Autoantibody Development under Treatment with Immune-Checkpoint Inhibitors

Immune-checkpoint inhibitors (ICIs) have improved the previously dismal prognosis of patients with various types of cancer, but at the cost of immune-related adverse events (irAEs), including arthritis, colitis, hepatitis, and various endocrinopathies (1). ICIs inhibit negative costimulatory signals to T cells, thereby enhancing antitumor T-cell responses (2). Because this mode of action is not antigen-specific, ICIs may also (re)activate otherwise dormant autoreactive T cells. This, in turn, might lead to a break in T-cell tolerance to not only tumor antigens but also autoantigens, resulting in activation of autoreactive B cells and ultimately the formation of autoantibodies. If true, the occurrence of autoantibodies may be associated with more frequent irAEs. Production of autoantibodies may indicate enhanced global immunogenicity, which may, in turn, be associated with better antitumor responses, as has been reported for changes in the T-cell repertoire (3–5). Therefore, we determined if autoimmune disease-associated autoantibodies were formed with ICI treatment and investigated their association with irAEs and clinical outcome.

For this analysis, we included 133 patients with late-stage melanoma who were treated with ipilimumab, a CTLA-4 inhibitor, and for whom pre- and posttreatment serum or plasma samples were available. Patients were treated with a maximum of four cycles of ipilimumab 3 mg/kg in an expanded access program or according to the label after approval at the Netherlands Cancer Institute or the Leiden University Medical Center. Patients were included if they were at least 18 years of age and had histologically or cytologically proven irresectable stage IIIc or IV melanoma, with measurable metastatic lesions according to the RECIST 1.1 criteria. Patients were treated with four cycles of intravenous 3 mg/kg ipilimumab every 3 weeks. Sixty-six (49.6%) patients were treated with anti–PD-1 therapy following ipilimumab: either 2 mg/kg intravenous pembrolizumab every 3 weeks or 240 mg intravenous nivolumab every 4 weeks. The study was conducted in accordance with the Declaration of Helsinki after approval by the institutional review boards of both centers. All patients signed informed consent for withdrawal of extra blood samples for biomarker analysis. According to the study protocol, serum or plasma for autoantibody determination was collected before initiation of ipilimumab treatment and 12 weeks after. Pre- and posttreatment serum was snap-frozen and stored at −80°C until autoantibody determination. Due to failed measurements, post-ipilimumab autoantibody status could not be determined in 4 patients (Supplementary Fig. S1).

Autoimmune hepatitis and primary biliary cirrhosis-associated antismooth muscle, antimitochondrial, and anti-liver/kidney microsome (LKM) antibodies were measured by indirect immunofluorescence assay (IFA) using mouse liver/kidney/stomach substrate (Aesku). Antinuclear antibodies (ANA) were determined in all patients by IFA using the HEp-2000 ANA Test System which uses human epithelioid cells stably transfected with the SSA/Ro autoantigen, cultured, and fixed directly on the test wells (Immuno Concepts). Patient serum samples at a dilution of 1:40 were incubated with antigen substrate for 30 minutes at room temperature to allow specific binding of autoantibodies to cell nuclei. After washing with phosphate-buffered saline to remove nonspecifically bound antibodies, the substrate was incubated with an anti-human antibody conjugated to fluorescein. After another washing step, the nuclear staining pattern was read using the international consensus on antinuclear antibody pattern (ICAP; ref. 6) by two experienced, independent readers trained in ANA-pattern reporting and blinded to time order and patient data of samples. In the case of lack of consensus, a third reader functioned as tiebreaker. All system reagents, conjugates, calibrators, and positive and negative controls were provided by and used according to instructions of the manufacturer (Immuno Concepts). All steps of the IFA were conducted using a Helmed fully automated IFA slide processor (Aesku).

