Tyrosine-protein kinase SYK-related gene signature in baseline immune cells associated with immune- related adverse events in melanoma

Abstract

Provided are methods for determining if an individual is likely to experience immune-related adverse events (irAEs) if treated for melanoma with a combination of immune checkpoint inhibitors that are an anti-Cytotoxic T-Lymphocyte-Associated Protein 4 (CTLA-4) antibody and an anti-programmed death-ligand 1 (PD-1) antibody. The methods include determining differential gene signatures as between individuals who had melanoma and were treated with a combination of a CTLA-4 antibody and an PD-1 antibody but did not experience severe irAEs after the treatment, with the gene signature for an individual who has melanoma but was not treated with an immune checkpoint inhibitor.

Description

[0001] Attorney Docket No.: 058636.00814 TYROSINE-PROTEIN KINASE SYK-RELATED GENE SIGNATURE IN BASELINE IMMUNE CELLS ASSOCIATED WITH ADJUVANT IMMUNOTHERAPY- INDUCED IMMUNE-RELATED ADVERSE EVENTS IN MELANOMA CROSS REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. provisional patent application no.63 / 679,452, filed August 5, 2024, the entire disclosure of which is incorporated herein by reference. STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH This invention was made with government support under P50 CA225450 and F99 CA274650 awarded by the National Institutes of Health. The government has certain rights in the invention. RELATED INFORMATION Predicting patients at risk of immune-related adverse events (irAEs) from melanoma ICI, particularly in adjuvant (AT) / neoadjuvant (NAT) treatments, is a previously unmet clinical need to guide less toxic treatment selection. Immune-checkpoint inhibition (ICI) therapies have recently been explored in the adjuvant (post-surgical tumor resection)1-4and neoadjuvant (pre-surgical resection)5settings where they demonstrated improved outcomes when compared to surgical intervention alone. Nevertheless, many patients treated with ICI develop severe immune-related adverse events (irAEs), resulting in significant complications that often require immunosuppressive therapy interventions and / or potential treatment discontinuation, thus negatively impacting patient quality of life, and reducing overall clinical benefits of ICI. In advanced, unresectable metastatic cancers, the risks of irAEs are a necessary trade-off compared to a poor prognosis in the absence of ICI intervention. In contrast, greater consideration must be given to the morbidity of severe irAEs in adjuvant (AT) / neoadjuvant (NAT) ICI treatment of patients, some of whom would be cured with surgery alone. In such patients, the unnecessary complications related to severe grade 3-5 irAEs are a critical concern that must be weighed when assessing the overall benefit of ICI. Furthermore, identifying biomarkers that can point to the biological mechanisms of irAE etiology can lead to the development of toxicity-reducing therapies that enable ICI administration while preventing severe irAEs. Developing baseline biomarkers that are easy to assess and can reliably, robustly, and reproducibly predict irAE risk is key to determining the appropriate ICI treatment regimen to maximize outcomes while minimizing debilitating irAEs. Pre-treatment biomarkers of ICI toxicity, primarily studied in the metastatic disease setting,6have been limited. Collectively, while some baseline predictive biomarkers of irAEs from peripheral blood have been proposed, these are not yet clinically validated, and also vastly unexplored in AT or NAT, which has been an urgent clinical task in patients treated at earlier, less advanced stages of melanoma. Thus, there is an ongoing an unmet need for alternative biomarkers of irAEs, and methods of use of such biomarkers in recommending treatment, and treating patients with alternative approaches. The present disclosure is related to this need. BRIEF SUMMARY This disclosure provides gene-expression profiles of circulating CD4+ and CD8+ T cells of 212 patients from the CheckMate-915 clinical trial to identify irAE-predictive biomarkers in AT. The disclosure reveals a gene expression signature in the SYK pathway correctly predicting ~60% of ipilimumab-nivolumab (COMBO)-treated patients developing severe grade 3-5 irAEs. The signature also predicts mild to moderate irAEs linked to SYK- related autoimmune traits, with 96% of patients experiencing ≥1 and 79% experiencing ≥2 of these SYK-related toxicities. The regression model incorporating the SYK signature improved the irAE prediction accuracy from 73% to 82%. The signature is independent of disease recurrence, suggesting its biomarker potential to select patients for less toxic treatment choices, without hampering therapeutic efficacy. The analysis of predicted protein-protein interactions among differentially expressed genes (DEGs) in peripheral CD4+cells revealed significant enrichment of the spleen tyrosine kinase (SYK) pathway, associated with severe irAEs in COMBO-treated patients. This gene- expression signature predicted severe-irAE COMBO patients (chi-square p-value=0.001) with 73% accuracy and was independent of disease recurrence (p=0.79). The irAE-predictive model incorporating this gene-expression signature demonstrated 82% accuracy (chi-square p-value=8.91E-06). The results were obtained by assessing post-surgical, pre-ICI treatment peripheral CD4+and CD8+T cells from clinical trial patients (CheckMate-915) treated with AT nivolumab (NIVO, n=130) or ipilimumab / nivolumab (COMBO, n=82). RNA-seq differential gene expression analysis was used to detect baseline differences associated with severe (grade 3-5) irAEs and to construct an irAE-predictive model using LASSO-regularized logistic regression. The disclosure accordingly reveals baseline gene-expression differences in key immune pathways of peripheral blood T cells from COMBO-treated patients with grade 3-5 irAEs, and defines a SYK-related gene signature correctly identifying ~60% of COMBO- treated patients with grade 3-5 irAE. This gene signature can be used a baseline biomarker of severe grade 3-5 irAE risk, which is particularly pertinent in AT. The described signatures allow for tailoring individual melanoma therapies using selected ICIs or other melanoma treatments. BRIEF DESCRIPTION OF THE FIGURES FIG.1. Heatmaps showing the significantly differentially expressed genes (DEG), defined as P < 0.01, |log2FC| >1, from RNA-seq DEG analysis of peripheral CD4+ and CD8+ T cells, comparing high-grade (3–5) vs. low-grade (0–2) irAEs. a., DEGs from COMBO CD4+ analysis (n = 78 samples); 87 genes (n = 63 upregulated in irAE; n = 24 downregulated in irAE). b., DEGs from NIVO CD4+ analysis (n = 124 samples); 43 genes (n = 7 upregulated in irAE; n = 36 downregulated in irAE). c., DEGs from COMBO CD8+ analysis (n = 78 samples); 92 genes (n = 34 upregulated in irAE; n = 58 downregulated in irAE). d., DEGs from NIVO CD8+ analysis (n = 121 samples); 75 genes (n = 29 upregulated in irAE; n = 46 downregulated in irAE). FIG.2. Pathway enrichment analysis of DEGs in peripheral CD4+ and CD8+ from patients in CheckMate 915. a., Overlap in DEGs between the treatments (NIVO / COMBO) and cell types (CD4+ / CD8+) comparing high-grade (3–5) vs. low-grade (0–2) irAEs. The intersection size (y-axis on the vertical bar chart) indicates the number of unique DEGs in each comparison. The set size (x-axis on the horizontal bar chart) indicates the total number of DEGs in each category. b., Top enriched pathways (adjusted P < 0.01) from Gene Ontology analysis using the overlapping DEGs between cell types and treatments. Each bar is labeled with the number of DEGs from the pathway that are present (e.g., COMBO CD4+ and CD8+ cells shared 9 DEGs of a total of 847 genes in the IL2 signaling pathway; CD4+ cells in COMBO and NIVO shared 5 DEGs in the same pathway). FIG.3. PPI volcano plot illustrating proteins encoded by significantly enriched gene sets from COMBO CD4+ DEGs between patients with high-grade (3–5) vs. low-grade (0–2) irAEs. The OR for the odds of observing the given DEGs at random is plotted on the x-axis. The y-axis plots the –log(P value) for the OR [a larger –log(P value) means the genes are less likely to be randomly observed]. The blue dots represent proteins associated with significant gene sets (OR >1.0, P value <0.05; darker blue signifies greater significance), indicating the DEGs associated with these proteins are unlikely to be randomly observed. FIGS.4A-4B. Differential expression of genes expressed in CD4+ T cells that encode proteins associated with SYK identifies the majority of COMBO-treated patients with high- grade irAEs.4A, Semisupervised hierarchical clustering using SYK-associated genes in PPI analysis for COMBO CD4+ DEGs between patients with high-grade (3–5) vs. low-grade (0– 2) irAEs (24 of 136 SYK-interacting genes; P value 9.96E.04, adj. P value 0.04). The cluster to the far right comprises the 24 patients with the most similar mean gene expression of these 24 genes and contains the majority of patients with severe irAEs.4B, Network diagram showing the immune response Fc-epsilon receptor I pathway identified from Gene Ontology on these 24 genes. Post-transl. m, post-translational modification. Network diagram from the PathCards Pathway Unification Database. FIG.5. Bubble plot showing irAEs shared among COMBO-treated patients in the SYK-pathway cluster (SYK-irAEs). To be included, the irAE was present in at least one TOXhigh(above dotted line; n = 13) patient and one TOXlow(below dotted line; n = 11) patient. Each row represents a patient with the SYK-GEP, and each column represents an SYK-irAE. The size of the bubble corresponds to the number of events of each toxicity that the patient experienced while on treatment, with larger circles indicating more frequent events. The color scale shows the CTCAE grade for the irAEs: orange and red events, more severe (grades 3–4); green and blue events, less severe (grades 1–2). ALT, alanine aminotransferase; AST, aspartate aminotransferase; TSH, thyroid-stimulating hormone. FIG.6. Model performance of the logistic regression model predicting severe irAEs (grades 3–5) using age, sex, stage, and 5-gene SYK-GEP expression (CD22, PAG1, CD33, HNRNPU, and FCGR2C) in COMBO-treated patients. a., ROC curve and the corresponding AUC plotting sensitivity and specificity of the prediction model. b., Waterfall plot for severe irAE prediction using the regression model parameters scaled so that prediction scores <0 are predicted to have no severe irAEs and scores >0 are predicted to have severe irAEs. The actual toxicity status of the patients is indicated by color (blue, no severe irAEs; salmon, severe irAEs). FIG.7. Sample size and power calculation for different effect size values. For a two- tailed alpha of 0.05, this plots the power at a given sample size (y-axis) to detect a given effect size (w) (x-axis). The dashed red and blue lines correspond to the NIVO and COMBO analytic cohort sample sizes, respectively. The dotted vertical lines illustrate the effect size for the gene expression profile (w=0.43) and regression model (w=0.66) identified in the COMBO CD4+cohort. As they both converge with the line corresponding to the COMBO sample size above the pink 95% power line, this indicates that there is >95% power to detect an effect of this magnitude. FIG.8. Top 10 Gene Ontology (GO) results for top DEGs (p<0.05) comparing low / no irAEs to high irAEs. a. GO for COMBO CD4+b. GO for Nivo CD4+c. GO for COMBO CD8+d. GO for Nivo CD8+. FIG.9. Semi-supervised hierarchical clustering of the DEGs comparing TOXlowvs TOXhighin CD4+cells from COMBO-treated patients (n=78) identified from PPI analysis. a. DEGs in the MAPK8 and MAPK14 pathways b. DEGs in the following pathways: (1) SMAD3 and SMAD4, (2) PRKCZ, (3) SRC, (4) CSNK2A1. FIG.10. Semi-supervised hierarchical clustering of the DEGs comparing TOXlowvs TOXhighin CD4+cells from NIVO-treated patients (n=124) identified from PPI analysis. The hierarchical clustering by SYK-GEP used genes identified as differentially expressed in COMBO CD4+analysis comparing TOXlowvs TOXhigh. FIG.11. Detailed annotation for semi-supervised hierarchical clustering of the genes associated with SYK in protein-protein interaction (PPI) analysis comparing TOXlowvs TOXhighin CD4+cells from COMBO-treated patients (n=78). Annotations include follow-up status (recurrence still in follow-up (RE / FU)), high / low toxicity, patients with SYK-irAEs without the SYK-GEP, patient without SYK-GEP with vitiligo, and patients who experienced colitis. FIG.12. Correlation plots for predictor variables in cross-validation logistic regression models. a: All clinical variables and genes in the SYK-GEP with clusters of positively and negatively correlated variables indicating multicollinearity. b: Final variables included in the regression model after excluding those with a Variance Inflation Factor (VIF) >1.5 and / or tolerance of <0.25. FIG.13. Waterfall plot for logistic regression model predicting severe (grade 3-5) irAEs using 24-gene SYK GEP. The model incorporates SYK-associated differentially expressed genes (DEGs) and clinical variables. In the waterfall plot the highlights indicate COMBO-treated patients with and without the 24-gene SYK-GEP by toxicity. FIG.14. Table showing differential gene expression of genes for the described 5 and 24 gene signatures. Positive log2FC = gene is upregulated in those with severe irAEs compared to control (those without severe irAEs). Negative log2FC = gene is downregulated in those with severe irAEs compared to control. DETAILED DESCRIPTION Unless defined otherwise herein, all technical and scientific terms used in this disclosure have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. Every numerical range given throughout this specification includes its upper and lower values, as well as every narrower numerical range that falls within it, as if such narrower numerical ranges were all expressly written herein. Although relevant subject matter will be described in terms of certain examples, other examples, including examples that do not provide all of the benefits and features set forth herein, are also within the scope of this disclosure. The steps of the methods described in the various examples disclosed herein are sufficient to carry out the methods of the present invention. Thus, in an example, the method consists essentially of a combination of the steps of the methods disclosed herein. In another example, the