Patent Application
20230141003
The instant application contains a Sequence Listing which has been submitted in .xml format and is hereby incorporated by reference in its entirety. Said .xml file is named “058636 00562.xml”, was created on Nov. 7, 2022, and is 790,199 bytes in size.
Adjuvant immune checkpoint blockade (ICB) produces clinical benefit for a subset of resected melanoma patients, but many individuals develop disease recurrence, and a substantial proportion develop immune-related adverse events (irAEs) (Eggermont et al., 2016; Weber et al., 2017). Toxicity can be severe enough to necessitate the interruption or permanent discontinuation of immunotherapy and may require treatment with systemic immunosuppressive agents. Even with appropriate clinical management, irAEs can lead to lifelong secondary conditions or, in rare cases, death (Michot et al., 2016; Postow et al., 2018). Thus, there is an urgent need to identify biomarkers of immunotherapy response and toxicity. Ideally, there would be a single assay that simultaneously risk-stratifies patients according to their likelihood of suffering recurrence or developing irAEs, which would help optimize patient selection for treatment. In patients who are at high risk for developing severe irAEs but proceed with treatment, it would facilitate monitoring and enable early intervention should toxicities develop. Thus, there is an ongoing need for identification and uses of biomarkers to predict responses to immune therapy and provide medical interventions based on the identification. The present disclosure is pertinent to this need.
The present disclosure provides compositions and methods for determining baseline serum autoantibodies (autoAbs) for use in patient selection in connection with treatment with immune checkpoint blockade agents. The patients can be stratified into categories that include being at high risk for cancer recurrence or not, and for predicting whether or not a patient will experience toxicity in response to immune checkpoint blockade therapy. The disclosure includes all compositions and methods that are used to assess the autoAbs. The disclosure also includes medical interventions to reduce the risk of recurrence of cancer, to reduce toxicity that is predicted to occur if immune blockade therapy is used, and to select patients who are candidates for immune blockade therapy with low risk of severe toxicity, recurrence, or a combination thereof. The disclosure includes administering the immune blockade therapy to selected patients, administering anti-cancer agents that are not predicted to cause toxicity to selected patients, and administering agents to reduce predicted toxicity in selected patients. Thus, the disclosure provides pretreatment autoAb profiles correlate with the development of severe immune related adverse events (IRAEs) and moreover that autoAb profiles can be used to predict disease recurrence following treatment with adjuvant immune checkpoint blockade. As such, the disclosure provides signatures of autoAbs that can be used to simultaneously predict immunotherapy response and toxicity.
The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.
Unless defined otherwise herein, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.
Unless specified to the contrary, it is intended that every maximum numerical limitation given throughout this description includes every lower numerical limitation, as if such lower numerical limitations were expressly written herein. Every minimum numerical limitation given throughout this specification will include every higher numerical limitation, as if such higher numerical limitations were expressly written herein. Every numerical range given throughout this specification will include every narrower numerical range that falls within such broader numerical range, as if such narrower numerical ranges were all expressly written herein. All protein sequences described herein include all isoforms of such proteins, e.g., proteins made from splice variants, and proteins that may vary from individual to individual in certain amino acids. Thus, all proteins described herein include proteins that have from 90.0-99.9% identity across their entire lengths to such proteins. The amino acid or polynucleotide sequence as the case may be associated with each GenBank or other database accession number of this disclosure is incorporated herein by reference as presented in the database on the effective filing date of this application or patent.
Aspects of this disclosure include each protein described herein, and all combinations of such proteins, wherein one or more of the proteins are present in vitro and are in contact with a biological sample obtained from an individual who has cancer. In embodiments, the individual from whom a first sample was obtained was not treated with any checkpoint inhibitor before the sample was obtained. In embodiments, the individual from whom a first sample is obtained has been diagnosed with any type of cancer. In embodiments, the cancer is a solid or liquid tumor. In embodiments, the cancer is renal cell carcinoma, breast cancer, prostate cancer, pancreatic cancer, lung cancer, liver cancer, ovarian cancer, cervical cancer, colon cancer, esophageal cancer, stomach cancer, bladder cancer, brain cancer, testicular cancer, head and neck cancer, melanoma or another skin cancer, any sarcoma, including but not limited to fibrosarcoma, angiosarcoma, adenocarcinoma, and rhabdomyosarcoma, and any blood cancer, including all types of leukemia, lymphoma, and myeloma. In a non-limiting embodiment, the biological sample is obtained from an individual who has been diagnosed with melanoma and was not treated with any checkpoint inhibitor before the sample was obtained.