Anticyclic citrullinated peptide 2 (CCP2) IgG, rheumatoid factor (RF) IgM, antigliadin IgG, and (if ANA was positive by IFA) antibodies to extractable nuclear antigens (ENA) were determined by EliA technique on a Phadia ImmunoCap 250 instrument (Thermo Fisher Scientific). This is a fully automated and high-throughput fluorescence enzyme immunoassay system used for routine diagnostic laboratory testing. The fluorescence signal of measured serum samples is compared with calibrators with known concentrations. For anti-CCP2 IgG, citrullinated synthetic peptides (second-generation antigen) were used as antigen, for RF IgM, aggregated rabbit IgG was used, for antigliadin IgG, synthetic deamidated gliadin peptides were used, and for ENA, a Symphony Well of various antigens was used: human recombinant U1RNP (RNP70, A, C), SS-A/Ro (60 kDa, 52 kDa), SS-B/La, Centromere B, Scl-70 protein, Jo-1 protein, and native purified Sm proteins. Anti-ENA–positive patients were further assayed for the following specific ENA antibodies by EliA (coated antigens in parentheses): anti-SSA (human recombinant SS-A/Ro (60 kDa, 52 kDa) proteins), anti-SSB (human recombinant SS-B/La protein), anti-RNP70 (human recombinant RNP70 protein), anti-U1RNP (human recombinant U1RNP (RNP70, A, C) proteins), anti-Smith (synthetic SmD3 peptide), anti-Jo1 (human recombinant Jo-1 protein), anti-CENP (human recombinant centromere protein B), anti-PMSCL (human recombinant PM-Scl protein), anti-RNAP3 (human recombinant RNA polymerase III protein), anti-Scl70s (human recombinant Scl-70 protein). All system reagents, conjugates, calibrators, and positive and negative controls were used according to the manufacturer's instructions.

Antithyroid peroxidase (TPO), antithyroglobulin (TG), and, in ANA-positive patients, anti-dsDNA were determined by noncompetitive chemiluminescent immunoassay (CLIA) using Immulite 2000 (Siemens Healthineers). These assays use a luminescent adamantyl dioxetane phosphate tracer and were performed using reagents provided by the manufacturer according to instructions in the package insert.

Information about demographics, treatment response, survival status, and the occurrence of irAEs was obtained from retrospective review of medical records. irAEs were recorded starting from the first ipilimumab treatment until one year later, death, or the start of different therapy (whichever occurred first), using Common Terminology Criteria for Adverse Events version 4.03: any grade arthralgia/arthritis, colitis, hypophysitis, primary adrenal insufficiency, primary thyroid dysfunction, dermatitis (rash, vitiligo, or psoriasis), uveitis, or grade 3–4 hepatitis. Primary thyroid dysfunction as an irAE during anti–PD-1 treatment (nivolumab or pembrolizumab) following ipilimumab treatment was determined in the same manner. Hematologic and serum parameters necessary for making the above diagnoses were determined at baseline, every 3 months during follow-up, at progressive disease, and according to the treating oncologist's clinical judgment. Three patients had preexisting hypothyroidism. Thyroid dysfunction was registered only as an irAE in these patients if symptoms were aggravated and a new medical intervention was indicated. Two of the four cases of arthralgia/arthritis constituted a flare of preexisting rheumatoid arthritis (RA). Survival was defined as time from start of ipilimumab to death of any cause, recorded between start of first ipilimumab treatment until January 2018, for a median follow-up time of 20.4 months (IQR: 8.8–40.8). Radiologic evaluation (CT or PET/CT scanning) was performed at baseline, week 12, and subsequently every 3 months until progression. Response was scored according to RECIST 1.1 criteria. Best overall response was defined as the best response recorded from start of ipilimumab until date of progression, death, or the start of a different therapy (whichever occurred first). Patients achieving a partial response or complete response were considered responding patients.