method consists of such steps. As used in the specification and the appended claims, the singular forms “a” "and” and “the" include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from “about” one particular value, and / or to “about” another particular value. When such a range is expressed, another example includes from the one particular value and / or to the other particular value. Similarly, when values are expressed as approximations, by the use of the antecedent “about” or “approximately” it will be understood that the particular value forms another example. The term “about” and “approximately” in relation to a numerical value encompass variations of + / -10%, + / - 5%, or + / - 1%. In examples, the disclosure provides for determining differential expression of genes as described herein, wherein the differential expression is relative to a control value. In examples, differential gene expression refers to expression levels which are at least 10%, 20%, 30%, 40%, or 50%, 60%, 70%, 80%, 90% higher or more, and / or 1.1 fold, 1.2 fold, 1.4 fold, 1.6 fold, 1.8 fold, 2.0 fold higher or more, and any and all whole or partial increments there between relative to a comparator. In examples, differential gene expression refers to expression levels which are at least 10%, 20%, 30%, 40%, or 50%, 60%, 70%, 80%, 90% higher or more, and / or 1.1 fold, 1.2 fold, 1.4 fold, 1.6 fold, 1.8 fold, 2.0 fold lower or more, and any and all whole or partial increments there between relative to a comparator. This disclosure includes all nucleotide sequences that are described by reference to a database entry. The disclosure includes such nucleotide sequences as they exist in the database as of the filing date of this disclosure. The disclosure includes all RNA equivalents of DNA sequences that may be in any database entry, as well as complementary sequences, and reverse complementary sequences. All splice variants of any mRNA are included within the scope of this disclosure. In examples, representative comparators are referred to herein as control expression values. In examples, a control expression value comprises a determination of mRNA levels transcribed from described genes in biological samples from individuals who were treated with an ICI but did not experience irAEs, or only experienced mild irAEs. In this regard, irAEs are known to be side effects that arise from treatment with immune checkpoint inhibitors, such as anti-PD-1 and anti-CTLA-4 agents. These therapies function by activating the immune system to recognize and attack cancer cells; however, this enhanced immune activity can also result in inflammation and damage to healthy tissues. irAEs can affect nearly any organ system, with commonly involved sites including the skin, gastrointestinal tract, liver, endocrine glands, lungs and other sites. Symptoms may range from mild, such as rash or fatigue, to severe or life-threatening conditions like colitis, pneumonitis, or autoimmune hepatitis or other autoimmune like conditions including psoriasis, vitiligo and erythema. These events may occur during therapy or even months after its discontinuation. Management typically involves immunosuppressive treatment, such as corticosteroids, with treatment decisions guided by the severity of the irAE. As described herein, mild or no irAEs is indicated by a severity grade of 0-2 (TOXlow). In contrast, severe irAEs are indicated by a severity score grades of 3-5 (TOXhigh). Those skilled in the art will understand how to determine irAE severity scores. In examples, severity scores are calculated according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE) version 4.0, from which the criteria are incorporated herein by reference. In examples, control expression values are obtained from individuals who were diagnosed with melanoma and were treated with an anti-CTLA-4 antibody, an anti-PD-1 antibody, or a combination thereof. In examples, the control expression value to which the patient sample is compared comprises a sample from one or more individuals who were diagnosed with melanoma and did not experience severe irAEs after treatment with one or more described checkpoint inhibitors as a treatment for the melanoma. In examples, the control threshold value may be represented by standardized curve(s), a cutoff or threshold value, and the like. In examples, the control threshold value is a severity grade of below 3. In examples, any described individual in this disclosure may be treated, or have been treated, with ICI as adjuvant therapy or neoadjuvant therapy. In examples, for an individual treated with a described combination therapy after performing a described differential gene expression test, the disclosure comprises testing a second, third, fourth, sample, etc. from the individual who is undergoing treatment and tested to monitor the appearance and / or severity of the irAEs, and thereby keep steady, change, adjust, discontinue, or not reinitiate treatment with the combination ICI therapy. In examples, gene expression via mRNA analysis may be determined from any suitable source, including but not necessarily limited to peripheral blood mononuclear cells (PBMCs). In examples, gene expression is determined from CD4+ cells, CD8+ cells, or a combination thereof. mRNA may be analyzed using a variety of known techniques that allow qualitative or quantitative expression analysis. In examples, the disclosure identifies genes that are differentially expressed as between individuals who did not experience irAEs or that indicated that ICI treatment should be discontinued, and an individual who has been diagnosed with melanoma, but has not yet been treated with ICI. The disclosure thus provides for making ICI-related treatment decisions for melanoma patients, and in examples includes treating the melanoma patient with a therapy that is influenced by the gene expression pattern determined from a biological sample from the individual. The genes that are differentially expressed are described in a tables of this disclosure which also provide sequence access numbers and indication of over and underexpression relative to control values. In an example, the disclosure provides a method for determining if an individual is likely to experience irAEs if treated for melanoma with a combination of ICIs that are an anti-cytotoxic T-Lymphocyte-Associated Protein 4 (CTLA-4) antibody and an anti- programmed death-ligand 1 (PD-1) antibody, and optionally treating the individual for the melanoma. The treatment may be determined based at least in part by taking into account the described differential expression of the genes. In an example, a method comprises testing a biological sample obtained from the individual before treating the individual with the anti-CTLA-4 antibody or the anti-PD-1 antibody or a combination thereof, and from the biological sample determining expression of a plurality of genes. In one approach, the plurality of genes comprises at least CD22, PAG1, CD33, HNRNPU, and FCGR2C. Expression of the CD22, PAG1, CD33, HNRNPU, and FCGR2C genes is compared to a control expression value for the CD22, PAG1, CD33, HNRNPU and FCGR2C genes obtained from individuals who had melanoma and did not experience irAEs after treatment with the anti-CTLA-4 antibody or the anti-PD-1 antibody or a combination thereof. A difference in expression of the CD22, PAG1, CD33, HNRNPU, and FCGR2C genes relative to the control expression value indicates the individual is likely to experience the irAEs, such as moderate or severe irAEs. In such a case, the individual may be treated with immune checkpoint inhibition that does not comprise a combination of the anti-CTLA-4 antibody and the anti-PD-1 antibody, or with an ICI an ICI that for example an anti-PD-1 antibody as a monotherapy in terms if ICI. Other ICIs may be used and include but are not necessarily limited to anti-PD-L1 antibodies, anti-LAG-3 antibodies, anti-TIM-3 antibodies, - anti-TIGIT antibodies, or anti-VISTA antibodies. Anti-PD-1 antibodies include but are not necessarily limited to Nivolumab, Pembrolizumab, Cemiplimab, Dostarlimab, Retifanlimab, Toripalimab, and Tislelizumab. Anti-PD-L1 Inhibitors include but are not necessarily limited to Atezolizumab, Durvalumab, and Avelumab. Anti-CTLA-4 include but are not necessarily limited to Ipilimumab and Tremelimumab. A non-limiting example of an anti-LAG-3 inhibitor is relatlimab. A non-limiting example of an anti-TIM-3 antibody is galunimab. A non-limiting example of an anti-VISTA antibody is pidilizumab. In examples, an individual who is identified using a described process may be treated with ICI using an immune checkpoint inhibitor as a monotherapy, or a combination of immune checkpoint inhibitors as described herein. In the case where no difference in expression of the CD22, PAG1, CD33, HNRNPU, and FCGR2C genes relative to the control expression value is determined, or a difference that is not statistically significant is determined, the method may include treating the individual with a combination of the anti-CTLA-4 antibody and the anti-PD-1 antibody, or other ICI agents as described herein or would be otherwise known in the art. Thus, in an example, “no difference” means no statistically significant difference. Accordingly, in an example, a difference means a statistically significant difference. While expression of the 5 gene signature discussed above is believed to be sufficient to perform the described characterization of the melanoma patient who did not receive an ICI, the disclosure further includes determining a difference in expression of genes in the biological sample obtained from the individual to determine expression of genes selected from the group consisting of RPS6KB2, EZR, PTK2, MAPK3, LYN, ITGB3, CTTN, CSF3R, UBC, GAB2, PIK3CA, GRB2, SWAP70, MSN, PIK3CG, PIK3AP1, PIK3CD, PRKCA, CLEC7A, or a combination thereof, and comparing expression of one or more of said genes to the control expression value for the same one or more genes obtained from individuals who did not experience severe irAEs after treatment with the anti-CTLA-4 antibody or the anti-PD-1 antibody or a combination thereof. In examples, treating the individual with a melanoma therapy that does not comprise administering to the individual a combination of an anti-CTLA-4 antibody and an anti-PD-1 antibody is indicated, or more of the following statements, in any combination, are true: expression of the CD22 mRNA for the individual is higher than its control expression value; expression of the PAG1 mRNA for the individual is lower than its control expression value; expression of the CD33 mRNA is higher than its control expression value; expression of the HNRNPU mRNA is higher than its control expression value; expression of the FCGR2C mRNA is higher than its control expression value; expression of the RPS6KB2 mRNA is higher than its control expression value; expression of the EZR mRNA is higher than its control expression value; expression of the PTK2 mRNA is higher than its control expression value; expression of the MAPK3 mRNA is lower than its control expression value; expression of the LYN mRNA is higher than its control expression value; expression of the ITGB3 mRNA is lower than its control expression value; expression of the CTTN mRNA is lower than its control expression value; expression of the CSF3R mRNA is higher than its control expression value; expression of the UBC mRNA is higher than its control expression value; expression of the GAB2 mRNA is higher than its control expression value; expression of the PIK3CA mRNA is higher than its control expression value; expression of the GRB2 mRNA is higher than its control expression value; expression of the SWAP70 mRNA is higher than its control expression value; expression of the MSN mRNA is higher than its control expression value; expression of the PIK3CG mRNA is higher than its control expression value; expression of the PIK3AP1 mRNA is higher than its control expression value; expression of the PIK3CD mRNA is lower than its control expression value; expression of the PRKCA is lower than its control expression value; and expression of the CLEC7A is higher than its control expression value. In contrast, when treating the individual with a melanoma therapy that does comprise administering to the individual a combination of an anti-CTLA-4 antibody and an anti-PD-1 antibody is indicated, any of the statements above may be not true. The disclosure includes the proviso that any described method can exclude testing for the rs7036417 genotype. The following examples are intended to illustrate aspects of the disclosure but are not intended to be limiting. Examples Materials and Methods Study Population: Checkmate 915 In this disclosure, we used peripheral blood mononuclear cells (PBMCs) from the CheckMate-915 clinical trial (ClinicalTrials.gov identifier NCT03068455), a 2-arm Phase III randomized study enrolling 1844 patients within 12 weeks after complete surgical resection of stage IIIB / C / D or stage IV melanoma. Subjects were treated with either single-agent Nivolumab (NIVO) or the combination of Ipilimumab and Nivolumab (COMBO). The CheckMate-915 protocol was approved by the institutional review boards (IRBs) of all study sites participating in the clinical trial, and written informed consent was provided by all patients before enrollment in accordance with the Declaration of Helsinki.22This disclosure provides results from RNA-sequencing (RNA-seq) of CD4+and CD8+cells isolated from PBMCs collected before ICI treatment initiation (n=212 samples; NIVO n=130; COMBO n=82). The samples (n=212) were randomly selected from the CheckMate-915 cohort and subsequently evaluated for the distribution of treatment response and demographic variables. We confirmed the clinical and demographic characteristics of the selected analytical cohort (Table 1) reflected those of the overall CheckMate-915 study population.22Treatment efficacy was determined by recurrence-free survival (RFS) as evaluated by the CheckMate-915 investigators,1with RFS outcomes dichotomized by those whose disease had recurred (RE) and those still in follow-up at the time of censoring (FU). Adverse events (AEs) while on study were assessed and graded by the CheckMate-915 investigators according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE) version 4.0. For this analysis, we assessed only immune-related AEs (irAEs), defined as AEs determined to be related to the ICI treatment received. These analyses included AE and efficacy outcomes from a minimum