://www.proteinatlas.org/, and Genotype-Tissue Expression RNA-seq dataset at https://www.gtexportal.org). Interestingly for this review, we reported the expression of CTLA-4 mRNA and protein in the human pituitary gland [24], a topic that will be explore in a later section. Humans with a defective copy of CTLA-4 (haploinsufficiency), due for example to a nonsense mutation in exon 1, develop a complex immune dysregulatory syndrome characterized by lymphocytic infiltration in multiple organs (duodenum, lungs, cerebellum), autoimmune manifestations, recurrent infections, and hypogammaglobulinemia [25]. Humans with polymorphisms in the 3’ untranslated region of CTLA-4 are at greater risk of developing autoimmune diseases such as type 1 diabetes mellitus and Graves disease [26].

Overall, a large body of data has shown that CTLA-4 normally functions as an inhibitor of immune (mainly T cell) responses. This inhibition occurs predominantly on effector T cells, where the appearance of CTLA-4 on the cell surface quenches their proliferation and activation. When CTLA-4 is mutated or contains certain polymorphisms, the host is at greater risk of developing autoimmune diseases because its T cells are no longer held in check, and so they become more activated, diversified, and capable of proliferating. While Dr. Bluestone focused on the role of CTLA-4 in autoimmune diseases, Dr. Allison hypothesized that blocking CTLA-4 with an antibody would be a way to unleash the T cells and make them more capable of destroying cancer cells.

Programmed cell death 1 (PD-1), like CTLA-4, is a member of the immunoglobulin superfamily, encoded by a gene located on chromosome 2 in humans and 1 in mice, that is mainly expressed on effector T cells upon activation and functions as a way to attenuate their function and prevent tonic signaling [27]. Although its name was chosen because of the original observation that it induced the death of stimulated 2B4.11 (a murine T cell hybridoma) [28], PD-1 is better viewed as an immune checkpoint, a marker of activated T cells that are on the pathway of becoming exhausted. The 228 amino acid protein has a similar architecture to CTLA-4, with an extracellular IgV-like domain, a transmembrane region, and a longer C-terminal cytoplasmic tail that contains immunoreceptor tyrosine-based inhibitory and switch motifs. Also similarly to CTLA-4, the 5-exon gene codes for different splice variants, one lacking exon 2, one lacking exon 3, one exons 2 and 4, and another lacking exons 2, 3, and 4 [29]. The variant lacking exon 3 does not contain the transmembrane domain and thus produces a soluble protein that resemble soluble CTLA-4 and has been shown to have clinical utility as a serum biomarker in HIV infection [30] and cancer [31]. Like CTLA-4, PD-1 carries on its inhibitory functions on T cells by targeting CD28, albeit from a different angle: PD-1, in fact, recruits to its cytoplasmic tail the tyrosine phosphatase SHP2 (also known as PTPN11) that in turn de-phosphorylates CD28 [32], as well as the proximal signaling elements of the T cell receptor [33]. PD-1 is normally found not only on effector T cells, but also on activated B cells, monocytes, and dendritic cells, as well as in non-lymphoid organs such as the testis and the cerebral cortex. PD-1 works by interacting with its ligands, PD-L1 and PD-L2, which are expressed widely in non-lymphoid tissues (https://www.proteinatlas.org), including the cancer cells. Continuous engagement of PD-1 with its ligands inhibits the effector T cell because it undergoes an epigenetic program of exhaustion [34]. The inhibitory effect of PD-1 occurs in the peripheral tissues, on T cells that are already activated, thus at later stages of the immune response. In other words, PD-1 is a negative regulator of already established immune responses and thus has a more restricted effect than CTLA-4.

Not surprisingly thus, deletion of both copies of PD-1 in mice causes autoimmune manifestations that are much less severe than those seen in CTLA-4 knockout mice and dependent upon the strain. The laboratory of Dr. Honjo showed that in the C56BL6 strain lack of PD-1 causes lupus-like manifestations [35], whereas in the BALB/c strain it leads to dilated cardiomyopathy [36]. This heart selectivity is also seen in mice where the PD-L1 is deleted [37]. In humans, no significant pathology has been associated with mutations in the PD-1, PD-L1, or PD-L2 genes. Only polymorphisms in the PD-1 gene have been associated with increased susceptibility to systemic lupus erythematosus (G allele at position 7146) [38], disease progression (A allele at position 7146) in multiple sclerosis [39].

3.2. Anti-cancer mechanisms of CTLA-4 and PD-1/PD-L1 blockade

Following the first decade of studies in mice, two fully humanized monoclonal antibodies blocking CTLA-4 (named ipilimumab and tremelimumab) were developed and began to be used in clinical trials of melanoma in June 2000. These antibodies bind to the region in the extracellular domain of CTLA-4 that contains the residues used for binding to B7-1 and B7-2, therefore they sterically hinder the active site of CTLA-4 [40], preventing CTLA-4 from carrying on its inhibitory functions. It became quickly evident that in a minority of patients these blocking antibodies could induced a prolonged tumor regression or disappearance, as it had never seen with previous melanoma treatments. The timing of this response, however, was not predictable and different from that seen with conventional chemotherapy: some patients treated with ipilimumab, for example, responded long after ipilimumab had been discontinued, or in a delayed fashion after the initial tumor had progressed. Therefore, it was more difficult for regulatory agencies to approve a new oncology drug based on the progression-free survival metrics because a more prolonged metrics was needed, that of overall survival. It took another decade of studies before the Food and Drug Administration granted approval to ipilimumab for the treatment of patients with advanced melanoma in March 2011. The approval depended upon the completion of two large phase 3 melanoma trials showing that ipilimumab significantly prolonged overall survival when compared to glycoprotein 100 vaccine [41] or dacarbazine chemotherapy [42]. Interestingly, the survival curves reported in Figure 1A of the Hodi paper [41] also showed the appearance of a right tail, a small subset of patients did not die but rather showed long-term, durable responses. In sum, blocking CTLA-4 with a monoclonal antibody not only prolonged the median survival but also cured a small proportion of patients.

The mechanisms through which administration of a CTLA-4 blocking antibody eliminates cancer cells are incompletely understood. The predicted mechanism of action is the one on effector T cells: removing the CTLA-4 inhibitory break makes these T cells more active and proliferative, enabling them to respond to overexpressed and mutated tumor antigens. Indeed, studies have shown an expansion of the CD4+, ICOS+, Tbet+, PD-1+ subset [43], an increased ability of the CD4+, Th17 subset to produce IL-17 [44], and an expansion of the CD8+ PD-1+ subset [45]. Interestingly, a recent case series reported two patients with stage IV melanoma treated with the combination of ipilimumab and nivolumab who underwent biopsy of the thymus. Pathology showed true thymic hyperplasia [46], suggesting that ICI administration leads to an expansion of the T cell emigrants from the thymus. This finding is consistent with the notion that CTLA-4 blockade increases the richness (i.e., number of unique T cell receptor Vβ CD3 regions) and abundance of the T cell receptor repertoire in peripheral T cells [47]. In other words, CTLA-4 blockade allows the emergence from the thymus of T cells clones that recognize antigens with low signal strength, clones that would have otherwise been deleted. As we will see in the next section, broadening of the T cell receptor repertoire has a cost because it increases the risk of developing immune-related adverse events. Another predictable mechanism of action of CTLA-4 blockade is the one on regulatory T cells, a subset of CD4+ T cells that specifically express the transcription factor forkhead boxprotein P3 (FoxP3) and constitutively CTLA-4 on their surface. CTLA-4 blocking antibodies likely deplete the number of Treg, and therefore leave effector T cells freer to execute their function, including the destruction of cancer cells. The depletion occurs in the tumor microenvironment and is done by macrophages that recognize with their Fc gamma 3 receptor (CD16) the constant region of the anti-CTLA-4 antibody bound to the regulatory T cells, a mechanism known as antibody-dependent cell-mediated cytotoxicity. Interestingly, melanoma patients who have a polymorphism (V158F) in CD16 that increases its affinity for binding antibodies respond better to ipilimumab [48].