We used McNemar test for paired data to test whether autoantibody positivity increased post-ipilimumab. Frequencies of irAEs in patients who developed antibodies versus those who did not were compared using Fisher exact tests. To test whether post-ipilimumab autoantibody positivity was associated with (i) the development of any irAEs under ipilimumab, primary thyroid dysfunction under subsequent anti–PD-1 therapy, or better overall response, and (ii) overall survival, binary logistic regression and Cox proportional hazards regression, respectively, were used, and adjusted for age, gender, treating hospital, and number of ipilimumab cycles received. All analyses were conducted using Stata statistical software, Special Edition, release 14.1 (StataCorp LP).

Mean age was 59 years [standard deviation (SD): 14], and 62% of patients were male. Of 127 patients with complete pre-ipilimumab autoantibody data, 26 (20%) were positive for any of the autoantibodies before treatment. In total, 29% (36/125) of patients with complete autoantibody data were autoantibody-positive after ipilimumab treatment. Of patients who were fully autoantibody-negative before ipilimumab treatment, 19.2% (19/99) developed any autoantibodies posttreatment (P < 0.0001). Predominantly anti-TPO (4.8%, 6/125) and anti-TG (6.0%, 8/132) appeared in patients who were negative for these autoantibodies at baseline (P = 0.03 and P = 0.008, respectively). For all other autoantibodies, posttreatment positivity did not greatly change (Supplementary Fig. S1).

A nonsignificant association was seen between the development of any autoantibody and irAEs: 15/19 (78.9%) patients who developed any autoantibody experienced irAEs compared with 46/80 (57.5%) patients who did not develop autoantibodies [OR, 2.92; 95% confidence interval (CI) 0.85–10.01; Fig. 1]. When disregarding autoantibody status pre-ipilimumab, patients with autoantibodies posttreatment also experienced more irAEs (Supplementary Fig. S2). No significant association between pre-ipilimumab autoantibody positivity and irAEs was observed: 8/26 (31%) patients who were autoantibody-positive pre-ipilimumab experienced irAEs compared with 38/100 (37%) patients who were autoantibody-negative pre-ipilimumab (OR, 1.61; 95% CI, 0.62–4.18; P = 0.33).

In a prespecified subgroup analysis, we focused only on the irAEs related to the tested autoantibodies (arthritis/arthralgia, hepatitis, thyroid dysfunction, colitis, adrenal insufficiency, dermatitis, or sicca symptoms). In this analysis, 14/19 (73.7%) patients who developed autoantibodies had irAEs related to the tested antibodies compared with 37/80 (46.3%) patients who did not develop autoantibodies, indicating a significant association between the development of autoantibodies and irAEs (OR, 3.64; 95% CI, 1.13–11.75). However, the appearance of a specific autoantibody did not associate with the occurrence of an irAE in the organ system affected by the disease for which the specific autoantibody has diagnostic value (Table 1).

We hypothesized that autoantibody development with ipilimumab treatment might predispose patients to irAEs during subsequent anti–PD-1 therapy. Following progression on ipilimumab treatment and after exclusion of patients who had thyroid dysfunction with ipilimumab (n = 12), 61 (50.4%) patients received anti–PD-1 therapy. In these patients, we found a significant association between the development of thyroid autoantibodies while on ipilimumab and subsequent thyroid dysfunction under PD-1 blockade: 4/9 (44.4%) patients who developed thyroid autoantibodies with ipilimumab and subsequently received anti–PD1-therapy had thyroid dysfunction under anti–PD-1 compared with only 7/48 (14.6%) patients who did not develop autoantibodies (OR, 6.26; 95% CI, 1.07–36.5; P = 0.04). The association between the development of thyroid autoantibodies while on ipilimumab and subsequent thyroid dysfunction under PD-1 blockade was even stronger when autoantibody status pre-ipilimumab was disregarded and all anti–PD-1-treated patients were included in the analysis (n = 60; one patient missing anti-TPO measurement): 7/11 (54.6%) patients who had thyroid autoantibodies after ipilimumab treatment developed thyroid dysfunction under anti–PD-1 compared with 7/49 (14.3%) patients who did not develop autoantibodies (OR, 9.96; 95% CI, 1.94–51.1; P = 0.006).