follow-up of 23.7 months with 24- month RFS rates.1Sample Isolation, Sequencing, and Genotyping. Pre-treatment PBMCs were used as a source of CD4+and CD8+T cell populations from 212 patients (NIVO n=130; COMBO n=82). CD4+and CD8+cells were negatively selected from the overall PBMC population using an EasySep Human CD4+or CD8+Enrichment Kit (STEMCELL Technologies, Vancouver, BC, Canada) per the manufacturer’s instructions and RNA and DNA were extracted using the AllPrep DNA / RNA kit (Qiagen, Germantown, MD). Ribosomal RNA-depleted RNA-seq libraries were generated from purified RNA specimens by the SMARTer Stranded Total RNA-Seq Kit v2 Pico Input Mammalian (TaKaRa Bio USA, Mountain View, CA) and sequenced on the NovaSeq 6000 (Illumina, San Diego, CA) using 150-base paired-end sequencing to a depth of ~30 million reads per sample. STAR aligner23was used to map and align the reads to the human reference genome version hg38 (median paired reads per sample 64,251,603 across all 411 sequenced samples including both CD4+and CD8+cells). Gene expression and transcript quantification were obtained with RSEM v1.3.1 using default parameters.24Genotypes for the SYK-associated variant, rs7036417, were assessed as previously described.9Differential Gene Expression Analysis The R package DESeq225was used to identify differentially expressed genes (DEGs) between those with mild or no irAEs (grades 0-2; TOXlow) compared to those with severe irAEs (grades 3-5; TOXhigh). Raw gene expression levels were normalized in DESeq2,25and sequencing outliers (>5 standard deviations from PCA mean) were excluded. We filtered artifactual / lowly-expressed genes, retaining only those with a read count of ≥20% of the sample size across all samples and those with ≥2 reads in > 10% of samples. The threshold for defining significant DEGs was a p-value <0.01 and an absolute log2fold change value >1 between the comparison groups. To address the potential for type I error when conducting multiple comparisons, the Benjamini-Hochberg Procedure for false discovery rate adjustment (FDR) was applied; an FDR-adjusted p-value of <0.05 was considered statistically significant. Predicted protein-protein interaction (PPI) analysis was performed using eXpression2Kinases (X2K), a computational tool for predicting cell signaling pathways. Given a specified DEG set (DEGs defined here by nominal p-value <0.05), X2K predicts transcription factors (TFs) likely to regulate expression of the provided genes, integrates these TFs with known protein-protein interactions, and performs kinase enrichment on the resulting interaction network.26Visualization of overlapping DEGs was generated using UpSet.27Biological relevance of the identified transcriptional signatures was tested by pathway modeling and interactome analyses.28-30Volcano plot for the significantly enriched PPI gene set was generated using Apptyers.31Statistical significance of the identified PPIs was determined using a chi-square test, calculating the p-value for the probability of seeing at least x genes out of n total proteins in the pathway. PPI networks for the DEGs were assessed using the PathCards Pathway Unification Database.32Semi-supervised hierarchical clustering was performed using the R package pheatmap,33which employs k-means clustering to group samples by the similarity of their mean gene expression. All RNA-seq analyses were conducted in R version 4.3.0. Statistical Analyses Sensitivity, specificity, and accuracy were calculated to evaluate the predictive potential of the gene expression profiles on treatment outcomes (toxicity and response). Briefly, sensitivity was defined as the proportion of patients correctly identified by the gene expression profile as having severe toxicity or disease recurrence (true positives; TP) divided by all patients (TP + false negatives; FN). Specificity was defined as the proportion of patients correctly identified by the gene expression profile who did not develop severe toxicity or disease recurrence (true negatives; TN) divided by all the patients (TN + false positives; FP). Accuracy was defined as the ratio of the correct results (TP + TN) to the entire sample (TP + TN + FP + FN). The chi-square test was performed to evaluate the associations between gene expression profiles and predicted versus observed treatment outcomes. Logistic regression models incorporating gene expression and clinical covariates (age, sex, disease stage, and rs7036417 genotype) were used to identify the variables most predictive of severe (grade 3-5) toxicity. To improve model stability and reproducibility, we performed least absolute shrinkage and selection operator (LASSO) regularization and 10- fold cross-validation using the R packages glmnet (version 4.1-7)34,35and caret (version 6.0- 94).36We defined the LASSO regularization parameter λ for the regression model such that it minimized the cross-validation error. To further improve model stability, we diagnosed potential multicollinearity among the predictor variables using the R package olsrr (version 0.5.3)37to define the Variance Inflation Factor (VIF), ^ ^ିோమ, and tolerance, ^1 െ ^^ଶ^, for eachvariable, where ^^ଶ ൌ 1 represents perfect multicollinearity.38 Variables with a VIF >1.5and / or tolerance <0.25 were excluded before fitting the final regression models. Correlation plots were generated using the corrplot R package (version 0.92).39To reduce overfitting bias, we performed 300 iterations of LASSO regularization and 10-fold cross-validation, each with a unique random seed. Receiver Operating Characteristic (ROC) curves and the corresponding Areas Under the Curve (AUCs) for the predictions were generated using the Rpackage pROC (version 1.18.2).36,40 ROC curves plot ^^^^^^^^^^^^^^^^^^^^^^ ^^^^. ^1 െ ^^^^^^^^^^^^^^^^^^^^^^^ forall possible cut-points between cases and controls (here, high or low toxicity), and pROC calculates the AUC using the trapezoidal rule.40Chi-square tests were performed comparing the actual toxicities to those predicted by the regression model. We performed a power analysis to assess the magnitude of the effect size detectable in the final analytic cohorts for NIVO and COMBO (Figure 7) using the R package pwr (version 1.3-0).41The effect size index (w) represents the magnitude of an effect independent of sample size; for chi-square tests, w=0.1 is considered small, w=0.3 medium, and w=0.5 large.41As can be seen from the curves in Figure 7, we have >95% power to detect an effect w≥0.40 at a two-sided alpha=0.05. Using the pwr.chisq.test command, we estimated the effect size index for the gene expression profile identified in the COMBO CD4+arm to be w=0.43, indicating we have >95% power to detect an effect of this magnitude. The effect size for the COMBO CD4+signature defined by the logistic regression model was w=0.66, corresponding to >99% power. Results This disclosure provides analysis of specimens from 212 randomly selected patients enrolled in the CheckMate-915 clinical trial who received anti-PD-1 (NIVO) or a combination of anti-PD-1 and anti-CTLA-4 (IPI / NIVO, COMBO)(Table 1).16% of the 130 NIVO-treated patients and 30% of the 82 COMBO-treated patients experienced a severe (CTCAE grade ≥3) immune-related adverse event (irAE), consistent with the established increase in irAE rates observed with COMBO treatment.42Detailed patient-level toxicity information, including the CTCAE grade and frequency of each irAE experienced on study, are provided for both NIVO and COMBO patients. We performed bulk RNA-seq on CD4+and CD8+cells isolated from PBMCs collected post-surgical resection before adjuvant ICI treatment (AT). Given the different toxicity profiles of the two treatments, we performed the differential expression (DE) analyses separately for each ICI arm (Figure 1a-d). In NIVO-treated patients we identified 43 significantly differentially expressed genes (DEGs) (defined as p<0.01, abs(log2FC)>1) in CD4+cells, and 75 DEGs in CD8+cells, with 4 DEGs passing FDR adjustment. In COMBO- treated patients, we identified 87 DEGs in CD4+cells and 92 DEGs in CD8+cells, with 2 genes passing FDR adjustment. Gene Ontology (GO) on DEGs with p<0.05 identified significant enrichment of key immune pathways in both CD4+and CD8+cells across both treatment arms (Figure 8). While the majority of the individual DEGs in these immune pathways were unique to each cell type and treatment, there was an overlap: for example, T- cell receptor-related genes were shared between CD4+and CD8+cells in NIVO patients. Genes involved in IL-2 signaling were shared among CD4+and CD8+cells in COMBO patients, as well as CD4+cells in both COMBO and NIVO patients. TNF-alpha signaling was also common between cell types in COMBO-treated patients, as well as in CD8+cells across treatments. (Figure 2a-b). In cell-specific analyses, we predicted protein-protein interactions (PPIs) for the DEGs (p <0.05) in each treatment arm, identifying hub proteins with the largest number of predicted interactions per each cell-specific DEG set (Table S1a-b). In COMBO CD4+cells, a significant proportion of the enriched encoded proteins were involved in the spleen tyrosine kinase (SYK) pathway (including MAPK8 / 14 and SMAD3 / 4) (Figure 3). In total, 24 of 136 SYK-interacting genes were present in the COMBO CD4+DEGs (p-value for PPI enrichment=9.96E-04, adjusted p-value 0.04). To test if we could develop expression signatures discriminating patients by irAE severity, we performed semi-supervised hierarchical clustering on the DEGs in the PPI-identified pathways. To limit overfitting, we sought to select the PPI signatures that achieved the best predictive capacity with the fewest genes. For example, using the 105 DEGs in the MAPK8 and MAPK14 pathways, we correctly predicted only 39% of TOXhighpatients (Figure 9a). Other top PPIs with larger gene signatures, such as SMAD3 / SMAD4 and others, achieved similar toxicity prediction (hierarchical clustering for DEGs in the other top significant pathways summarized in Figure 9b). In contrast, for the SYK pathway, with only 24 DEGs, the predictive ability to identify COMBO-treated patients with severe irAEs increased to 57% with an accuracy of 73% (chi- square p-value=0.001) (Figure 4a-b, Table 2a, Table S2). Of note, only one of these 24 genes, SWAP70, was significantly DE in NIVO CD4+cells (p-value=0.004). When assessing the predictive potential of the SYK-GEP in CD4+cells of NIVO-treated patients, it identifies 14% of TOXhighpatients (3 of 21) (Figure 10), suggesting that this SYK-related gene expression profile (SYK-GEP) is specific to CD4+cells of COMBO-treated patients. When assessing the toxicity profiles of the COMBO-treated patients with the SYK- GEP, we noticed a common pattern of enrichment of irAEs in organ systems previously linked with increased SYK activity (referred to here as SYK-irAEs),43-64shared among both TOXhighand TOXlow(Figure 5). In particular, all events of psoriasis and erythema in COMBO-treated patients in our analysis were in patients with the SYK-GEP. Six of seven patients with vitiligo among all analyzed COMBO patients were in this cluster, as were the only two COMBO-treated patients in who developed colitis (summarized in Figure 11). Each of these toxicities has been reported to have a direct link with SYK upregulation.43-46,52 When assessing SYK-irAE events of any grade, 79% of all patients with the SYK- GEP (19 / 24) experienced between two to seven different irAEs, in contrast to only 28% of patients without the GEP (15 / 54) (Figure 11, Table 2b) (chi-square p-value=2.4E-05). Analyzing the per-patient enrichment of the same irAE, we found that 58% of patients from the SYK-GEP cluster (14 / 24) experienced >1 event of the same SYK-irAE, in contrast to 35% (19 / 54) of patients outside of the cluster (chi-square p-value=0.06). Additionally, out of the 43 total observed irAEs in TOXlowpatients with the SYK-GEP, 40 were SYK-irAEs. Furthermore, although by definition none of the subjects in the “low toxicity” group experienced a severe irAE, 50% of TOXlowpatients with the SYK-GEP also experienced a moderate (grade 2) event in these shared SYK-irAEs (Figure 5). We analyzed whether the SYK-GEP was associated with disease recurrence (the treatment efficacy outcome in CheckMate-915) and found no association with recurrence when comparing all TOXhighand TOXlowpatients (chi-square p-value=0.99), or those with and without the SYK-GEP (chi-square p-value=0.79)(Figure 11, Table 2c). These results suggest that the SYK-GEP predicts only SYK-irAE risk and not ICI treatment tumor response. We developed an irAE prediction model for COMBO-treated patients incorporating the SYK DEGs and clinical parameters. First, the logistic regression model using only clinical variables showed no significant association with severe (grade 3-5) toxicity (AUC=0.50). To define the most parsimonious regression model, with maximal prediction precision and generalizability, we performed least absolute shrinkage and selection operator (LASSO) regularization, using k-fold cross-validation on the LASSO-regularized models to reduce potential model over-fitting. To further increase model stability, we diagnosed highly correlated predictor variables, which can produce unstable regression coefficients. Excluding highly correlated genes (illustrated in Figure 12a.) reduced the 24-gene SYK DEG signature to five genes (CD22, PAG1, CD33, HNRNPU, and FCGR2C), which were included in the final model along with age, sex, disease stage, and rs7036417 genotype (Figure 12b). Finally, we performed 300 iterations of LASSO regularization and 10-fold cross-validation with unique random seeds to further increase analytical reproducibility. Overall, prediction of severe (grade 3-5) irAE with these variables in COMBO-treated patients was highly stable over 300 model iterations (AUC mean, median, and mode = 0.80, standard error (SE) = 4.46E-04). The most reproducible model included age, stage, and CD22, PAG1, CD33, HNRNPU, and FCGR2C gene expression (AUC mean, median, and mode also = 0.80 for this model, SE = 3.22E-04) (Figure 6a), correctly identifying almost all subjects without severe toxicity (specificity = 96.4%) (Table S3a). This parsimonious model, with only 5 out of the 24 SYK-related genes, identifies 48% of severe toxicity patients, with an overall accuracy of 82.1%, (chi-square p-value = 8.91E-06) (Table S8a, Figure 6b). Stratifying to just those patients identified by the larger 24-gene SYK GEP, the 5-gene model shows the highest sensitivity, correctly predicting 77% of