In embodiments, a second, third, fourth, sample, etc. can be obtained from an individual who is undergoing treatment and tested to monitor the effect of the treatment, and keep steady, change, adjust, or discontinue treatment with a checkpoint inhibitor. In embodiments, the treatment is adjusted to prevent or mitigate the onset of irAEs, such as by administering an agent to the individual as described further herein.
A method of the present disclosure comprises screening for the presence of one or more sets of antibodies in a biological sample (such as blood, serum, plasma etc.) from an individual who is being considered as a candidate for therapy with one or more immune checkpoint inhibitors, and based upon the antibody profile, identifying the appropriate immune checkpoint inhibitors for administration to the individual, or determining that the individual should not be treated with a checkpoint inhibitor, or determining that the immune checkpoint inhibitors should be administered in conjunction with toxicity mitigation agents/process. The checkpoint inhibitors may be anti-PD-1, anti-CTLA-4, or a combination thereof. In embodiments, the checkpoint inhibitors are Ipilimumab used as a single checkpoint inhibitor therapy, or Nivolumab as a single checkpoint inhibitor therapy, or a combination of Ipilimumab and Nivolumab as a combination checkpoint inhibitor therapy.
In non-limiting embodiments, the disclosure provides a set of distinct autoAb signatures that can be used to predict the following two treatment outcomes for patients with advanced melanoma who receive adjuvant immunotherapy: (i) severe (Grade 3 or 4) versus non-severe (Grade 1 or 2) immune related adverse events; and (ii) disease recurrence versus no disease recurrence. Methods of grading of immune related adverse events is known in the art and are described further below. The distinct signatures predict ipilimumab efficacy, ipilimumab toxicity, nivolumab efficacy, and nivolumab toxicity. The efficacy and toxicity signatures for each treatment are combined so that patients can be stratified into one of four predicted outcomes: (i) efficacy and no severe toxicity; (ii) efficacy and severe toxicity; (iii) no efficacy and no severe toxicity; and (iv) no efficacy and severe toxicity. In certain approaches, predicting an individual will not have recurrence of cancer after treatment with a checkpoint inhibitor indicates the checkpoint inhibitor, or combination of checkpoint inhibitors, will have efficacy.
The signatures are developed by analyzing pre-checkpoint inhibitor samples for binding to a plurality of proteins. In embodiments, the plurality of proteins used in the described compositions and methods are selected from Table X.
With respect to compositions and methods of this disclosure, antibodies, if present in the biological sample, bind with specificity to one or more proteins that are present in an assay that is designed to determine the presence, absence, and/or amount of such antibodies. Thus, in embodiments, the disclosure comprises exposing a biological sample to a protein array. In embodiments, the protein array comprises at least 50%, 60%, 70%, or 80% of the proteins in the human proteome. In embodiments, the protein array pertains to the proteins known as of the date of the filing of this application or patent. In embodiments, the protein array comprises at least one protein from Table X. In embodiments, the plurality of proteins attached to the substrate comprises fewer than 21,000 proteins. In embodiments, the plurality of proteins attached to the array comprise or consist of 1-283 proteins from Table X, inclusive, and including all numbers and ranges of numbers between 1-283. Thus, in embodiments, the only proteins in the plurality of proteins attached to the substrate are selected from the proteins of Table X. In embodiments, the array comprises all of the proteins described in Table X.