The mechanisms through which blocking PD-1 or its ligands lead to killing of the cancer cells are also incompletely understood, but one thing is certain: blockade of the PD-1 pathway is conceivably the most important breakthrough in the history of cancer therapy, now used in many cancer types in thousands of clinical trials (Table 2). Nivolumab, the first anti-PD-1 antibody, was initially given to patients in October 2006, its results published as a full paper in 2012 [49], and then approved by the FDA for the treatment of refractory melanoma and non-small cell lung cancer in 2014 and 2015, respectively. Other anti-PD1 and anti-PD-L1 antibodies followed and received FDA approval through accelerated pathways. PD-1 blockade reinvigorates pre-existing, anti-tumor effector T cells that were present at the tumor site but exhausted, enabling them to kill cancer cells despite their expression of PD-L1 and PD-L2. The main mediator of the PD-1 blockade effect seems to be a subset of effector CD8 T cells (cytotoxic T cells) that expresses the transcription factor TCF1, and preferentially undergoes proliferative expansion after PD-1 blockade [50]. The tumors that respond best to PD-1 blockade are those driven by viral infections, such as the Merkel cell carcinoma of the skin [51] induced by the Merkel cell polyomavirus [49], those having high mutational burden [52], such as the hereditary nonpolyposis colorectal cancer that has mutations in the DNA mismatch repair proteins [53], and the nodular-sclerosis type of Hodgkin lymphoma, which features an amplification of the short arm of chromosome 9 (at position 24.1) leading to constitutive expression of PD-L1 and PD-L2 and activation of the Janus kinase 2-STAT pathway [54]. More controversial is the effect of PD-1 blockade on regulatory T cells, a fraction of which expresses PD-1. A recent study showed that administration of an anti-PD1 antibody, rather than depleting the PD-1+ Tregs, actually leads to an expansion of their number and function, expansion that was associated to accelerated tumor progression in patients with gastric cancer [55].

3.3. Mechanisms of toxicity of ICIs: an overview

As the first success stories of tumor regression were noted in the early 2000s, it also become apparent that patients treated with ICIs developed a number of side effects resulting from inflammation of the targeted organ. The first patient with ICI hypophysitis was reported at the National Cancer Institute in a cohort of 14 melanoma patients treated with ipilimumab who had shown a variety of toxicities [6], and then expanded by the same group to a case series dedicated to hypophysitis [7]. The toxicities caused by ICIs were different from those seen with conventional chemotherapy or molecularly targeted agents, and collectively named immune-related adverse events (irAEs). We now know that irAEs can affect almost any organ or tissue in the body (see Figure 2 in [56]), so much so that every new symptom developing in a cancer patient treated with ICI must be considered an irAE until proven otherwise. They manifest at variable times during ICI therapy and do not follow the cyclical pattern typically seen with conventional chemotherapy. Several difficulties have surrounded the irAEs due to naming and classification uncertainties, lack of centralized repositories, and paucity of pathological data from the organ/tissue affected by irAEs.

The naming and classification have been initially a challenge, in part because oncologists who followed the ICI patients and wrote the paper did not have specialty-level knowledge of the autoimmune diseases mimicked by the irAE. Hypophysitis is a good example of that. The first published clinical trial using an ICI reported irAEs as Table 3 [41]. Looking at the endocrine section of that table one sees six entries: hypothyroidism, hypopituitarism, hypophysitis, adrenal insufficiency, increase serum TSH, decreased serum ACTH. With the exception of increased TSH and hypophysitis itself, all of the other entries could be found in patients with hypophysitis, therefore, it is not possible to know how many patients did actually develop hypophysitis after ICI administration. The National Cancer Institute recognized this difficulty early on and developed a set of terms to be used when classifying irAEs in clinical trials, called Common Terminology Criteria for Adverse Events. The current version (v5, released in November 2017) contains 837 terms, 14 of which are endocrine disorders, and now include hypophysitis. Each term is classified into five grades of increasing clinical severity up to death (grade 5). A second difficulty when analyzing irAEs is the lack of a centralized data repository. The FDA does maintain the side effect repository resource, which stores side effects extracted from the FDA drug labels, but its most update version (4.1) does not include ICIs. Data about irAEs, therefore, mainly come from review of published papers and text mining of the FDA drug labels. Khoja and colleagues reviewed 122 ICI clinical trials published between 2003 and 2015 and selected 48 for their analysis, totaling about 7,000 patients, based on the quality of the data [57]. They reported that colitis, hypophysitis, rash, and pruritus were significantly more common with CTLA-4 blockade than PD-1/PD-L1 blockade, whereas pneumonitis, myalgia, hypothyroidism, arthralgia, and vitiligo were more common with PD-1/PD-L1 blockers. They also noted that different tumor histiotypes treated with the same ICI give rise to different irAEs profiles: for example, patients treated with anti-PD-1 develop more commonly pneumonitis when the underlying cancer is non-small cell lung cancer and colitis when it is melanoma, overall suggesting that the heterogeneity of the tumor microenvironment contributes to the irAE profile. Wang and Xu manually extracted irAEs terms from FDA drug labels of 6 ICIs (nivolumab, pembrolizumab, durvalumab, avelumab, and ipilimumab) and mapped them to terms in the Medical Dictionary for Regulatory Activities [58]. They then calculated the frequencies of pairwise irAEs among the 6 ICIs and compared them to those of 1,507 FDA-approved, non-ICI drugs, expressing the results as Jaccard similarity coefficient (number shared divided by total number in the pair minus number shared). They reported that ICIs with the same target have more similar irAEs than ICIs with different targets, suggesting that certain irAEs are caused on-target effects of the ICI. A final difficulty related to irAEs is that with few exceptions such as the colon the organ/tissue affected by the irAE is not pathologically examined, therefore it becomes very difficult to understand the mechanisms of irAE development.