TOXhipatients (10 / 13; Figure 13, dark orange bars) and 91% TOXlowpatients (10 / 11; Figure 13, dark blue bars), for an overall accuracy of 83.3% (chi-square p-value=8.90E-04) (Table S3b). Taken together, these results indicate that a parsimonious model incorporating baseline 5-gene SYK-related GEP and clinical variables can effectively identify a significant proportion of COMBO-treated patients at risk of developing severe irAEs. Discussion of Examples Melanoma treatment has been significantly altered with the advent of ICI therapies, first in the metastatic65-67and now in the adjuvant (AT)1,2and neoadjuvant (NAT)5settings. While ICI therapies have increased patient survival, severe ICI-induced irAEs can significantly complicate treatment outcomes and quality of life. As such, there has been a pressing clinical need to identify reliable and minimally invasive baseline (pre-treatment) markers of irAE risk, particularly in AT, where many patients are cured by surgery alone. Several biomarkers of irAE development have been described,6but none have reached clinical utility, and only limited efforts have been expended to define biomarkers in AT to date. In this disclosure, for the first time in AT, we explored whether host baseline transcriptional profiles of circulating T cells associate with toxicity outcomes in both single- agent NIVO and COMBO treatments. We identified significant upregulation of genes in key immune pathways including IL-2, IFN-gamma, TCRs, and Toll-like receptors in patients with severe irAEs from both treatments, suggesting pre-existing transcriptional differences in inflammatory pathways that may predispose these patients to irAEs upon ICI-mediated immune stimulation. Our main findings also suggest that baseline upregulation of genes involved in the SYK pathway is associated with combined anti-PD-1 / anti-CTLA-4 toxicity in AT. This signature is unique to COMBO-treated patients, suggesting an involvement of anti- CTLA4.9Among the COMBO-treated patients in the study, this gene signature is unique to CD4+T cells and is not observed in CD8+T cells, aligning with prior evidence that SYK regulation modulates the activity of regulatory T cells, including CD4+T cells.68 Although this signature involves genes interacting with SYK, we did not observe significant differential expression of SYK itself.69While altered SYK expression has been observed in autoimmune conditions including systemic lupus erythematosus (SLE)70-72and rheumatoid arthritis (RA),73the differences in receptor signaling and signal transduction in these patients may be due to activated pSYK.69Additionally, SYK gene expression among immune cells is more prevalent in B-cells,72,73which may also help explain why we do not observe SYK DE in the transcriptomes of the T cells we analyzed.71,74Herein we have identified a pre-treatment gene expression profile (GEP) that correctly identifies almost 60% of patients experiencing severe irAEs upon COMBO ICI treatment in melanoma AT (SYK-GEP). In patients with the SYK-GEP, we observed an enrichment of irAEs in the skin, GI, liver, and endocrine organs. These organ sites have all been previously associated with diseases involving SYK-related pathways (referred to here as SYK-irAEs). For example, SYK induces autoantibody-mediated skin inflammation,43,44and clinical trials with drugs targeting SYK to treat autoimmune vitiligo are ongoing45while preclinical studies suggest an association between SYK upregulation and psoriasis.46Upregulation of SYK is also associated with liver diseases including liver fibrosis, alcoholic liver disease, and acute liver failure.47-50SYK activation in the GI tract stimulates inflammatory cytokine and chemokine production51and elevated SYK expression is observed in patients with inflammatory bowel disease75and colitis.52While SYK is not directly related to thyroid conditions, endocrine disorders have been shown to be associated with the upregulation of SMAD3 / 4, MAPK8 / 14, SRC, CSNK2A1, PRKCZ, and RPS6KA3, all of which are overexpressed in the SYK-cluster and associated with SYK in the PPI analysis .55-64Among patients with the SYK-GEP, 96% experienced a grade 1 SYK-irAE, and 75% experienced at least one grade 2 SYK-irAE. Even among the TOXlowpatients, 50% experienced at least one grade 2 SYK-irAE. Many patients also experienced repeated instances of these SYK-irAEs. For example, a TOXlowpatient with the SYK-GEP experienced 11 grade 1 diarrhea episodes. Although these do not meet the CTCAE grading criteria for a “severe” irAE, the cumulative negative impact of even low-grade events on patients’ quality of life is significant. Patients with the SYK-GEP also tended to develop various SYK-irAEs in these organ sites; 79% of all SYK-GEP patients experienced at least two different SYK-irAEs, with some patients experiencing up to seven distinct SYK-irAEs in both the TOXhighand TOXlowgroups. This is in contrast to only 28% of patients without the SYK-GEP with at least two SYK-irAEs. Further, several toxicities directly associated with SYK activity, including psoriasis,46vitiligo,45and colitis,52were nearly exclusively experienced by SYK-GEP patients. Taken together, both the frequency and specificity of these SYK-irAEs support a role of SYK in mediating irAE development in the SYK-GEP patients. The disclosure also shows that the SYK-GEP is not associated with tumor recurrence (the treatment efficacy outcome in CheckMate-915) in COMBO-treated patients, suggesting that it is an independent predictor of irAEs. The interactome analyses described herein suggest possible molecular pathways driving SYK-mediated irAE development. The results implicate the involvement of PAG1, a negative regulator of T cell activation that can also decrease the activity of LYN, a kinase that activates SYK following B cell receptor signaling.76,77In the 24-gene SYK-GEP, PAG1 was downregulated while LYN was upregulated, suggesting these patients may exhibit increased T cell activation and more activated SYK. Existing treatments to reduce activated SYK in patients with autoimmune disease include abatacept, a CTLA-4 fusion protein that inhibits T cell activation and results in decreased pSYK in circulating B cells in patients with RA.73While administering CTLA-4 to address toxicities induced by anti-CTLA-4 treatment would not be desirable in this treatment context, the disclosure may provide opportunities to target axes upstream of pSYK, such as the one involving PAG1 and LYN, to reduce anti-CTLA-4- mediated toxicity without compromising efficacy. This may lead to a greater mechanistic understanding of the cell-specific roles of these genes, their interaction with SYK, and their association with irAE development, towards more selective treatments for the AEs without compromising possibly beneficial therapy outcomes. Given the plausible biological underpinning of the association between the SYK-GEP and irAE development, the disclosure provides refinement of the gene signature to those most associated with irAE risk and incorporate clinical predictors of toxicity to better identify the patients who would benefit most from such toxicity-based treatment selection. To achieve this, we developed and cross-validated a LASSO-regularized predictive model incorporating clinical information and SYK-GEP expression. We employed LASSO regularization as it tends to produce parsimonious models, performing feature selection by shrinking to zero the coefficients of the variables with the least contribution to model prediction. The resulting model is less likely to be over-fit to the data, as the regularization penalty reduces variance, and contains only those variables with the greatest contribution to the model. Repeated iterations of cross-validation also ensured the model was trained and validated on many randomly selected cross-validation splits, further reducing the risk of over-fitting by identifying variables that were most stable over repeated iterations. With 82.1% accuracy for predicting severe irAEs in COMBO-treated patients, the model correctly predicts nearly all TOXlowpatients (98% specificity), thereby delineating those who, due to predicted high toxicity, would require further treatment stratification. The model is parsimonious, with only 5 out of 24 SYK-related genes, further increasing its use for routine clinical testing. Although the rs7036417 genotype, previously found by us to associate with single-line anti-CTLA-4 toxicity, was also selected by LASSO as a variable in multiple predictive models, its addition did not significantly improve the performance of the final selected model. While the predictive utility of the SYK-GEP model in conjunction with rs7036417 would possibly be beneficial to test in anti-CTLA4 monotherapies, the data from COMBO indicate that the SYK-GEP is sufficient to identify TOXhipatients without the need for rs7036417 genotype information. An aspect of this disclosure is a well-established clinical trial (CheckMate-915), for which detailed CTCAE grading information is available for all irAEs, including data on type, severity, and organ-specificity. An additional strength is the sample size (n=212 patients), helping ensure adequate analytical power to compare irAE profiles between treatment types. This disclosure is one of the largest to date for irAE-based peripheral blood gene expression analysis. The disclosure also provides evidence of minimally invasive baseline PBMC-based toxicity biomarkers, which are of great clinical importance for treatment stratification, particularly in AT. This disclosure includes discussion on transcriptional data from T cells, but other immune cell types may be involved in irAE development; for example, SYK signaling in autoimmune conditions has been shown to modulate B cell activity and differentiation.78,79Also, while the SYK-GEP is derived from the COMBO arm of CheckMate-915, this combination therapy is not approved in AT.80-84Nevertheless, COMBO is currently being explored for patients with resectable disease (e.g. the IMMUNED85and NADINA86trials), and this disclosure provides opportunities for validation in these emerging treatment regimens, as well as in the metastatic setting for which COMBO is currently an approved standard-of-care treatment. This disclosure includes a sufficiently large sample size to minimize the likelihood of false positive findings (type I errors), with >95% power to detect an effect size of this magnitude (Figure 7), but without validation in an independent sample set. To address this limitation, the statistical approach (e.g. cross-validated LASSO regularization models) was designed to limit the risk of overfitting the model to our dataset, thereby improving the generalizability of these findings. Furthermore, the use of LASSO regularized models yielded the parsimonious 5-gene signature, which further enhances the feasibility of validation in future cohorts through targeted quantitative PCR assays, (e.g. TaqMan). Given the narrow gene target (5-gene signature) and minimal input material requirement, such targeted testing will allow for a high-throughput, accurate, and, relatively low-cost assessment of these findings in large-scale populations across treatment regimens and cancer types, even at cancer centers with limited resources. In conclusion, and without intending to be bound by any particular theory, it is considered this disclosure provides the first evidence of a baseline transcriptomic biomarker in circulating CD4+T cells identifying melanoma patients likely to develop SYK-associated irAEs when treated with COMBO ICI in the adjuvant setting. While it has been proposed that an increased risk of irAEs may be associated with better immune-mediated tumor control, this disclosure shows that this transcriptional profile is not associated with improved recurrence-free survival. In light of these findings, this disclosure indicates that this signature, which is detectable before treatment initiation, may not only identify patients more likely to incur irAEs upon anti-CTLA-4-based ICI regimens but could help guide toxicity-driven adjuvant treatment decisions without impacting overall ICI clinical efficacy. REFERENCES This reference listing is not an indication that any reference is material to patentability 1 Weber, J. S. et al. Adjuvant Therapy of Nivolumab Combined With Ipilimumab Versus Nivolumab Alone in Patients With Resected Stage IIIB-D or Stage IV Melanoma (CheckMate 915). J Clin Oncol 41, 517-527 (2023). doi.org:10.1200 / jco.22.00533 2 Khushalani, N. I. et al. Phase II clinical and immune correlate study of adjuvant nivolumab plus ipilimumab for high-risk resected melanoma. Journal for ImmunoTherapy of Cancer 10, e005684 (2022). doi.org:10.1136 / jitc-2022-005684 3 Kirkwood, J. M. et al. Adjuvant nivolumab in resected stage IIB / C melanoma: primary results from the randomized, phase 3 CheckMate 76K trial. Nat Med 29, 2835-2843 (2023). doi.org:10.1038 / s41591-023-02583-2 4 Luke, J. J. et al. Pembrolizumab versus placebo as adjuvant therapy in completely resected stage IIB or IIC melanoma (KEYNOTE-716): a randomised, double-blind, phase 3 trial. Lancet 399, 1718-1729 (2022). doi.org:10.1016 / s0140-6736(22)00562-1 5 Patel, S. et al. LBA6 Neoadjvuant versus adjuvant pembrolizumab for resected stage III- IV melanoma (SWOG S1801). Annals of Oncology 33, S1408 (2022). doi.org:10.1016 / j.annonc.2022.08.039 6 Hommes, J. W., Verheijden, R. J., Suijkerbuijk, K. P. M. & Hamann, D. Biomarkers of Checkpoint Inhibitor Induced Immune-Related Adverse Events—A Comprehensive Review. Frontiers in Oncology 10 (2021). doi.org:10.3389 / fonc.2020.585311 7 Gowen, M. F. et al. Baseline antibody profiles predict toxicity in melanoma patients treated with immune checkpoint inhibitors. Journal of Translational Medicine 16, 82 (2018). doi.org:10.1186 / s12967-018-1452-4 8 Johannet, P. et al. Baseline Serum Autoantibody Signatures Predict Recurrence and Toxicity in Melanoma Patients Receiving Adjuvant Immune Checkpoint Blockade. Clin Cancer Res 28, 4121-4130 (2022). doi.org:10.1158 / 1078-0432.Ccr-22-0404 9 Ferguson, R. et al. Germline immunomodulatory expression quantitative trait loci (ieQTLs) associated with immune-related toxicity from checkpoint inhibition. European Journal of Cancer 189, 112923 (2023). doi.org:10.1016 / j.ejca.2023.05.011 10 Chaput, N. et al. Baseline gut microbiota predicts clinical response and colitis in metastatic melanoma patients treated with ipilimumab. Annals of Oncology 28, 1368- 1379 (2017). doi.org:10.1093 / annonc / mdx108 11 Dubin, K. et al. Intestinal microbiome analyses identify melanoma patients at risk for checkpoint-blockade-induced colitis. Nature Communications 7, 10391 (2016). doi.org:10.1038 / ncomms10391 12 Valpione, S. et al. Sex and interleukin-6 are prognostic factors for autoimmune toxicity following treatment with anti-CTLA4 blockade. Journal of Translational Medicine 16, 94 (2018). doi.org:10.1186 / s12967-018-1467-x 13 Kurimoto, C. et al. Predictive and sensitive biomarkers for thyroid dysfunctions during treatment with immune-checkpoint inhibitors. Cancer Science 111, 1468-1477 (2020). doi.org:10.1111 / cas.14363 14 Tarhini, A. A. et al. Baseline circulating IL-17 predicts toxicity while TGF-β1 and IL-10 are prognostic of relapse in ipilimumab neoadjuvant therapy of melanoma. Journal for ImmunoTherapy of Cancer 3, 39 (2015). doi.org:10.1186 / s40425-015-0081-1 15 Khan, S. et al. Immune dysregulation in cancer patients developing immune-related adverse events. British Journal of Cancer 120, 63-68 (2019). doi.org:10.1038 / s41416- 018-0155-1 16 Lozano, A. X. et al. T cell characteristics associated with toxicity to immune checkpoint blockade in patients with melanoma. Nat Med 28, 353-362 (2022). doi.org:10.1038 / s41591-021-01623-z 17 Refae, S. et al. Germinal Immunogenetics predict treatment outcome for PD-1 / PD-L1 checkpoint inhibitors. Investigational New Drugs 38, 160-171 (2020). doi.org:10.1007 / s10637-019-00845-w 18 Groha, S. et al. Germline variants associated with toxicity to immune checkpoint blockade. Nature Medicine 28, 2584-2591 (2022). doi.org:10.1038 / s41591-022-02094-6 19 Taylor, C. A. et al. IL7 genetic variation and toxicity to immune checkpoint blockade in patients with melanoma. Nat Med 28, 2592-2600 (2022). doi.org:10.1038 / s41591-022- 02095-5 20 Titmuss, E. et al. Exploration of Germline Correlates and Risk of Immune-Related Adverse Events in Advanced Cancer Patients Treated with Immune Checkpoint Inhibitors. Current Oncology 31, 1865-1875 (2024). 