Peripheral blood autoantibody signatures might be useful biomarkers of immunotherapy outcome. Signatures predicting melanoma recurrence and toxicity during adjuvant immunotherapy were recently presented. Whether autoantibodies are bystanders or have a pathophysiologic role is unknown, and further efforts are needed to investigate potential mechanisms and determine causation.

See related article by Johannet et al., p. 4121

In this issue of Clinical Cancer Research, Johannet and colleagues report on the predictive value of autoantibody signatures for adjuvant immunotherapy of melanoma (1).

Anti-PD-1 antibody therapy (pembrolizumab or nivolumab) is routinely used in stage III melanoma to reduce the risk of recurrence, with pembrolizumab also approved for stage IIB/C disease (2). HRs for improvement in relapse-free survival (RFS) vary from approximately 0.55 to 0.70 depending on the comparator against placebo (KEYNOTE-054; KN-054) or ipilimumab (CHECKMATE-238; CM-238). For resected stage IV melanoma, an advantage for ipilimumab plus nivolumab was observed compared with nivolumab in the IMMUNED trial. This is in contrast, however, to low-dose ipilimumab plus nivolumab in the CM-915 trial. In management of patients with resected high-risk melanoma, reduction in recurrence must be weighed against toxicity, noting that grade 3–4 adverse events ranged from 25% to 37% with anti-PD1 monotherapy and up to 71% with ipilimumab and nivolumab combination (2).

Whether toxicity to immune checkpoint blockade is a strong surrogate for long-term outcome remains an area of debate though some adjuvant studies (KN-054) have suggested this. Such a connection is potentially consistent with the broad role that the PD1:PD-L1 axis plays in human physiology surrounding T-cell regulation and limitation of autoimmunity. Various mechanisms are proposed to play a role in the development of immune-related adverse events (irAE; ref. 3) including but not limited to cross-reactivity of T cells between tumor and healthy tissue (ir-vitiligo and proposed for ir-myocarditis and ir-orchitis), the expression of immune checkpoints on normal tissue (as demonstrated for CTLA-4 in hypophysis), increased levels of inflammatory cytokines (colitis) and increasing levels of preexisting autoantibodies. Expression of PD1 is not only defined to T cells but is also notably observed on B cells. Here, however, PD1 is a regulator of B-cell activation and blockade increases B-cell activation and the production of inflammatory cytokines. PD1 blockage on memory B cells might further be a mechanism to induce the production of self-antigen reactive autoantibodies. Despite this, the autoantibodies typically described in organ-specific autoimmune diseases are not routinely found in the respective irAE. An exception to this might be immune-related thyroiditis where baseline antithyroglobulin antibodies are associated with the development of thyroid dysfunction.

In the investigation of Johannet and colleagues, 950 pretreatment serum samples from patients in the CM-238 and -915 trials were tested for autoantibodies using the HuProt Human Proteome Microarray. The authors defined training, test and validation sets for patients treated with nivolumab (CM-238: 75% training, 25% test; CM-915: 100% validation) as well as ipilimumab (CM-238) and nivolumab + low-dose ipilimumab (CM-915), both split in 75% training and 25% test sets. Baseline demographics were similar in the test groups with the exception of toxicity for the nivolumab cohorts, with lower rates of severe toxicities in CM-915. Autoantibody profiles at baseline were compared between patients by recurrence as well as high-grade toxicities.

In comparing recurrence during nivolumab, an autoantibody signature recurrence score was defined in the training set which split patients in the validation cohort with a median RFS of 17 months (high score) and a median not reached (low score; AUC 0.82). Similarly, the toxicity score separated patients with 0.75 in the validation set. Using clinical parameters, prediction was improved adding PD-L1 expression. Patients with PD-L1 <5% and a high recurrence score had the worst RFS. For ipilimumab and ipilimumab + nivolumab, similar recurrence and toxicity scores could be defined, though there was no significant overlap between autoantibody signatures. Those autoantibodies associated with disease recurrence were enriched for antigens related to negative regulation of the immune system and all profiles were enriched for autoantibodies against inflammatory response antigens. Analysis of differentially expressed autoantibodies, associated with severe toxicity, demonstrated overlapping enrichment of antigens involved with immune-related pathways. Patients’ median overall autoantibody signal intensity was not associated with disease recurrence or severe toxicity. Moreover, there was no association between disease recurrence and the development of severe toxicity. A missing but clinically relevant analysis might have also been the correlation between lack of recurrence, toxicity and potential overlapping autoantibody profiles. An autoantibody profile that would predict toxicity in a patient with a low risk for recurrence would have a great value in clinical practice, so as to avoid adjuvant therapy, especially in stage IIB/C melanoma.

In our own investigation of stage IV melanoma, more than 300 serum samples of patients with metastatic disease treated with anti-PD1, ipilimumab or combination therapy were analyzed for autoantibodies using a Luminex-based array (4). Here, a set of 47 autoantibodies was found to be predictive for the development of irAE (including both increased and decreased risk of irAE). While differences in autoantibody identification likely exist between studies due to the assay used, we similarly observed associations with better or worse clinical outcomes. Across both studies, the validation of an autoantibody signature that predicts severe toxicity with ipilimumab plus nivolumab would be a high priority given alternative combination immunotherapy approaches with relatlimab and nivolumab. This would be an even greater relevance if such a signature also correlated with clinical outcome. On an organ specific level, an autoantibody profile for irAE with a high mortality rate such as myocarditis, or in the adjuvant setting a high likelihood for hypophysitis or diabetes, would be invaluable in facilitating shared decision making for the choice or avoidance of therapy.