Table 3:

Key features of Primary Hypophysitis and Hypophysitis secondary to immune checkpoint inhibitors

Features	Primary Hypophysitis	Hypophysitis secondary to ICI	p value
No. of patients	1342	276
No. of articles	672	124
Publication time span, years (range)	102 (1917 to 2019)	16 (2003 to 2019)
No. (%) of females	918 (68%)	77 (28%)	<0.0001
No. (%) of males	412 (31%)	199 (72%)	<0.0001
F:M ratio	2:1	1:4	<0.0001
Mean age at onset, years (SD)	42 (17)	61 (12)	<0.0001
Mean time to onset	Unknown (years)	14 (13.5 weeks)
Symptoms at presentation, No./Tot (%)
Headache	495 of 1035 (48 %)	119 of 267 (45 %)
Visual disturbances	336 of 1041 (32 %)	17 of 267 (6 %)	<0.0001
Low cortisol	381 of 1015 (38 %)	215 of 267 (81 %)	<0.0001
Low thyroxine	162 of 989 (16 %)	47 of 266 (18 %)
Low sex steroids	215 of 1006 (21 %)	29 of 266 (11 %)	<0.0001
Polydipsia and polyuria	351 of 1021 (34 %)	4 of 266 (2 %)	<0.0001
Endocrine abnormalities at diagnosis, No./Tot (%)
Secondary hypocortisolism	519 of 859 (60 %)	237of 248 (96 %)	<0.0001
Secondary hypothyroidism	436 of 871 (50 %)	145 of 232 (63 %)	0.003
Secondary hypogonadism	430 of 789 (55 %)	117 of 197 (59 %)
Hyperprolactinemia	297 of 768 (39 %)	17 of 154 (11 %)	<0.0001
GH deficiency	229 of 628 (37 %)	23 of 123 (19 %)	<0.0001
Central Diabetes insipidus	409 of 654 (63 %)	4 of 163 (3 %)	<0.0001
MRI findings, No./Tot (%)
Abnormal	900 of 929 (97 %)	125 of 194 (64 %)	<0.0001
Normal	29 of 929 (3 %)	69 of 194 (36 %)	<0.0001
Diagnosis established by, No./Tot (%)
Surgical pathology	814 of 1342 (61 %)	0 of 276 (0 %)	<0.0001
Autopsy	43 of 1342 (3 %)	1 of 276 (1 %)	<0.0001
Clinical and Imaging criteria,	450 of 1342 (33 %)	209 of 276 (78 %)	<0.0001
Only clinical criteria	35 of 1342 (3 %)	58 of 276 (21 %)	<0.0001
Pathologic variants, No./Tot (%)
Lymphocytic	554 of 1342 (41 %)	0
Granulomatous	158 of 1342 (12 %)	0
IgG4 plasmacytic	63 of 1342 (5 %)	0
Mixed forms	41 of 1342 (3 %)	0
Xanthomatous	33 of 1342 (2 %)	0
Necrotizing	4 of 1342 (1 %)	1 of 276 (1 %)
Unknown	489 of 1342 (36)	275 of 276 (99 %)
Treatment, No./Tot (%)
Mass reduction with or w/ HRT	675 of 880 (77 %)	142 of 242 (59 %)
- Surgery	491 of 675 (73 %)	0 of 142 (0 %)	<0.0001
- Lympholytic drugs	178 of 675 (26 %)	142 of 142 (100 %)	<0.0001
- RTX/SRS	6 of 675 (1 %)	0 of 142 (0 %)
HRT only	174 of 880 (20 %)	95 of 242 (39 %)	<0.0001
None	31 of 880 (3 %)	5 of 242 (2 %)
Median follow-up time, years (iqr)	1.2 (2.5)	0.8 (1.6)	0.0002
Outcome, No./Tot (%)
Resolved spontaneously	12 of 627 (2 %)	0 of 164 (0 %)
Improved after treatment w/o need of HRT	144 of 627 (23 %)	10 of 164 (6 %)	<0.0001
Required long-term HRT	382 of 627 (61 %)	145 of 164 (88 %)	<0.0001
Recurrence after treatment	44 of 627 (7 %)	1 of 164 (1 %)	<0.0001
Death	45 of 627 (7 %)	8 of 164 (5 %)

Abbreviations:

No.: number, SD: standard deviation, HRT: Hormonal replacement therapy, RTX: radiotherapy, SRS: stereotactic radiosurgery, iqr: interquartile range

Four mechanisms have been postulated to explain the onset of irAEs, and elegantly reviewed by Postow and colleagues [59]. The first one, which can be referred to as “shared antigens”, is the observation that administration of ICIs induces the emergence of effector T cells having a T cell receptor that recognizes both a tumor neoantigen and a normal tissue antigen. For example, in patients with melanoma administration of ipilimumab induces the appearance of CD8 T cells that recognize antigens expressed in both the normal and malignant melanocytes. These CD8 T cells, which produce high levels of interferon-gamma and express the chemokine receptor CXCR3, kill the cancer cells and also induce depigmentation (vitiligo) in the normal skin. Indeed, the appearance of vitiligo in melanoma patients undergoing immunotherapy is a good prognostic sign: patients who develop vitiligo have a 4-fold lower risk of death than patients who do not [60]. A similar scenario was proposed in two patients who developed fulminant myocarditis after combined PD-1 and CTLA-4 blockade [61], where sequencing of the T cell receptor repertoire in heart, skeletal muscle, and tumor showed the dominance of one T cell receptor clone in one of the patients. A second proposed mechanism is that ICI treatment causes the expansion of T cell populations capable of releasing pro-inflammatory cytokines such as IL-17 [44], mechanism invoked in cancer patients who develop colitis as an irAE. The third mechanism, named “epitope spreading”, postulates that ICIs cause a general, nonspecific inflammation secondary to the tumor lysis that expands (“spreads”) pre-existing autoimmune responses. For example, patients with non-small cell lung cancer who have pre-existing thyroid antibodies are at much greater risk of developing hypothyroidism after PD-1 blockade than patients who do not have these antibodies at baseline [62]. Similarly, an analysis of 23 autoantibodies commonly tested in clinical laboratories applied to melanoma patients treated with ipilimumab showed that these antibodies develop in about 20% of the patients who did not have these antibodies before treatment [63]. The last mechanism is that one we proposed based on the findings of CTLA-4 expression in pituitary endocrine cells [24], that is the fact that the target recognized by the ICI, such as CTLA-4 and PD-1, can be found not only on T lymphocytes (the intended target) but also on other cells, a topic that will be discussed in the following section. Similar considerations apply to the expression of PD-L1 on the beta cells of the endocrine pancreas [64], or on hypothalamic cells [65].

3.4. Mechanisms of pituitary toxicity after ICIs

With the caveats discussed above in mind regarding the definition and data repository of irAEs, it is reasonable to consider best the estimates of hypophysitis occurrence after ICI provided by endocrinologists who are familiar with hypophysitis and participate to the care of these oncology patients. These estimates report an incidence of hypophysitis of 12% after CTLA-4 blocking antibodies, and 0.5% after PD-1 blocking antibodies [66], estimates confirmed in a recent meta-analysis of 38 randomized clinical trials [67]. It is unknown why hypophysitis, a disease that is exceedingly rare in its primary form, becomes so prevalent after CTLA-4 blockade. It also remains unknown why “only” 12% of the treated patients develop it, whereas the remaining majority (88%) does not. These clinical observations highlight the existence of complex interplays between genetics, immune system, and environment (as it is typically seen in primary autoimmune diseases), therefore of disease mechanisms not easily identifiable.