21 Weidhaas, J. et al. Germline biomarkers predict toxicity to anti-PD1 / PDL1 checkpoint therapy. J Immunother Cancer 10 (2022). doi.org:10.1136 / jitc-2021-003625 22 Larkin, J. et al. Combined Nivolumab and Ipilimumab or Monotherapy in Untreated Melanoma. New Engl J Med 373, 23-34 (2015). doi.org:10.1056 / NEJMoa1504030 23 Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics (Oxford, England) 29, 15-21 (2013). doi.org:10.1093 / bioinformatics / bts635 24 Li, B. & Dewey, C. N. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC bioinformatics 12, 323 (2011). doi.org:10.1186 / 1471-2105-12-323 25 Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550 (2014). doi.org:10.1186 / s13059-014-0550-8 26 Clarke, D. J B. et al. eXpression2Kinases (X2K) Web: linking expression signatures to upstream cell signaling networks. Nucleic Acids Research 46, W171-W179 (2018). doi.org:10.1093 / nar / gky458 27 Lex, A., Gehlenborg, N., Strobelt, H., Vuillemot, R. & Pfister, H. UpSet: Visualization of Intersecting Sets. IEEE Trans Vis Comput Graph 20, 1983-1992 (2014). doi.org:10.1109 / tvcg.2014.2346248 28 Chen, E. Y. et al. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC bioinformatics 14, 128 (2013). doi.org:10.1186 / 1471-2105-14-128 29 Kuleshov, M. V. et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res 44, W90-97 (2016). doi.org:10.1093 / nar / gkw377 30 Xie, Z. et al. Gene Set Knowledge Discovery with Enrichr. Current Protocols 1, e90 (2021). doi.org:10.1002 / cpz1.90 31 Clarke, D. J. B. et al. Appyters: Turning Jupyter Notebooks into data-driven web apps. Patterns 2, 100213 (2021). doi.org:10.1016 / j.patter.2021.100213 32 Belinky, F. et al. PathCards: multi-source consolidation of human biological pathways. Database (Oxford) 2015 (2015). doi.org:10.1093 / database / bav006 33 Kolde, R. Pheatmap: pretty heatmaps. R package version 1, 726 (2012). 34 Friedman, J., Hastie, T. & Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw 33, 1-22 (2010). 35 Tay, J. K., Narasimhan, B. & Hastie, T. Elastic Net Regularization Paths for All Generalized Linear Models. J Stat Softw 106 (2023). doi.org:10.18637 / jss.v106.i01 36 Kuhn, M. Building Predictive Models in R Using the caret Package. Journal of Statistical Software 28, 1 - 26 (2008). doi.org:10.18637 / jss.v028.i05 37 olsrr: Tools for Building OLS Regression Models (2024). 38 Kim, J. H. Multicollinearity and misleading statistical results. Korean J Anesthesiol 72, 558-569 (2019). doi.org:10.4097 / kja.19087 39 Wei, T. & Simko, V. R package 'corrplot': Visualization of a Correlation Matrix. (Version 0.92). (2021). 40 Robin, X. et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC bioinformatics 12, 77 (2011). doi.org:10.1186 / 1471-2105-12-77 41 pwr: Basic Functions for Power Analysis (2020). 42 Park, R., Lopes, L., Cristancho, C. R., Riano, I. M. & Saeed, A. Treatment-Related Adverse Events of Combination Immune Checkpoint Inhibitors: Systematic Review and Meta-Analysis. Front Oncol 10, 258 (2020). doi.org:10.3389 / fonc.2020.00258 43 Németh, T., Virtic, O., Sitaru, C. & Mócsai, A. The Syk Tyrosine Kinase Is Required for Skin Inflammation in an In Vivo Mouse Model of Epidermolysis Bullosa Acquisita. J Invest Dermatol 137, 2131-2139 (2017). doi.org:10.1016 / j.jid.2017.05.017 44 Braegelmann, C. et al. Spleen tyrosine kinase (SYK) is a potential target for the treatment of cutaneous lupus erythematosus patients. Experimental Dermatology 25, 375-379 (2016). doi.org:10.1111 / exd.12986 45 Qi, F., Liu, F. & Gao, L. Janus Kinase Inhibitors in the Treatment of Vitiligo: A Review. Front Immunol 12, 790125 (2021). doi.org:10.3389 / fimmu.2021.790125 46 Alzahrani, K. S. et al. Inhibition of spleen tyrosine kinase attenuates psoriasis-like inflammation in mice through blockade of dendritic cell-Th17 inflammation axis. Biomed Pharmacother 111, 347-358 (2019). doi.org:10.1016 / j.biopha.2018.12.060 47 Zhao, Y., Liu, R., Li, M. & Liu, P. The spleen tyrosine kinase (SYK): A crucial therapeutic target for diverse liver diseases. Heliyon 8, e12130 (2022). doi.org:10.1016 / j.heliyon.2022.e12130 48 Qu, C. et al. Tyrosine kinase SYK is a potential therapeutic target for liver fibrosis. Hepatology 68, 1125-1139 (2018). doi.org:10.1002 / hep.29881 49 Bukong, T. N. et al. Inhibition of spleen tyrosine kinase activation ameliorates inflammation, cell death, and steatosis in alcoholic liver disease. Hepatology 64, 1057- 1071 (2016). doi.org:10.1002 / hep.28680 50 Sun, Y. et al. Identification of Hub Genes and Potential Molecular Mechanisms in Patients with HBV-Associated Acute Liver Failure. Evol Bioinform Online 16, 1176934320943901 (2020). doi.org:10.1177 / 1176934320943901 51 Gong, W. et al. The ubiquitous role of spleen tyrosine kinase (Syk) in gut diseases: From mucosal immunity to targeted therapy. Int Rev Immunol 41, 552-563 (2022). doi.org:10.1080 / 08830185.2021.1962860 52 Gong, W. et al. CCL4-mediated targeting of spleen tyrosine kinase (Syk) inhibitor using nanoparticles alleviates inflammatory bowel disease. Clin Transl Med 11, e339 (2021). doi.org:10.1002 / ctm2.339 53 Pan, M. et al. Knockdown and Inhibition of SYK Tyrosine Kinase Decreases mTORC1 Activity and Enhances Lysosomal Biogenesis and Autolysosomal Maturation in Pancreatic Cancer Cell Lines. The FASEB Journal 34, 1-1 (2020). doi.org:10.1096 / fasebj.2020.34.s1.01879 54 Rohila, D. et al. Syk Inhibition Reprograms Tumor-Associated Macrophages and Overcomes Gemcitabine-Induced Immunosuppression in Pancreatic Ductal Adenocarcinoma. Cancer Research 83, 2675-2689 (2023). doi.org:10.1158 / 0008- 5472.Can-22-3645 55 Wang, X. et al. Phosphorylated c-Jun NH2-terminal kinase is overexpressed in human papillary thyroid carcinomas and associates with lymph node metastasis. Cancer Lett 293, 175-180 (2010). doi.org:10.1016 / j.canlet.2010.01.007 56 Wang, Y. et al. The role of SMAD3 in the genetic predisposition to papillary thyroid carcinoma. Genet Med 20, 927-935 (2018). doi.org:10.1038 / gim.2017.224 57 Mohamed, A. H. et al. Prognostic Roles of ZNF703 and SMAD4 Expression in Patients with Papillary Thyroid Cancer and Association with Nodal Metastasis. Indian Journal of Surgical Oncology 13, 169-177 (2022). doi.org:10.1007 / s13193-022-01519-5 58 Cui, D., Zhao, Y. & Xu, J. Activation of CXCL5-CXCR2 axis promotes proliferation and accelerates G1 to S phase transition of papillary thyroid carcinoma cells and activates JNK and p38 pathways. Cancer Biol Ther 20, 608-616 (2019). doi.org:10.1080 / 15384047.2018.1539289 59 Zhang, P., Yang, M., Zeng, L. & Liu, C. P38 / TRHr-Dependent Regulation of TPO in Thyroid Cells Contributes to the Hypothyroidism of Triclosan-Treated Rats. Cell Physiol Biochem 45, 1303-1315 (2018). doi.org:10.1159 / 000487558 60 Rose, M. M. et al. BCL2L11 Induction Mediates Sensitivity to Src and MEK1 / 2 Inhibition in Thyroid Cancer. Cancers 15, 378 (2023). 61 Beadnell, T. C. et al. Src-mediated regulation of the PI3K pathway in advanced papillary and anaplastic thyroid cancer. Oncogenesis 7, 23 (2018). doi.org:10.1038 / s41389-017- 0015-5 62 Shu, C. et al. CircNDST1 promotes papillary thyroid cancer progression via its interaction with CSNK2A1 to activate the PI3K-Akt pathway and epithelial- mesenchymal transition. J Endocrinol Invest 46, 545-557 (2023). doi.org:10.1007 / s40618-022-01928-x 63 Williams, A. T. et al. Genome-wide association study of thyroid-stimulating hormone highlights new genes, pathways and associations with thyroid disease. Nature Communications 14, 6713 (2023). doi.org:10.1038 / s41467-023-42284-5 64 Zhan, S., Li, J., Wang, T. & Ge, W. Quantitative Proteomics Analysis of Sporadic Medullary Thyroid Cancer Reveals FN1 as a Potential Novel Candidate Prognostic Biomarker. Oncologist 23, 1415-1425 (2018). doi.org:10.1634 / theoncologist.2017-0399 65 Hodi, F. S. et al. Improved Survival with Ipilimumab in Patients with Metastatic Melanoma. New Engl J Med 363, 711-723 (2010). doi.org:10.1056 / NEJMoa1003466 66 Robert, C. et al. Nivolumab in Previously Untreated Melanoma without BRAF Mutation. New Engl J Med 372, 320-330 (2014). doi.org:10.1056 / NEJMoa1412082 67 Hodi, F. S. et al. Nivolumab plus ipilimumab or nivolumab alone versus ipilimumab alone in advanced melanoma (CheckMate 067): 4-year outcomes of a multicentre, randomised, phase 3 trial. The Lancet Oncology 19, 1480-1492 (2018). doi.org:10.1016 / S1470-2045(18)30700-9 68 Park, J. E. et al. The role of Syk in peripheral T cells. Clin Immunol 192, 50-57 (2018). doi.org:10.1016 / j.clim.2018.04.007 69 Deng, G.-M., Kyttaris, V. C. & Tsokos, G. C. Targeting Syk in Autoimmune Rheumatic Diseases. Frontiers in Immunology 7 (2016). doi.org:10.3389 / fimmu.2016.00078 70 Grammatikos, A. P., Ghosh, D., Devlin, A., Kyttaris, V. C. & Tsokos, G. C. Spleen tyrosine kinase (Syk) regulates systemic lupus erythematosus (SLE) T cell signaling. PLoS One 8, e74550 (2013). doi.org:10.1371 / journal.pone.0074550 71 Krishnan, S. et al. Differential Expression and Molecular Associations of Syk in Systemic Lupus Erythematosus T Cells1. The Journal of Immunology 181, 8145-8152 (2008). doi.org:10.4049 / jimmunol.181.11.8145 72 Iwata, S. et al. Increased Syk phosphorylation leads to overexpression of TRAF6 in peripheral B cells of patients with systemic lupus erythematosus. Lupus 24, 695-704 (2015). doi.org:10.1177 / 0961203314560424 73 Iwata, S. et al. Activation of Syk in Peripheral Blood B Cells in Patients With Rheumatoid Arthritis: A Potential Target for Abatacept Therapy. Arthritis & Rheumatology 67, 63-73 (2015). doi.org:10.1002 / art.38895 74 Wilcox, R. A. et al. Inhibition of Syk protein tyrosine kinase induces apoptosis and blocks proliferation in T-cell non-Hodgkin's lymphoma cell lines. Leukemia 24, 229-232 (2010). doi.org:10.1038 / leu.2009.198 75 Antonini, F., Pezzilli, R., Angelelli, L. & Macarri, G. Pancreatic disorders in inflammatory bowel disease. World J Gastrointest Pathophysiol 7, 276-282 (2016). doi.org:10.4291 / wjgp.v7.i3.276 76 Barua, D., Hlavacek, W. S. & Lipniacki, T. A computational model for early events in B cell antigen receptor signaling: analysis of the roles of Lyn and Fyn. J Immunol 189, 646-658 (2012). doi.org:10.4049 / jimmunol.1102003 77 Ullah, M. A. et al. PAG1 limits allergen-induced type 2 inflammation in the murine lung. Allergy 75, 336-345 (2020). doi.org:doi.org / 10.1111 / all.13991 78 Ackermann, J. A. et al. Syk tyrosine kinase is critical for B cell antibody responses and memory B cell survival. J Immunol 194, 4650-4656 (2015). doi.org:10.4049 / jimmunol.1500461 79 Roders, N. et al. SYK Inhibition Induces Apoptosis in Germinal Center-Like B Cells by Modulating the Antiapoptotic Protein Myeloid Cell Leukemia-1, Affecting B-Cell Activation and Antibody Production. Frontiers in Immunology 9 (2018). doi.org:10.3389 / fimmu.2018.00787 80 Weber, J. et al. Adjuvant Nivolumab versus Ipilimumab in Resected Stage III or IV Melanoma. New Engl J Med 377, 1824-1835 (2017). doi.org:10.1056 / NEJMoa1709030 81 Ascierto, P. A. et al. Adjuvant nivolumab versus ipilimumab in resected stage IIIB-C and stage IV melanoma (CheckMate 238): 4-year results from a multicentre, double- blind, randomised, controlled, phase 3 trial. Lancet Oncol 21, 1465-1477 (2020). doi.org:10.1016 / s1470-2045(20)30494-0 82 Larkin, J. et al. Adjuvant Nivolumab versus Ipilimumab in Resected Stage III / IV Melanoma: 5-Year Efficacy and Biomarker Results from CheckMate 238. Clin Cancer Res 29, 3352-3361 (2023). doi.org:10.1158 / 1078-0432.Ccr-22-3145 83 Eggermont, A. M. M. et al. Adjuvant pembrolizumab versus placebo in resected stage III melanoma (EORTC 1325-MG / KEYNOTE-054): distant metastasis-free survival results from a double-blind, randomised, controlled, phase 3 trial. Lancet Oncol 22, 643-654 (2021). doi.org:10.1016 / s1470-2045(21)00065-6 84 Eggermont, A. M. M. et al. Five-Year Analysis of Adjuvant Pembrolizumab or Placebo in Stage III Melanoma. NEJM Evid 1, EVIDoa2200214 (2022). doi.org:10.1056 / EVIDoa2200214 85 Livingstone, E. et al. Adjuvant nivolumab plus ipilimumab or nivolumab alone versus placebo in patients with resected stage IV melanoma with no evidence of disease (IMMUNED): final results of a randomised, double-blind, phase 2 trial. Lancet 400, 1117-1129 (2022). doi.org:10.1016 / s0140-6736(22)01654-3 86 Lucas, M. W. et al. The NADINA trial: A multicenter, randomised, phase 3 trial comparing the efficacy of neoadjuvant ipilimumab plus nivolumab with standard adjuvant nivolumab in macroscopic resectable stage III melanoma. Journal of Clinical Oncology 40, TPS9605-TPS9605 (2022). doi.org:10.1200 / JCO.2022.40.16_suppl.TPS9605