It is not currently clear whether autoantibodies are bystanders or have a functional role for the induction of irAE, either generally or by organ specificity (Fig. 1). Johannet and colleagues noted that the autoantibodies associated with severe toxicity were enriched against immune-related pathway antigens. This is a relatively nonspecific finding however and the exact make-up of the proposed toxicity scores or specific autoantibodies included is unclear. Further work in this field is needed to identify individual or more specific combinations of autoantibodies that underly pathologic mechanisms. We have suggested that there may be autoantibodies with pathologic relevance for the induction of, or even protection from, irAE. This raises an important issue surrounding the heterogeneity between different assays. High priority consideration should be given to potential molecular mechanisms beyond more general biomarker signatures, especially when focusing on irAE.

Figure 1.

Impact of autoantibodies on clinical outcome after immunotherapy. Patients with melanoma receiving immune checkpoint blockade generate a spectrum of autoantibodies that correlate with toxicity, treatment response, and recurrence in the adjuvant setting. Whether these antibodies are antigen specific or a more nonspecific epiphenomenon remains to be determined.

Figure 1.

Impact of autoantibodies on clinical outcome after immunotherapy. Patients with melanoma receiving immune checkpoint blockade generate a spectrum of autoantibodies that correlate with toxicity, treatment response, and recurrence in the adjuvant setting. Whether these antibodies are antigen specific or a more nonspecific epiphenomenon remains to be determined.

Close modal

An autoantibody signature that correlates with prognosis would be of substantial clinical interest in deciding for or against adjuvant therapy. In non–small cell lung cancer, a 13-biomarker autoantibody signature was defined in 157 patients before surgery by an immunome protein array of >1,600 antigens (5). Patients with a high score had decreased survival with a 5-year survival rate of below 8%. For melanoma, it would be interesting to evaluate autoantibody profiles in the placebo arms of adjuvant trials as a prognostic biomarker for recurrence. Such investigations could be used in the design of randomized trials or in clinical decision-making surrounding treatment choice based on the reduction of recurrence versus tolerance of toxicity.

To conclude, Johannet and colleagues contribute to growing evidence that autoantibody profiles have intriguing potential as tools to assess patient prognosis, toxicity, and treatment decision. Further efforts are needed to mechanistically understand these phenomena and implement meaningful measurement of them in clinical practice.

J.C. Hassel reports personal fees from BMS, GSK, MSD, Novartis, Pierre Fabre, Roche, Sanofi, and Sunpharma and grants from BMS and Sunpharma outside the submitted work. J.J. Luke reports service on data safety monitoring boards of AbbVie, Immutep, and Evaxion; scientific advisory board membership with (no stock) 7 Hills, Bright Peak, Exo, Fstar, Inzen, RefleXion, and Xilio, and (stock) Actym, Alphamab Oncology, Arch Oncology, Kanaph, Mavu, NeoTx, Onc.AI, OncoNano, Pyxis, STipe, and Tempest; consultancy with compensation from AbbVie, Bayer, Bristol-Myers Squibb, Castle, Checkmate, Codiak, Crown, Day One, Duke St, EMD Serono, Endeavor, Flame, Genentech, Gilead, Glenmark, HotSpot, Kadmon, Janssen, Ikena, Immunocore, Incyte, IO Biotech, Macrogenics, Merck, Nektar, Novartis, Partner, Pfizer, Regeneron, Roivant, Servier, STINGthera, Synlogic, and Synthekine; and research support from (all to institution for clinical trials unless noted) AbbVie, Astellas, AstraZeneca, Bristol-Myers Squibb, Corvus, Day One, EMD Serono, Fstar, Genmab, Ikena, Immatics, Incyte, Kadmon, KAHR, Macrogenics, Merck, Moderna, Nektar, Next Cure, Numab, Palleon, Pfizer, Replimmune, Rubius, Servier, Scholar Rock, Synlogic, Takeda, Trishula, Tizona, and Xencor. No other disclosures were reported.

J.J. Luke is supported by the NIH/NCI (UM1CA186690-06, P50CA254865-01, and P30CA047904-32).

1.
Johannet
P
,
Liu
W
,
Fenyo
D
,
Wind-Rotolo
M
,
Krogsgaard
M
,
Mehnert
JM
, et al
.
Baseline serum autoantibody signatures predict recurrence and toxicity in melanoma patients receiving adjuvant immune checkpoint blockade
.
Clin Can Res
2022
;
28
:
4121
30
.
2.
Eggermont
AMM
,
Hamid
O
,
Long
GV
,
Luke
JJ
.
Optimal systemic therapy for high-risk resectable melanoma
.
Nat Rev Clin Oncol
2022
;
19
:
431
9
.
3.
Postow
MA
,
Sidlow
R
,
Hellmann
MD
.
Immune-related adverse events associated with immune checkpoint blockade
.
N Engl J Med
2018
;
378
:
158
68
.
4.
Hassel
JC
,
Zucht
H-D
,
Mangana
J
,
Dummer
R
,
Pföhler
C
,
Wistuba-Hamprecht
K
, et al
.
Autoantibodies as predictors for survival and immune-related adverse events in checkpoint inhibition therapy of metastasized melanoma
.
J Clin Oncol
38
:
15s
,
2020
(
suppl; abstr 10011
).
5.
Patel
AJ
,
Tan
T-M
,
Richter
AG
,
Naidu
B
,
Blackburn
JM
,
Middleton
GW
.
A highly predictive autoantibody-based biomarker panel for prognosis in early-stage NSCLC with potential therapeutic implications
.
Br J Cancer
2022
;
126
:
238
46
.