For the CTLA-4 mechanism, we postulated that a reason for the exquisite pituitary toxicity was the “ectopic” expression of CTLA-4 on non-immune cells, such as those of the adenohypophysis (see Figure 6 in [4]). In brief, we envisioned that the injected anti-CTLA-4 antibody bound not only to its intended targets (CTLA-4 expressed on the surface of activated effector T cells and regulatory T cells), but also to CTLA-4 expressed on pituitary cells. Since ipilimumab is an IgG1, it can bind complement and activate the classical complement cascade to induce antibody-dependent cell-mediated cytotoxicity (a type 2 hypersensitivity reaction). Indeed we showed complement deposition on pituitary cells in mice injected with a CTLA-4 blocking antibody [24], and in a patient who died after tremelimumab treatment [4], a CTLA-4 blocking antibody of the IgG2 isotype that can also bind complement. We also envisioned that these initial, innate events of the immune response would then trigger an activation of the adaptive immune system leading to a more classical type of autoimmune damage (type 4 hypersensitivity).

The establishment of multi-center resources to study gene expression and regulation in a variety of human tissues, such as the GTEx project (supported by the Common Fund of the Office of the Director of the National Institutes of Health) is revealing novel insights into the biology of CTLA-4 and PD-1, which were traditionally viewed as “T cell only” molecules. The GTEx project has currently data on 51 human tissues, each contributed by an average number of 327 donors (range from 9 to 803). Analysis of the mRNA expression of CTLA-4 and PD-1 from the GTEx data confirmed that the spleen is the highest expresser of both CTLA-4 and PD-1 (Figure 2A). Interestingly, however, other non-lymphoid organs such as atrium, pituitary, thyroid, and testis express them. We proved CTLA-4 expression at the protein level and localized it specifically in the cytosol of adenohypophyseal cells (Figure 2B), mainly those producing prolactin but other cell types express it as well. There is significant variability among individuals in the expression of these checkpoints, a variability that could impart explain why some patients develop toxicity after administration of ICI whereas others do not. A recent study has taken advantage of this “ectopic” expression of CTLA-4 on pituitary endocrine cells and successfully used anti-CTLA-4 and anti-PD-1 to treat a patient with aggressive, hypermutated ACTH-secreting pituitary carcinoma [68]. Other interesting deductions can be drawn from the analysis of Figure 2A. For example, the atrium is the second highest expresser of PD-1 after the spleen: notably mice where the PD-1 gene is deleted develop autoimmune dilated cardiomyopathy [36]. Equally interesting, the testis, an organ classically considered by immunologists as an “immune-privileged site”, expresses high levels of CTLA-4, not very different from those seen in the spleen.

Figure 2.

Expression of CTLA-4 and PD-1. A) mRNA expression of CTLA-4 and PD-1 in human tissues collected by the Genotype-Tissue Expression RNA-seq project. B) Protein expression of CTLA-4 in pituitary endocrine cells (40x).

For the PD-1/PD-L1 mechanism, there are no published experimental articles that allow us to postulate a pathway responsible for the pituitary toxicity. Therefore, the pathogenesis of hypophysitis in the 64 published patients who developed it after PD-1 blockade or the 5 who developed it after PD-L1 blockade (Table 1) is unknown. Some speculations can be offered here, centered on the dominance of the ACTH deficiency and the time of disease onset. In contrast to CTLA-4 induced damage, which involved almost with equal frequencies the ACTH-, TSH-, and gonadotropin-secreting cells, the one induced by PD-1 blockade overwhelmingly predilects that ACTH-secreting cells (see Table 4 in a later section). Perhaps ACTH-secreting cells are the pituitary cell type that expresses the highest levels of PD-1. As shown in Figure 2, the GTEx data reveal that the pituitary gland is the 6^th highest expresser of PD-1 (following spleen, atrium, ileum, whole blood, and lung), but it remains to be determined on what type of pituitary cells PD-1 is found. Corticotroph selectivity could also relate to the notion that ACTH-secreting express Fc receptors [69], and are thus theoretically more capable of binding the infused anti-PD-1 antibody as it flows through the hypophyseal portal circulation. Furthermore, in contrast to hypophysitis secondary to CTLA-4, which develops at a predictable interval of 10 weeks after the first infusion, the form caused by PD-1 blockade has a much more variable presentation: some patients develop it within a few weeks, most of them at an average of 27 weeks, and a few more than a year later. The early-onset cases likely unfold through a mechanism akin to that seen in destructive (painless) thyroiditis: there is an initial, transient phase where the targeted thyrocytes are destroyed and release their pre-stored thyroid hormones, leading to biochemical and at times clinical thyrotoxicosis. This initial phase is then followed by the more permanent hypothyroid phase. Interestingly, a renal cell carcinoma patient that was treated with nivolumab and developed isolated ACTH deficiency had a brief initial phase were serum ACTH and cortisol levels were markedly elevated without signs or symptoms of Cushing disease [70]. Being asymptomatic, this initial phase of hypercortisolemia may go undetected. All in all, much remains to be learned about the biology of the CTLA-4 and PD-1 immune checkpoints.

Table 4:

Comparison between Hypophysitis caused by CTLA-4 blockade and Hypophysitis caused by PD-1/PD-L1 blockade.

Features	CTLA-4 blockade	PD-1 or PD-L1 blockade	p value
No. of patients	192	69
Sex M, No./Tot (%)	142 of 192 (74%)	50 (72%)
Mean age at onset, years (SD)	60.7 (12)	63.8 (12)	0.059
Type of Cancer, No./Tot (%)
Skin	180 of 192 (94%)	20 of 69 (29%)	<0.001
Lung	1 of 192 (0.5%)	35 of 69 (51%)	<0.001
Genito-urinary	4 of 192 (2%)	6 of 69 (9%)	0.014
Others	7 of 192 (3.5%)	8 of 69 (11%)	0.015
Mean time to onset, weeks (SD)	10.5 (4.8)	27.1 (22.4)	<0.001
Mean doses to onset, No. (SD)	3.4 (1.1)	10.3 (7.2)	<0.001
Symptoms at onset No./Tot (%)
Low cortisol	139 of 185 (75%)	62 of 68 (91%)	0.005
Low thyroxine	39 of 184 (21%)	5 of 68 (7%)	0.010
Low sex steroids	29 of 185 (16%)	0 of 68 (0%)	0.001
Polydipsia and polyuria	1 of 184 (.5%)	2 of 68 (3%)
Headache	111 of 185 (60%)	3 of 68 (4%)	<0.001
Visual disturbances	15 of 185 (8%)	0 of 68 (0%)	0.015
Endocrine abnormalities at diagnosis, No./Tot (%)
Secondary hypocortisolism	156 of 165 (95%)	67 of 69 (97%)
Secondary hypothyroidism	139 of 163 (85%)	2 of 56 (4%)	<0.001
Secondary hypogonadism	107 of 142 (75%)	6 of 46 (13%)	<0.001
GH deficiency	22 of 81 (27%)	1 of 35 (3%)	0.003
PRL deficiency	26 of 102 (25%)	0 of 46 (0%)	<0.001
Hyperprolactinemia	7 of 102 (7%)	9 of 46 (20%)	0.101
Diabetes Insipidus	2 of 118 (2%)	1 of 40 (3%)
Hyponatremia	32 of 83 (39%)	32 of 52 (62%)	0.009
MRI findings, No./Tot (%)
Abnormal	114 of 141 (81%)	8 of 45 (18%)	<0.001
Normal	27 of 141 (19%)	37 of 45 (82%)	<0.001
Diagnosis established by, No./Tot (%)
Clinical and Imaging criteria	179 of 192 (93%)	20 of 69 (33%)	<0.001
Only clinical criteria	12 of 192 (6%)	61 of 69 (67%)	<0.001
Autopsy	1 of 192 (1%)	0 of 69 (0%)
Treatment, No./Tot (%)
Lympholytic drugs w/ or w/o HRT	121 of 163 (74%)	12 of 64 (19%)	<0.001
HRT only	38 of 163 (23%)	52 of 64 (81%)	<0.001
None	4 of 163 (3%)	0 of 64 (0%)
ICI discontinuation, No./Tot (%)
No	37 of 67 (56%)	38 of 54 (70%)
Yes, temporarily	2 of 67 (3%)	11 of 54 (20%)	0.002
Yes, permanently	27 of 67 (41%)	5 of 54 (10%)	<0.001
Outcome, No./Tot (%)
Required long-term HRT	93 of 104 (89%)	43 of 52 (90%)
Improved after treatment w/o need of HRT	6 of 104 (5%)	3 of 52 (6%)
Resolved spontaneously	1 of 104 (1%)	0 of 52 (0%)
Death	6 of 104 (5%)	2 of 52 (4%)
Recurrence after treatment	0 of 104 (0%)	0 of 52 (0%)