[0002] TABLES Table 1. Demographic data for the analytic cohort comprised of patients from CheckMate-915, used in the study. Ipilimumab / Nivolumab (COMBO) CheckMate-915 Patient Demographics Nivolumab (NIVO) CohortCohort Total In Follow-Up Recurrence In Follow-Up Recurrence SexFemale 92 (43.4%) 30 (39.0%) 22 (41.5%) 20 (42.6%) 20 (57.1%) Male 120 (56.6%) 47 (61.0%) 31 (58.5%) 27 (57.4%) 15 (42.9%)Age (years)Mean (SD) 54.1 (14.5) 55.1 (15.1) 54.2 (13.0) 53.2 (14.9) 52.9 (15.1) Median [Min, Max] 54.0 [16.0, 82.0] 56.0 [19.0, 82.0] 53.0 [31.0, 79.0] 54.0 [22.0, 79.0] 53.0 [16.0, 78.0] StageStage IIIB 68 (32.1%) 26 (33.8%) 15 (28.3%) 17 (36.2%) 10 (28.6%) Stage IIIC 107 (50.5%) 37 (48.1%) 30 (56.6%) 23 (48.9%) 17 (48.6%) Stage IIID 5 (2.4%) 1 (1.3%) 1 (1.9%) 1 (2.1%) 2 (5.7%)Stage IV 32 (15.1%) 13 (16.9%) 7 (13.2%) 6 (12.8%) 6 (17.1%)BRAF StatusWildtype 101 (47.6%) 36 (46.8%) 28 (52.8%) 23 (48.9%) 14 (40.0%) Mutant 66 (31.1%) 22 (28.6%) 16 (30.2%) 14 (29.8%) 14 (40.0%) Unknown 45 (21.2%) 19 (24.7%) 9 (17.0%) 10 (21.3%) 7 (20.0%) Severe irAE, Grade ≥3Severe irAE 46 (21.7%) 16 (20.8%) 5 (9.4%) 15 (31.9%) 10 (28.6%) No Severe irAE 166 (78.3%) 61 (79.2%) 48 (90.6%) 32 (68.1%) 25 (71.4%) RFS in monthsMean (SD) 19.0 (11.6) 25.0 (8.81) 10.2 (8.63) 26.8 (8.69) 8.78 (6.74) Median [Min, Max] 23.8 [0.03, 36.0] 27.6 [0.03, 36.0] 7.85 [0.79, 30.5] 28.8 [0.03, 36.0] 6.80 [0.26, 24.9]