We next investigated the association of autoantibody development following the initial ipilimumab treatment with survival and response. During the median follow-up time of 20.4 months (IQR, 8.8–40.8), 92 (69%) patients died after a median of 11.2 months (IQR 7.3–21.9), 87 patients (65%) had stable or progressive disease, and 46 patients (35%) achieved complete or partial response. Patients who developed autoantibodies had a minor survival benefit compared with those that stayed autoantibody-negative, although this was not significant (HR for all-cause death: 0.66; 95% CI, 0.34–1.26; P = 0.21; Fig. 2). There was no significant association between the presence of a specific autoantibody and survival (Supplementary Table S1).

There was also a trend toward an association between development of any autoantibody and treatment responses (OR for response: 2.64; 95% CI, 0.85–8.16; P = 0.09). When assessing specific autoantibodies, only the development of thyroid autoantibodies was significantly associated with treatment response (OR for response: 5.43; 95% CI, 1.38–21.4; P = 0.02).

To determine whether the observed associations between autoantibody development and clinical outcomes were due to an association of irAEs with both entities, we also investigated the association between irAEs and clinical outcomes. No association between occurrence of irAEs and survival or treatment response was found (HR for all-cause death: 1.12; 95% CI, 0.70–1.79], P = 0.64; OR for response: 1.53; 95% CI, 0.68–3.46, P = 0.31). The estimates of the association between autoantibodies and survival and treatment response reported above did not greatly change after adjusting the analyses for occurrence of irAEs: HR for all-cause death: 0.65; 95% CI, 0.34–1.25, P = 0.20; OR for response: 2.50; 95% CI, 0.80–7.84, P = 0.12. All associations of autoantibody status with survival and treatment response were similar when autoantibody status pre-ipilimumab was disregarded and all patients were included in the analysis (Supplementary Table S2). Patients who were autoantibody-positive pre-ipilimumab had no survival or response benefit compared with patients who were autoantibody-negative (HR for all-cause death: 1.16; 95% CI, 0.67–2.03, P = 0.59; OR for response: 0.53; 95% CI, 0.19–1.46; P = 0.22).

In this study, we found that ipilimumab treatment induced development of autoantibodies in a fifth of melanoma patients. Our analyses revealed a trend for association between autoantibodies and irAEs under ipilimumab, and a much stronger, significant association between ipilimumab-induced thyroid autoantibodies and thyroid dysfunction under subsequent PD-1 blockade. Lastly, we found a minor survival and response benefit in patients who developed autoantibodies, specifically in those who developed thyroid autoantibodies.

We determined the presence autoantibodies both pre- and posttreatment with ICI therapy and linked these data to irAEs and clinical outcomes. Our results expand previous findings regarding the presence thyroid autoantibodies in patients with ICI-induced thyroid dysfunction (7–10) by showing that these autoantibodies also develop in the absence of overt thyroid dysfunction. Antithyroid antibodies are common in populations without overt thyroid disease, associated with or induced by concomitant autoimmune disease (i.e., type 1 diabetes mellitus, RA, and Celiac disease; refs. 11–17), mutations in CTLA-4 (18, 19), upregulation of MHC class II molecules on thyrocytes leading to thyroid antigen presentation to autoreactive cells (20), or as we show here, ipilimumab treatment. Our data also confirm previous studies reporting that patients rarely develop RA autoantibodies (21–24) or autoimmune hepatitis antibodies even in the presence of the related irAE (25–28).

Our findings demonstrated that the development of thyroid autoantibodies predisposes euthyroid ipilimumab-treated patients to subsequent thyroid dysfunction under anti–PD-1 therapy. The association between thyroid autoantibodies and thyroid dysfunction under anti–PD-1 therapy has been described previously (29, 30). Our results confirm that it is clinically useful to monitor patients with preexisting thyroid autoantibodies closely for thyroid dysfunction with anti–PD-1 therapy.