Abbreviations:

SD: standard deviation, HRT: Hormonal replacement therapy, ICI: immune-checkpoint inhibitor

4. Clinical Features of ICI hypophysitis

4.1. Comparison between ICI hypophysitis and primary hypophysitis

ICI hypophysitis differs from primary hypophysitis in regard to numerous clinical, diagnostic, treatment and outcome features, which we have summarized in Table 3 and Figure 3. ICI hypophysitis is more common in males (199 of 273, 72%, versus 412 of 1330, 31%, p < 0.0001) and older ages (61 ± 11.7 vs 42 ± 16.5 years, p < 0.0001, Figure 3A). These two features may reflect the demographics of the underlying cancer considering that melanoma and lung cancer, the tumor types where ICIs have been used the most, are more common in males. But it may also reflect true biological differences, considering that the male predominance in ICI hypophysitis remains even after adjusting for the male predominance in melanoma patients[71].

Figure 3.

Comparison of ICI hypophysitis to primary hypophysitis. A) Distribution of primary and ICI hypophysitis patients by sex, and age at diagnosis. B) Distribution of primary and ICI hypophysitis patients by hormonal deficiencies at diagnosis. * < 0.005, *** < 0.0001 by Pearson chi2. C) Follow-up time in primary hypophysitis and ICI hypophysitis patients. **< 0.001 by Wilcoxon rank sum test. D) Distribution of primary and ICI hypophysitis patients by outcome.

Symptoms derived from decreased cortisol are significantly more common in ICI hypophysitis (215 of 267, 81%) than primary hypophysitis (381 of 1015, 38%, p< 0.0001). They are diverse and include fatigue, appetite loss, nausea, vomiting, malaise, dizziness, and mild cognitive defects. They can be subtle and easily confused with the symptoms of the underlying cancer (thus leading to a delayed clinical recognition and diagnosis), or life-threatening. Polyuria/polydipsia, visual disturbances, hyperprolactinemia, and hypogonadism related symptoms , on the contrary, are less frequent in ICI hypophysitis (Table 3), likely due to the milder and often transient enlargement of the pituitary gland [72].

Analysis of the pituitary function at the time of diagnosis reveals interesting dichotomies, and offers insights into disease pathogenesis. ACTH deficiency is significantly more prevalent in ICI hypophysitis (237 of 248, 96%) than primary hypophysitis (519 of 859, 60%, p<0.0001, Table 3 and Figure 3). Similar finding for the TSH deficiency (145 of 232 versus 436 of 871, p= 0.003, Table 3). On the contrary, central diabetes insipidus, prolactin abnormalities (hyperprolactinemia or prolactin deficiency), and GH deficiency are more common in primary hypophysitis and infrequently found in ICI hypophysitis (Table 3 and Figure 3B). The broader involvement of the various hormonal axes suggests a more mature and diversified immune response in primary hypophysitis, a disease that likely developed for quite some time (likely measured in years) before being diagnosed. On the contrary, the more selective involvement of the ACTH and TSH axes in ICI hypophysitis points toward an immune toxicity that unfolds more rapidly (in weeks).

Brain MRI shows pituitary abnormalities in the overwhelming majority of primary hypophysitis patients (990 of 929, 97%), whereas is reported as normal in about a third of ICI hypophysitis patients (69 of 194, 36%, p < 0.0001, Table 3). This difference in part reflects the milder severity of ICI hypophysitis, which rarely features compression of the optic chiasm or thickening of the stalk. But it may also relates to the fact that baseline brain MRI are not typically taken in cancer patients undergoing immunotherapy, so subtle enlargement of the pituitary gland, albeit present, may not be detected without comparison with previous images [73]. In addition, ICI hypophysitis can resolve rapidly upon glucocorticoid treatment and therefore have a normal MRI appearance according to timing of the scan [74]. A normal brain MRI, thus, does not rule out ICI hypophysitis.

ICI hypophysitis is diagnosed on clinical and imaging grounds in most patients (209 of 261, 78 %), or clinical findings only in the remaining ones. None of the published patients underwent pituitary biopsy, and therefore pathological data about this condition are lacking. Biopsy is indeed not indicated, unless there is evidence of other pituitary pathologies or metastasis [74]. Pituitary pathology was available from the autopsy of a patient with mesothelioma who developed hypophysitis and died after tremelimumab administration [4]. Pathology showed a near complete destruction of the anterior pituitary gland, which became largely replaced by necrotic tissue, resembling what is seen in the necrotizing form of primary hypophysitis. Primary hypophysitis, on the contrary, is mainly diagnosed by surgical pathology (814 of 1,342, 61%), clinical and imaging criteria (450, 33%), autopsy (43, 3 %), or clinical findings only (35, 3%).

The treatment and management of hypophysitis is, at this time, symptomatic. It includes three main approaches: mass reduction with or without replacing the defective endocrine axes, hormonal replacement therapy only, and watchful waiting. Mass reduction with or without hormonal replacement therapy is the most commonly used approach in both primary and ICI hypophysitis (675 of 880, 77% vs 142 of 242, 59%). It can be achieved by surgery, lympholytic drugs, or radiotherapy/stereotactic radiosurgery. Surgery is the most common form of treatment in primary hypophysitis, being performed in 491 of 675 patients (73%). Lympholytic drugs such as high-dose glucocorticoids, on the contrary, have been the treatment of choice in ICI hypophysitis (142 of 142, 100 % vs 178 of 675, 26%, p < 0.0001). Recent studies, however, suggest they not improve the outcome of ICI hypophysitis and may negatively impact the overall survival of melanoma patients [72,73,75]. High-dose glucocorticoids should be considered to patients presenting with visual disturbances, and possible severe headache or other autoimmune side-effects [74]. Radiotherapy or stereotactic radiosurgery is reserved to a few primary hypophysitis patients (6 of 672, 0.9 %), typically those with refractory disease.