[0003] Table 2. The sensitivity / specificity calculations for prediction of irAEs in COMBO-treated patients with the SYK gene expression profile (GEP) and those without.a. Comparing patients with TOXhigh and TOXlow by SYK GEPTOXhighTOXlowTrue Positive False PositiveGEP TotalsSensitivity: 56.5%13 (7.08) [4.96] 11 (16.9) [2.07] 24 Specificity: 80.0% No SYK GEP False Negative True Negative Accuracy: 73.1% 10 (15.9) [2.2] 44 (38.1) [0.92] 54 55 Chi-square statistic = 10.2, TOX Totals 78 (Grand Total) p-value=0.001** (Yates correction, 8.51, p- value=0.004)b. Comparing COMBO-treated patients with at least 2 SYK irAEs to those with <2 SYK irAEs by SYK GEP c. Comparing COMBO-treated patients who have recurred and those still in follow up by SYK GEPRecurrence Follow Up True Positive False PositiveGEP Totals Sensitivity: 32.4%SYK GEP 11 (10.5) [0.03] 13 (13.5) [0.02] 24 Specificity: 70.5% No SYK GEP False Negative True NegativeAccuracy: 53.9%23 (23.5) [0.01] 31 (30.5) [0.01] 54 Overall chi-square statistic = Outcome Totals 34 44 78 (Grand Total) 0.07, p-value=0.790 (Yates correction, 4.0E-05, p- value=0.985) Key: Numbers in cells: observed / actual total (expected total) [per-cell chi-square statistic]. Sensitivity = True Positive / (True Positive + False Negative) Specificity = True Negative / (True Negative + False Positive) Accuracy = (True Positive + True Negative) / Grand Total 32404.00.062166982517060000.00.025132 / 23 / 5538 4KP D A A M M S 23 444 404.0400.0.065418611184629007080 0000.0 00.0.0653 211 5 / 5 57 / / 286274 Z1C K K C P R R A P S M 45 6 565620 030 068 811 194 4140444.00.00.0297 130665325 099 20900100100.00.0.0466 053 4 / 1 53 / 4 / 7 2071A2K N K C S B C Y S U 78 9 56565 5 506 6 6.0 0 0 00.0.0.0.05512123620 9 958 8 87 22 71 00 47 8 524 6 702 2 2 200.00000000.0.0.0.0535 596 519 9 5 / 1 1 1 33 / 20 / 30 / 30 / 384E 11K1K1N C K T G S H C K S C S L A 1121314151 56506407.00.0.0922227400833 269921 703 30000.00.0.08363907 / 6 / 36 4 / 4 894B CB23D K K A R S P G M S 617181 75992129750011.10.0.02846518825376264758050060.0.00 0.0576076 45 / / 11 2 / 8 54212K K P1D A S C MRI910212 948558 8485 585148 767.141 1610.0.0.0.07025193686467122510810898406 240900911120 0 01. . .0 00 0 0.0.0539 790 91563 3 / / 1 1 11 7 / / / 2 78202121 2 8M P A B K D S O B D A H U C M C R 2232425262 571738 809.1.1.1.10 0 0 0.016501 8 75 01 5 00622 3 9872 9 634849303141516170 0 01. . .0 00 0 0.0.0111 124 39662 / 19 / 2 / 1 / 3 / 2121342141A 4 A D09B3K K N D P U E S B P A J N HKIM 8292031323 978 231164652 803 99 0 0 11 21 2 21 5.2. .2 20 0 0.0.0.091695 6 1 65 43 7 1 10 53 0 6 16 33 4 0 355 4 2 988 909 6 71911 2 70 02 2 2 2.0 0 0 00.0.0.0.0.0263 0 8 1 916 / 1213121823 / 23 / 28 / 10 / 28 / 16321D5A2A2C R KTOAPK N M TLT S S S P C M H C 435363738393 B WM S C T U F S CN S HJT A A R L Psvni 86 67ebe2da mrr ooCcg- S 215323ai uju345 tdla.36n Av0.30.0ereffi 605498 5D-0E -41 742 7+4 eu7E 39783958l . .1 4 2Dav1 19310454 C- .0.0O P 00.0VINp5r alr 866 2251613o / 1 1fe2 / 2 / 8I v 1 / 5 / P O6 92 2 1PdetA1ci2 2F d4e n1R M Rr iL R U PeK :toK P O S S1 rT A P R MbSP B M elbkn2 3a a4 444TR1 2 4040404 4 4 184 9 4 380 4 1 1 1 864 6- - -0-0-0-16 15 51503160795 4 0 9E8E E E E E9 39 84 7 7 588 4534950 25 5194.775814839 93 27 7 7 383 9 6 8 2771.2.3.4.4.511 160 00.0.1701702802 240 00 38034 0 6610405060 00 8000 0.00.00.00. .00 0.00.00.00.00. .00 0.019822255576996675 8 2 4 5 0 0 3 6 9 0 95 38 8 5 1 1 7 5 4 0 3 4 8 / / 2 / 3 / 6 / 6 / 1 / 1 / 1 / 1 3 2 2 3 5 2 4 1 1 28 6 5 6 8 8 4 4 4 / 4 / 5 / 4 / 4 / 5 / 6 / 4 / 5 / 3 / 3 / 4K23 9 1 8 2 1AD KBBAP K2C D6 1 1 21R A P1A T2K3 1K K K S B K1N F R P AR1A F K T P K N A F A P A A R N S M DKA E DGM R R T A S E S M A C GIP E H M HIS T A M S P C 45 6 7 8 90111213141516171819102122232 148 56 23 49574366979534706 5 3 1 4 4 5 3 2 4 1 1 1 4 71781 0 7 0 5 1 0 8 7 74 35975944358919273298989888323498 7541255065339324 636 0 7 9 4 9 4 9 2 9 2 2 2 0 32 3 0 808 3 7 0 9 00091858995199604377734471 1 1 5 59 9 0 1 3 3 4 4 5 7 7 9 0 14 4 4 7 10 0 1 101 12 5 435 3 4 5 6 6 6 7 80 0 0. 101010 01010 0101020202020 03 40404 4 4 4 4 4. .0. . .0.0.0.0.0.0 0.0 0 0 0 0 0 00.0 0 0 0.0.0.0.0.0.0.0.0.0.0.0.0.0.0751 264127437393886474895 3 6 1 1 9 3 1 2 4 5 5 5 7 9101 3 17 8 1 2 3 5 6 2 2 2 2 2 2 2 9 / 3 / / / / 3 / 3 / 3 / 1 / 3 / 5 / 1 / 3 / 3 / 2 / 2 / 2 / 6 2 1 1 1 1 1 1 1 43 4 3 4 3 4 4 4 3 4 5 3 4 4 3 3 3 / 5 / 3 / 2 / 2 / 2 / 2 / 2 / 2 / 2 / 41A L 31 3P K1R1P1B Q31A1A C B N B CL1 1 4A0A A1 1 2F A R D1B A C OKA D01B2F R R3IC1 1A U3A L4T H A6P A9K R U A3V H B HKIP M H B E N P R T C WD K M S S V E C C A Y H R P A C P L R S D R COP S R M B S H C A B1NKA D A C U M C M S GIP C C G 526272829203132333435363738393041424344454647484940515 1 D9P L S Y H; M1;O61XIF OIF ;;11AR2 1B K D1P P F N S B G SHR S TC RA ;3;4;2; F;S S P2BGARIMPIB E1T ZLI134.2 59.975 3.77.5.171718060989 4.3.8 83.5.5.5342525252 24 4 45 5020 042042002028080 068 8 816 6 6 6.10.10.10.10.02463959 93 5 95 15 29 24 32 2 440 0 880 1 1 74904949 9.0 04 40.0.00.00.00020929 05 5 12 3 / / 1 14 4 / 2 / 2 / 2Z A1 9D8H3B P 5 G A S WPTM A Y TIS C 2535455565 8..0442.22.07.09.3391.07.119..20.97.69016.84365027.6075 ..5.3.9129914 5 9 4 1 1 4 6 1 3 10. 7 3 9 8 8 7 38611 4 4 1 1 5 3 4 5 5 3 2 5 1 1 1 9 1 9 543.62B K36 2K3A0U 7 P GC12P D AA7B NR3 2 1C3 23P NCRA CC C S R K P A N G T F C B G B3A N R3G3 3K E P Z T T S B AKIR D WS NKSICKIKPIL R E P MYLTIC C U GAPP G C M H P F P R P C 518.19121 8 7617 5131 4 5515151710211 7516121714.0.8.2.7.7.13. .1.7.1.9.5.3.9.5.7.1.2.9.8. .3 2 9 8 8 0 359 2 4 7 8 8 8 6 8 5 8 29 36582398204297355 903366818735730713856460622427 9 6 0 5 5 895 3 7 2 7 0 3 4 5 05 1 4 21 0 1 1 2 2 011 1 0 0 1 1 1 1 14 5 7 5 70 0 0 0 0 0 0 0 0 0 0 0 0 0 0 01 1 2 1 1 1 10 0 0 0 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0 000000000000000000000GS GS GS GS GS GS GS GS GS GS GS GS GS GS GS GS GS GS GS GS GS GS GS N N N N N N N N N N N N N N N N N N N N N N N E E E E E E E E E E E E E E E E E E E E E E E 79 1 5 7 5 3 5 9 71 2 4 2 7 0 9 1 7 585 4 9 6 2 4 7 7 8 7 2347212182740595868686775801011121314151516171,ega(sE r i )4-3edar g.(u elra< evv-epuleas,8v.- ci=.dec1r %% it,pls sitn%%o% i9. 9. 3etdneai :: tstc6e y7 0t :9.:38otmayptiyvt:ic yecrraroicci ytxiyvt:ic ycnd it if a u o iif aoisetsaires increuqcs-sietttiads ien creucert- eS pcS AhCY(tc eS pcS A r l sve-rde dsp r e ncidsx p tor e -lcairxo etce eddeGtciddnr(la ern egoP87 utcP amd C.ae2szvgniRirGyti ] ra] ]aluC gF ci]xw eolee 62v .5[ev1i.p1[ mo ytiwolee4v24.3 ev7i.2[r d-n otlXsliatis )7 e1urta)8. 55c ciXsliati[et )s )4ur a 69Oa,ayO Fo .Tge 5PxoO Fo0.Tge .S UutctiT P9(N4(E GTT P5(N5(S Pa ci 23ANg x5Kla 101L Rn oY utpoNirtrH a TplSca hAue ] ]tiPe ] ]o,f3m 3o txic chvgi i4t.i31[e8v6.2w [eE hvgi i4ti 7.2[e2v3.2rD taCxA hs, irtXo) esO P3e8.li3ata)2s Ghs. 3Fge 912 ohtK Xo) es it [P6Y Oe9.l5a a)Fge 40.m1 an GmTur( ( 1 N2rofS Tur(70N(o 3i AnsPu, oT1 1ixf2siT1h wys rth n2ufgiholtilagcat miw holoDnod X Xi od X X C. C c eytOxTn eytO 3d lSn l taci icO dixT ToToils tuci icO T T ela,re eoau fndiexobe v rP Ttc o rP T agaOACTts.a .b p,a0v.11= citsi oistcyt07ta itee irci1x :9:y 8t:s ceru t qc dvtitcifcaoruqcs-sie et is icus sehta ci nec-ci taCY( deerS p S AhCYp(-ledsolatmo.)l sT av y6ti 4tc)loytTi icix22a5toxoT Tdnaortld dnaetcarG(ut icdeG r(42a 4gP5nirapmocyt]i4]1cw0oev2.0 ev .01 P 1 EixleGoXsliti[)euiT Oaso 36. rt [a)4Tg .244KlYaT F P1( e4(S u1N3t t4u coAhtiP Eev]7]w hi 0 e403eG1s gitioK hs .ho 1[) esvli.t0[) 0t Y X P7rS O3. a ag361of oTeur0(Fe .N9(xN9siT1lrtatahgmiw hol yti slcaod TnX Xitooi et ytiO OxoT scuicT TTf dinxlaoer oTutCPc.cA