Although several studies report an association between irAEs and clinical outcome under ICIs (31, 32), we did not find such an association in this study. A relationship between immune-related thyroid dysfunction and clinical outcome has been described previously for various types of cancer immunotherapy, including IL2 (33–35), interferon-2α (35–37), and pembrolizumab (29). Some of these studies also found a response or survival benefit under IL2 (33, 34) or interferon-2α (36) for patients who developed thyroid autoantibodies. These findings are in line with our observations that patients developing thyroid autoantibodies with ICI have a better treatment response.

Our results indicate that CTLA-4 inhibition may lead to loss of B-cell self-tolerance. ICIs execute their function in an antigen-independent manner by diversifying the T-cell repertoire against a multitude of tumor antigens (3, 5). In this study, we found that tolerance to nontumor autoantigens is broken as well and that breaking of tolerance may be associated with signs of clinical autoimmunity, under CTLA-4 inhibition as well as subsequent PD-1 blockade. We did not observe that specific autoantibodies induced disease in their related organ systems, but our results lacked the power to test this hypothesis. Breaking of B-cell tolerance and development of autoantibodies was also associated with better treatment response (statistical trend) and a survival benefit (though nonsignificantly). Previous studies have shown that greater expansion of the T-cell repertoire by ICIs is associated with better response (3, 5, 38). We hypothesize that this expansion is paired with T cell–dependent activation of autoreactive B-cells and autoantibody production. If this is the case, autoantibodies may function as a marker for effective ICI-induced immunogenicity, and it is this enhanced immunity (rather than the autoantibodies themselves) which leads to reactions against both clinically favorable (e.g., tumor) and unfavorable (e.g., nontumor/self) tissues. This could explain the link found in this study between autoantibodies and both treatment response and irAEs.

The main limitation of this study is a lack of power due to the limited number of patients. This may explain why some of our findings failed to reach statistical significance. We also had limited clinical data for which to correct our analyses. However, our study tested all patients treated with ipilimumab (not just the subset that developed irAEs) for a broad panel of autoantibodies in a longitudinal manner, facilitating a pre- and posttreatment comparison of autoantibody prevalence.

We conclude that the development of autoantibodies is common with ipilimumab treatment and that autoantibody presence is associated with the development of irAEs and a trend for better overall response and survival. These results indicate a promising avenue for future research in the quest for biomarkers predicting ICI therapy toxicity and efficacy.

J.B. Haanen reports receiving a commercial research grant from Bristol-Myers Squibb, Mesoscale Discovery, Novartis, and Neon Therapeutics and is a consultant/advisory board member for Bristol-Myers Squibb, Mesoscale Discovery, Pfizer, Novartis, Roche, Ipsen, Bayer, Neon Therapeutics, Immunocore, Seattle Genetics, Gadeta, and Celsius. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E.C. de Moel, E.H. Kapiteijn, J.A. Bakker, T.W.J. Huizinga, J.B. Haanen, R.E.M. Toes, D. van der Woude

Development of methodology: E.H. Kapiteijn, J.A. Bakker, D. van der Woude

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.C. de Moel, E.A. Rozeman, E.H. Kapiteijn, E.M.E. Verdegaal, A. Grummels, J.A. Bakker, J.B. Haanen, R.E.M. Toes

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.C. de Moel, E.H. Kapiteijn, A. Grummels, J.A. Bakker, T.W.J. Huizinga, J.B. Haanen, R.E.M. Toes, D. van der Woude

Writing, review, and/or revision of the manuscript: E.C. de Moel, E.A. Rozeman, E.H. Kapiteijn, E.M.E. Verdegaal, J.A. Bakker, T.W.J. Huizinga, J.B. Haanen, R.E.M. Toes, D. van der Woude

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.C. de Moel, E.H. Kapiteijn, A. Grummels, J.A. Bakker

Study supervision: J.A. Bakker, D. van der Woude

This project was funded by a ZonMw (the Netherlands Organisation for Health Research and Development) Veni-grant (91617100).