Hormonal replacement therapy only, particularly glucocorticoid replacement, is more common in ICI treated patients than in those with primary hypophysitis (95 of 242, 39% vs 174 of 879, 20%, p < 0.0001). A watchful waiting approach has been used in limited cases of both primary and ICI hypophysitis (31 of 880, 3% vs 5 of 342, 2%).

Follow-up information was available in about half of the published patients. Although variability was large, follow up time is significantly longer in primary than ICI hypophysitis (median 1.2, iqr 2.5 vs median 0.8, iqr 1.6 years p = 0.0002, Figure 3C). During follow-up, the ICI treatment is withdrawn in about 40 % of patients (49 of 123) and resumed only in 15 patients.

Outcomes of patients with ICI hypophysitis is variable (Figure 3D). Long-term hormonal replacement is more frequently required in ICI hypophysitis patients than in those with primary hypophysitis (145 of 164, 88 % vs 382 of 627, 61 %, p < 0.0001). Pituitary deficit recovery after mass reducing treatment is uncommon in ICI hypophysitis (10 of 164, 6 %), whereas it occurs in about a fourth of the primary hypophysitis patients. Spontaneous resolution is not reported in ICI treated patients, and very rare (2%) in those with primary hypophysitis. Recurrence of hypophysitis is unusual in ICI hypophysitis and more frequent in the primary forms (1of 164, 1 % vs 44 of 627, 7%, p < 0.0001), a difference that may be biased by the shorter period of follow-up of ICI treated patients. Death is reported with a frequency of about 5% in both ICI and primary hypophysitis, although all eight patients treated with ICI died because of underlying, tumor-related causes.

4.2. Comparison between ICI hypophysitis caused by CTLA-4 blockade and that caused by PD1/PD-L1 blockade

Although the number of published hypophysitis cases caused by PD1/PD-L1 blockade is smaller than those caused by CTLA-4 blockade (69 vs 192, Table 4), more and more PD1-related cases are appearing in the literature, providing sufficient data for interesting comparisons (Figure 4A). These comparisons are mainly centered on clinical presentation, diagnosis, treatment, and outcome (Table 4).

Figure 4.

Comparison of ICI hypophysitis caused by CTLA-4 blockade to that caused by PD1/PD-L1 blockade. A) Number of published patients with hypophysitis secondary to CTLA-4 or PD-1/PDL-1 blockade from 2003 to 2019. B) Time of hypophysitis onset according to the type of ICI and the treatment regimen (mono- or combination therapy). C) Distribution of patients with hypophysitis secondary to CTLA-4 or PD-1/PDL-1 blockade by hormonal deficiencies at diagnosis D) Distinctive features of hypophysitis secondary to CTLA-4 or PD-1/PDL-1 blockade.

Hypophysitis from CTLA-4 blockade develops on average 10.5 weeks after the first administration, with a very narrow 95% confidence interval ranging from 9.8 to 11.2 weeks (Figure 4B). This time of presentation is significantly smaller than that observed with PD-1 blockade (27 weeks, 95% CI from 20.9 to 33.1, p< 0.001), or PD-L1 blockade (27.8 weeks, 95% CI from 0 to 58, p= 0.001), and instead no different from that seen when a combination therapy is used (10.3 weeks, 95% CI from 4.4 to 16.2, p= 0.876). In addition, the variability among the PD-1 and PD-L1 categories was enormous, with a case presenting within a week from the first injection and another more than 3 years later. This prolonged and delayed mode of action of anti-PD-1 drugs could be ascribed to a distinct pathogenetic mechanisms, in keeping with the notion that nivolumab remains bound to PD-1 on T cells for over 2 months following a single injection, regardless of the dose [76]. It could also reflect an ascertainment bias, since the rarity and more challenging diagnosis can delay its identification. These striking differences suggest that hypophysitis caused by CTLA-4 blockade or PD1/PD-L1 blockade are different diseases, arising through distinct mechanisms.

Symptoms from hypocortisolism are more frequent with PD-1 than CTLA-4 blockade (91% vs. 75%; p=0.005). On the contrary, symptoms of hypothyroidism (21% vs 7%; p=0.010), hypogonadism (16% vs. 0%; p=0.001), headache (60% vs. 4%; p<0.001) and visual disturbances (8% vs. 0%; p=0.015) are more common in hypophysitis induced by anti-CTLA4.

The diagnosis of hypophysitis in patients treated with PD-1 blockade is largely made on clinical grounds only (67% vs. 6%; p<0.001), whereas clinical and radiological tools are more frequently used to diagnose hypophysitis after CTLA-4 blockade (93% vs. 33%; p<0.001). Radiologic abnormalities such as pituitary enlargement, stalk widening, or altered contrast enhancement at MRI, are in fact present in most anti-CTLA-4 treated patients (81%) and consistent with the greater incidence of headache and hypopituitarism. They are instead seen in less than a fifth of the anti-PD-1/PD-L1 treated ones (18%; p<0.001), a difference that could reflect a delay in hypophysitis diagnosis for the PD-1 patients where the initial phase where radiological abnormalities would be present is missed. Curiously, two patients treated with PD-1 blockade featured a mild pituitary atrophy at MRI [77,78], possibly presenting a later stage of the disease after a morphologically silent phase.

Independently of the type of ICI used, the functional defects predominantly involve the anterior hypophysis, whereas diabetes insipidus is rare in both anti-CTLA4 and anti-PD-1/PD-L1 hypophysitis, with no differences between the groups (2% vs. 3%; p=0.747). Secondary adrenal insufficiency is the most common endocrine abnormality in both anti-CTLA-4 and anti-PD-1/PD-L1 hypophysitis patients, with no difference among the groups (95% vs. 97%; p=0.400). CTLA-4 patients, however, develop more frequently oligo- or pan-hypopituitarism, featuring TSH impairment (85% vs. 4%; p<0.001) or gonadotroph cells impairment (75% vs. 13%; p<0.001) that instead were rarely seen with PD-1 blockade (Figure 4C). The high selectivity for ACTH in anti-PD-1 treated patients appears to go along with a more severe hypocortisolism, as reported by a recent multi-centered study [66,71]. As discussed above, this unique sensitivity of ACTH-secreting cells to the damage caused by PD-1 blocking antibodies likely reflects a specific mechanism of action that remains to be detected.

The treatment of ICI hypophysitis is medical independently of the type of ICI, and almost exclusively based on the use of glucocorticoids. CTLA-4 patients, however, receive more commonly pharmacological doses to reduce the size of the pituitary gland, whereas replacement doses are prescribed more frequently in anti-PD1 induced hypophysitis (81% vs. 23%; p<0.001).

After a diagnosis of hypophysitis, ICI administration is maintained in most patient of both groups (70% in PD-1 blockade vs. 56% in CTLA-4 blockade; p=0.088). When suspension is requested, CTLA-4 blockade is more often associated to permanent (41% vs. 10%; p<0.001), and PD-1 blockade (20% vs. 3%; p=0.002) to temporary, discontinuation. This different medical practice further supports the hypothesis that these two forms of hypophysitis arise through different mechanisms.