Claims

What is claimed is:

1. A method for determining if an individual is likely to experience immune-related adverse events (irAEs) if treated for melanoma with a combination of immune checkpoint inhibitors (ICIs) that are an anti-Cytotoxic T-Lymphocyte-Associated Protein 4 (CTLA-4) antibody and an anti-programmed death-ligand 1 (PD-1) antibody, and optionally treating the individual for the cancer, the method comprising: a) testing a biological sample obtained from the individual before treating the individual with the anti-CTLA-4 antibody or the anti-PD-1 antibody or a combination thereof, and from the biological sample determining expression of a plurality of genes, wherein said plurality of genes comprises at least CD22, PAG1, CD33, HNRNPU, and FCGR2C; b) comparing expression of the CD22, PAG1, CD33, HNRNPU, and FCGR2C genes to a control expression value for the CD22, PAG1, CD33, HNRNPU and FCGR2C genes obtained from individuals who had melanoma and did not experience irAEs after treatment with the anti-CTLA-4 antibody or the anti-PD-1 antibody or a combination thereof; wherein a difference in expression of the CD22, PAG1, CD33, HNRNPU, and FCGR2C genes relative to the control expression value indicates the individual is likely to experience the irAEs, and optionally treating the individual with immune checkpoint inhibition that does not comprise a combination of the anti-CTLA-4 antibody and the anti- PD-1 antibody, or treating the individual with an immune checkpoint inhibition (ICI) agent that is not the anti-CTLA-4 or the anti-PD-1 antibody; or wherein no difference in expression of the CD22, PAG1, CD33, HNRNPU, and FCGR2C genes relative to the control expression value is determined, the method optionally further comprising treating the individual with a combination of the anti-CTLA-4 antibody and the anti-PD-1 antibody, and wherein optionally, the anti-CTLA-4 antibody is ipilimumab, or the anti-PD-1 antibody is nivolumab, or the anti-CTLA-4 antibody is ipilimumab and the anti-PD-1 antibody is nivolumab.

2. The method of claim 1, further comprising determining a difference in expression of genes in the biological sample obtained from the individual to determine expression of genes selected from the group consisting of RPS6KB2, EZR, PTK2, MAPK3, LYN, ITGB3, CTTN, CSF3R, UBC, GAB2, PIK3CA, GRB2, SWAP70, MSN, PIK3CG, PIK3AP1, PIK3CD, PRKCA, CLEC7A, or a combination thereof, and comparing expression of one ormore of said genes to a control expression value for the same one or more genes obtained from individuals who did not experience irAEs after treatment with the anti-CTLA-4 antibody or the anti-PD-1 antibody or a combination thereof.

3. The method of claim 1 or claim 2, wherein the determining the expression of the plurality of genes from the individual and the control expression value is determined by measuring mRNA levels for each of said genes.

4. The method of claim 3, wherein at least one of the following is true: c) expression of the CD22 mRNA for the individual is higher than its control expression value; d) expression of the PAG1 mRNA for the individual is lower than its control expression value; e) expression of the CD33 mRNA is higher than its control expression value; f) expression of the HNRNPU mRNA is higher than its control expression value; g) expression of the FCGR2C mRNA is higher than its control expression value; h expression of the RPS6KB2 mRNA is higher than its control expression value; i) expression of the EZR mRNA is higher than its control expression value; j) expression of the PTK2 mRNA is higher than its control expression value; k) expression of the MAPK3 mRNA is lower than its control expression value; l) expression of the LYN mRNA is higher than its control expression value; m) expression of the ITGB3 mRNA is lower than its control expression value; n) expression of the CTTN mRNA is lower than its control expression value; o) expression of the CSF3R mRNA is higher than its control expression value; p) expression of the UBC mRNA is higher than its control expression value; q) expression of the GAB2 mRNA is higher than its control expression value; r) expression of the PIK3CA mRNA is higher than its control expression value; s) expression of the GRB2 mRNA is higher than its control expression value; t) expression of the SWAP70 mRNA is higher than its control expression value; u) expression of the MSN mRNA is higher than its control expression value; v) expression of the PIK3CG mRNA is higher than its control expression value; w) expression of the PIK3AP1 mRNA is higher than its control expression value; x) expression of the PIK3CD mRNA is lower than its control expression value; y) PRKCA is lower than its control expression value; orz) CLEC7A is higher than its control expression value.

5. The method of claim 4, wherein at least all of c)-g) are true.

6. The method of claim 4, comprising treating the individual with a melanoma therapy that does not comprise administering to the individual a combination of an anti-CTLA-4 antibody and an anti-PD-1 antibody.

7. The method of claim 3, wherein at least one of the following is not true: c) expression of the CD22 mRNA for the individual is higher than its control expression value; d) expression of the PAG1 mRNA for the individual is lower than its control expression value; e) expression of the CD33 mRNA is higher than its control expression value; f) expression of the HNRNPU mRNA is higher than its control expression value; g) expression of the FCGR2C mRNA is higher than its control expression value; h) RPS6KB2 is higher than its control expression value; i) EZR is higher than its control expression value; j) PTK2 is higher than its control expression value; k) MAPK3 is lower than its control expression value; l) LYN is higher than its control expression value; m) ITGB3 is lower than its control expression value; n) CTTN is lower than its control expression value; o) CSF3R is higher than its control expression value; p) UBC is higher than its control expression value; q) GAB2 is higher than its control expression value; r) PIK3CA is higher than its control expression value; s) GRB2 is higher than its control expression value; t) SWAP70 is higher than its control expression value; u) MSN is higher than its control expression value; v) PIK3CG is higher than its control expression value; w) PIK3AP1 is higher than its control expression value; x) PIK3CD is lower than its control expression value; y) PRKCA is lower than its control expression value; or z) CLEC7A is higher than its control expression value.

8. The method of claim 7, wherein at least all of c)-g) are not true.

9. The method of claim 8, comprising treating the individual with a melanoma therapy that comprises administering to the individual a combination of the anti-CTLA-4 antibody and the anti-PD-1 antibody, and wherein optionally, the anti-CTLA-4 antibody is ipilimumab, or the anti-PD-1 antibody is nivolumab, or the anti-CTLA-4 antibody is ipilimumab and the anti-PD-1 antibody is nivolumab.

10. The method of claim 1, wherein determining if the individual is likely to experience irAEs comprises determining if the irAEs are likely to be grade of 3-5, and optionally administering to the individual one or more ICI agents that are not the anti-CTLA-4 antibody or the anti-PD-1 antibody.

11. The method of claim 1, wherein determining if an individual is likely to experience irAEs comprises determining if the irAEs are likely to be absent or a grade of 1 or 2, and optionally administering to the individual one or more ICI agents that are the anti-CTLA-4 antibody and the anti-PD-1 antibody, and wherein optionally, the anti-CTLA-4 antibody is ipilimumab, or the anti-PD-1 antibody is nivolumab, or the anti-CTLA-4 antibody is ipilimumab and the anti-PD-1 antibody is nivolumab.

Images