Most patients with hypophysitis secondary to ICI required long-term hormone replacement therapy, with no significant differences between anti-CTLA-4 and anti-PD-1/PD-L1 groups (89% vs. 90%; p=0.345). However, only short/medium term follow-up was available for most patients and in many cases the hormone withdrawal attempt was postponed after the end of oncologic treatment. Longer follow-up studies are needed to clarify whether the hypopituitarism induced by ICIs is permanent and whether there are differences between the type of ICI. Independently from the type of treatment, no recurrence of hypophysitis is reported in any of the 276 published patients. Eight deaths recorded during follow up (6 in CTLA-4 group, 5% and 2 in PD-1/PD-L1 group, 4%, Table 4), all ascribed to the underlying cancer.

4.3. ICI hypophysitis caused by combination immunotherapy

Patients on combination immunotherapy are at higher risk of developing hypophysitis compared with those treated with monotherapy [67]. Interestingly, Rassy et al. have recently showed a sub-additive effect of ICIs–chemotherapy combinations on grade 3 to 5 adverse events [79]. In our analysis, unfortunately, the number of patients on combination therapies was too small (11, 3,9 %, Table 1) for definitive conclusions or comparisons with hypophysitis patients treated with monotherapy.

5. Practical approaches to the diagnosis and management of ICI hypophysitis

ICI hypophysitis requires a high degree of suspicion to be diagnosed. As indicated in the preceding sections, symptoms can be subtle, time of onset variable, MRI findings minimal, and endocrine abnormalities of diverse extent and severity. Prompt diagnosis is, however, very important for cancer patients treated with immunotherapy because correction of hypophysitis markedly improves their quality of life. Considering that pituitary defects of ICI hypophysitis are often long-lasting, the need for appropriate and personalized endocrine care is evident, as recently reviewed by Chiloiro et al. that designed an elegant flow-chart of diagnosis and treatment of ICI hypophysitis according to their personal experience [80].

In this section we will offer practical considerations about the most distinctive clinical features of ICI hypophysitis, the proper medical treatment, and when to suspend or discontinue the ICI treatment.

The most characteristic clinical and endocrinological abnormality in ICI hypophysitis is, in our opinion, the secondary (central) adrenal insufficiency. The defective release of ACTH from the corticotroph cells of the adenohypophysis causes the reduced production of cortisol from the adrenal glands, which is the basis for the symptomatology. The onset of fatigue (asthenia, weakness, tiredness, exhaustion) in a cancer patient treated with ICI should alert the clinician about hypocortisolemia. Similarly, loss of appetite, malaise, nausea and vomiting, abdominal or muscular pain are signs and symptoms of adrenal insufficiency. More rarely, patients have unexplained fever or varying degrees of cognitive dysfunction.

The cornerstone to establish a biochemical diagnosis of hypocortisolism is the measurement of basal serum cortisol and ACTH in the morning (not later than 8AM). Mornings are chosen because of the circadian rhythm of CRH, ACTH, and cortisol that reach the highest level at 6 AM and the nadir at 11 PM. Undetectable cortisol with low (or inappropriately normal) ACTH is diagnostic of secondary adrenal insufficiency. When cortisol is low but still measurable (between 3 and 15 μg/dL), a stimulation test of the adrenal cortex is performed using synthetic ACTH, such as cosyntropin or tetracosactrin (polypeptides containing the first 24 of the total 39 amino acids that make up the natural ACTH). Following the basal cortisol measurement and the intra-muscular injection (0.25 mg) of synthetic ACTH, stimulated cortisol is then measured at 30 and 60 min. An increase in cortisol greater than 18 μg/dL at any time point indicates a normal ability of the adrenal cortex to produce cortisol. In primary adrenal insufficiency (for example the one caused by Addison disease), this cortisol response is abolished. In secondary adrenal insufficiency (as the one caused by ICI hypophysitis), the cortisol response is also impaired but dependent upon timing. The adrenal cortex, in fact, becomes atrophic when ACTH levels are low, but the atrophy takes time to fully develop, typically around 6 weeks. Therefore, the ACTH stimulation test may yield a normal cortisol increase if the secondary adrenal insufficiency is of recent onset [81,82]. Occasionally, a more global stimulation test of the hypothalamo-pituitary-adrenal axis is performed to diagnose adrenal insufficiency. The gold standard used to be the insulin tolerance test, but it may cause severe (even fatal) hypoglycemia and thus is not recommended in this setting. A safer alternative is the CRH stimulation test, which assess the response of both ACTH and cortisol to the stimulant. This test has been used in some ICI hypophysitis patients but the definition of an inadequate response to diagnose adrenal insufficiency remains controversial.

In addition to measuring the basal (and/or stimulated) levels of cortisol and ACTH, we recommend to monitor the serum sodium levels in cancer patients treated with ICI. In fact, hyponatremia (defined as a serum sodium concentration below 135 mEq/L), although most commonly caused by kidney diseases that impair their ability to excrete water, recognizes numerous causes [83], one of them being adrenal insufficiency. Low cortisol induces hyponatremia by increasing the levels of ADH through several mechanisms (Figure 5) [84]. It stimulates ADH both directly and indirectly via stimulation of hypothalamic CRH [85]. Low cortisol also leads to a reduction of blood pressure and cardiac output that in turn stimulate the baroreceptor-mediated release of ADH [86]. In addition, low cortisol upregulates the expression of aquaporin-2 water channels in the kidney tubuli, thus increasing their sensitivity to the action of ADH [87]. Finally, elderly patients have an increased release of ADH independent of cortisol, and are thus more prone to develop hyponatremia [88]. The increased ADH levels and action cause a hypotonic hyponatremia that is similar to that seen in cancer patients, especially lung cancer patients, whose tumor cells ectopically produce ADH, leading to the syndrome of inappropriate secretion of ADH (SIADH). The distinction between the two conditions is critical in order to initiate the correct treatment: fluid restriction in SIADH or corticosteroid replacement in adrenal insufficiency. Although serum sodium was not systematically documented in all articles that described the ICI hypophysitis patients presented here, hyponatremia was mentioned in 70 of the 276 (25%) patients. Interestingly, we noted that hyponatremia was three times more common in patients treated with anti-PD1 (31 of 64, 48%) than in those treated with anti-CTLA4 (32 of 192, 16%, p< 0.0001). This difference may simply reflect the exquisite sensitivity of ACTH-secreting cells to damage by anti-PD1 antibodies, or a yet-to-be uncovered mechanism.

Figure 5.

Mechanisms of hyponatremia in ICI hypophysitis. Low cortisol induces hyponatremia by increasing the levels of ADH through several mechanisms (direct stimulation of ADH, stimulation of CRH release, reduction of blood pressure and cardiac output, and up-regulation of aquaporin-2 expression in the kidney tubuli). Increased serum ADH levels lead to free water retention and hypotonic hyponatremia.

The treatment of secondary adrenal insufficiency is based on glucocorticoids, and their type and mode of administration dependent on timing [89]. When the onset is acute, also called adrenal crisis or Addisonian crisis, this represents a life-threatening medical emergency [90]. It requires the prompt im or iv administration of hydrocortisone (100 mg), followed by either iv saline infusion and repeated doses of hydrocortisone (50-100 mg every 6 hours) or continuous hydrocortisone infusion (at a rate of 1–2 mg/hour). Over-replacement in the short term does not appear to be detrimental. For long-term replacement, the daily dose of hydrocortisone can be calculated based on the mean total daily production of cortisol which is 5-10 mg/m², corresponding to a total dose of 15–20 mg daily. Usually, the daily dose is split into three administrations, using 10 mg in the morning, 5 mg at lunch time, and 5 mg in the early evening (no later than 6 pm to avoid insomnia or polyuria). This regimen better approximates the circadian rhythm and reduces the risk of glucocorticoids side effects [81,82].

Cortisone acetate, which requires reduction by 11 β-hydroxysteroid dehydrogenase to produce biologically active cortisol, can also be used for long-term replacement. It is given in two daily doses (25 mg in the morning and 12.5 mg in midafternoon), an administration scheme that may reduce circulating cortisol.

Thyrotropin and gonadotropin deficiencies are also frequent at diagnosis, but are often reversible, thus the initiation of a replacement treatment is less urgent and can be delayed under close clinical and biochemical follow-up. Levothyroxine replacement therapy should be started only after excluding adrenal insufficiency, and dosed to maintain serum free thyroxine in the middle-to-upper part of the normal range. Blood for the measurement of thyroid hormones should be drawn before levothyroxine administration [81]. Depending on the age of the patient and presence of contraindications, hypogonadism can be corrected by the administration of synthetic sex hormones in order to prevent bone loss and osteoporotic fractures in women, and muscular mass in men.

A topic that has received great attention is the dose of glucocorticoids that should be given to cancer patients once a diagnosis of ICI hypophysitis is established. Early cases were treated with the high doses that are typically used in patients with primary hypophysitis. But more recent studies suggest that high doses of glucocorticoids are not necessary, and associated with reduced efficacy of ICI and poorer prognosis of the underlying cancer [75]. Unless the patient presents with severe neurological symptoms (such as headache or visual field defects), we also endorse the use of replacement glucocorticoids doses, rather than anti-inflammatory doses.

A final practical point concerns the issue of whether to stop the ICI treatment or not once a diagnosis of ICI is established. It naturally depends on the individual patient situation, but our review of the literature and personal experience suggest not to interrupt the ICI cycles. The cycles can be temporarily suspended until the acute endocrinological situation is resolved and appropriate hormone replacement instituted, but most patients should be able to resume them and continue as planned. These suggestions are in line with the most recent guidelines [91-93].

6. Conclusion

Hypophysitis secondary to cancer immunotherapy is an emerging clinical entity that requires prompt diagnosis and treatment. A multidisciplinary team is ideal to manage these patients. Identification of serological markers that could predict the onset and/or outcome of ICI hypophysitis are desirable.

7. Expert opinion

Hypophysitis developing in cancer patients treated with ICIs is an increasingly recognized and complex clinical entity. We now recognize that the form developing upon CTLA-4 blockade features headache, mild pituitary enlargement, and some degrees of pan-hypopituitarism; whereas the one from PD-1 blockade is mostly characterized by isolated and severe ACTH deficiency, no mass effect symptoms, and no imaging abnormalities. Recognition of these distinctive features will aid the clinical diagnosis, and avoid unnecessary and detrimental administration of high dose glucocorticoids. Similarly, serial measurements of TSH, screening for the presence of pituitary antibodies, and comparison of pituitary imaging (by MRI or positron emission topography) before and during ICI treatment are useful diagnostic tools [72,94]. Nevertheless, diagnosis and treatment of this condition remain a challenge. An obstacle is the lack of serological markers that are specific for hypophysitis, such as the availability of serum antibodies directed against unique pituitary autoantigens. A very recent paper [95] reported two novel candidate autoantigens in this patient population: integral membrane protein 2B, and guanine nucleotide-binding protein G(olf) subunit alpha [95]. They were recognized by serum antibodies more readily in cases with hypophysitis (No. = 8) than in controls without it (No. = 21). It is important to note, however, that these two proteins are not uniquely expressed in the anterior pituitary gland ITM2B has the highest levels in the thyroid gland, and GNAL in the caudate nucleus), and that the cDNA library used to test the patient serum reactivity was a commercial library prepared from the whole human brain (not the pituitary gland). Thus, confirmation of these candidate autoantigens is needed, but the study first shows the utility of using specific serum autoantibodies to diagnose ICI hypophysitis. We expect that more and more studies of this nature will appear during the next decade. Another obstacle in the field of immune toxicities from cancer immunotherapy is our lack of understanding about organ selectivity of the immune response. As in the primary forms of autoimmune diseases (the ones that develop independently of cancer immunotherapy), we still do not know why a particular organ is preferentially targeted by the immune response as opposed to other organs (the so-called conundrum of specificity). For example: why is the pituitary gland uniquely damaged in patients undergoing CTLA-4 blockade as opposed to other glands such as the adrenals and the pancreatic islets? The understanding of this target organ selectivity has eluded immunologists for decades. The relative rapidity of onset of the immune-related adverse events seen after cancer immunotherapy and the fact that for these patients we have a baseline where no immune abnormalities are present (the time before the start of cancer immunotherapy) grant us a unique opportunity: a view of the human immune system that was never possible for patients with primary autoimmune diseases. This view will generate remarkable advances regarding the functions of our immune system during health and disease. We also expect to see more experimental studies devoted to elucidate the link between immune checkpoints blockade and development of hypophysitis. The pathogenesis of this condition, in fact, remains unknown. Unraveling the disease mechanism(s) will be a pivotal step toward the development of diagnostic and prognostic biomarkers, as well as the identification of patients “at risk” for ICI hypophysitis and other immune related adverse events. Finally, larger prospective studies aimed to refine current diagnostic and therapeutic protocols are likely to appear in the near future, all geared toward a personalized treatment that for most patients will consist of long-term hormone replacement and continuation of the ICI regimens.

Article highlights:

• Hypophysitis secondary to immune checkpoint inhibitors (ICIs) is an emerging and important problem in cancer patients treated with this form of immunotherapy.

• The mechanisms of pituitary toxicity following ICI administration remain to be elucidated. For the CTLA-4, an “ectopic” expression of this immune molecule on adenohypophyseal cells has been demonstrated and proposed as a disease pathway. For the PD-1, experimental articles have yet to be published, although its “ectopic” expression on corticotrophs cells could be a mechanism considering the exquisite predilection for ACTH-deficiency these patients exhibit.

• ICI hypophysitis has some distinctive features from primary hypophysitis, such as male predominance, older age at diagnosis, shorter time of onset, severe hypocortisolism, rarity of posterior pituitary and optic chiasm involvement, and often persistent hypopituitarism.

• Hypophysitis caused by CTLA-4 blockade is also distinct from that caused by PD-1/PD-L1 blockade, likely reflecting different mechanisms of toxicity. Hypophysitis from CTLA-4 blockade often leads to pan-hypopituitarism and is associated with mild pituitary enlargement. The one from PD-1 blockade, instead, is characterized by isolated and severe ACTH deficiency, no mass effect symptoms, and no imaging abnormalities.

• The diagnosis of ICI hypophysitis diagnosis requires a high degree of suspicion and is based mainly on clinical and imaging grounds. It remains a challenge, also considering that specific serological markers are lacking.

• The mainstay of treatment is hormonal replacement. The use of high dose of glucocorticoids should be reserved to cases with severe headache and visual disturbances. ICI can be suspended or delayed until the acute symptomatology is resolved.