Table X includes Tables A, B, C, D, E, and F. When reference to Table X is made, unless stated otherwise, all of Tables A, B, C, D, E, and F are included. Each of Tables A, B, C, D, E, and F include a plurality of proteins and an indication of what the auto-antibody signature prediction associated with auto-antibodies to each protein is. Specifically, Table A refers to proteins used for measuring autoantibodies to predict cancer recurrence when a patient treated with Ipilimumab as a monotherapy. Table B refers to proteins used for measuring autoantibodies to predict severe toxicity when a patient is treated with Ipilimumab as the only checkpoint inhibitor. Table C refers to proteins used for measuring autoantibodies to predict cancer recurrence when a patient is treated with Nivolumab as the only checkpoint inhibitor. Table D refers to proteins used for measuring autoantibodies to predict severe toxicity when a patient is treated with Nivolumab as the only checkpoint inhibitor. Table E refers to proteins used for measuring autoantibodies to predict cancer recurrence when a patient treated with Ipilimumab and Nivolumab as a combination therapy. Table F refers to proteins used for measuring autoantibodies to predict severe toxicity when a patient is treated with Ipilimumab and Nivolumab as a combination therapy.
Thus, the disclosure provides one or more arrays that comprise substrates with one or more of the described proteins in Table X that are reversibly or irreversibly attached to the substrate. The disclosure includes the described proteins and substrates that are in contact with a sample from an individual who has cancer, such as melanoma. The disclosure also includes the described protein-substrate combinations wherein auto-antibodies in the patient sample are bound to at least some of the proteins on the substrate.
Table X relates to Table Y. Table Y provides amino acid sequence information for the described proteins. The left column of Table Y provides modified sequences that were used in embodiments of the disclosure. The right hand column provides unmodified human protein sequences that may also be used in embodiments of the disclosure.
In embodiments, the amount of antibodies bound to a proteome array is scored, for example according to the Common Terminology Criteria for Adverse Events (CTCAE). Samples may be divided into groups as further set forth herein. Thus, the disclosure provides compositions and methods for antibody profiling. The antibody profiling may be carried out prior to treatment, any time during the treatment or any time after the treatment. The profiling may be carried out once or multiple times over any period of time.
In one aspect, the present disclosure provides methods for enhancing the efficacy of treatment of cancer, such as melanoma, with immune checkpoint inhibitors. The disclosure also provides panels for detection of subsets of antibodies that can form a basis for treatment decisions in the treatment of cancer, such as melanoma. The disclosure also provides kits for detection of specific antibodies.
In embodiments, a method of this disclosure comprises: a) obtaining a sample of a biological sample, such as blood, plasma or serum, b) determining antibodies using a protein array; and c) based on the profile of the antibodies, determining that the individual is not a candidate for a checkpoint inhibitor, or administering one or more immune checkpoint inhibitors to the individual. The method can further comprise administering to the individual agents to mitigate expected or observed toxicity from the checkpoint inhibitors.
The amount of antibodies, or a change in the level of antibodies, means a level that is measured against a suitable reference, such as a reference value. The reference may be established from a population of relevant individuals from which group the distinction is to be made. For example, the reference can be an average value from a group of individuals who have not shown toxicity, shown mild toxicity, or shown severe toxicity to the particular treatment. These values could be used as references for no toxicity, mild toxicity or severe toxicity, or site-specific toxicity. Other references can be obtained in a similar manner. For no toxicity, individuals who have not been treated at all may also be used.
The presence, absence and amount of antibodies in a patient sample can be detected by methods that are known in the art. For example, any type of immunological assay or antigen binding assay may be used. A commonly used assay is ELISA. Detection of the antigen-antibody complex is generally done by using detectable (fluorescent, luminescent, chemiluminescent, radioactive etc.) labels.
In one embodiment, a patient who is predicted to develop severe toxicity (or severe toxicity affecting specific organ/tissue sites or likely requiring treatment termination) could be monitored for the development of toxicity, or could be treated with a different dosage of immune checkpoint inhibitor(s). Such monitoring could allow clinicians to intervene e.g. with steroids, to mitigate toxicity (immune related adverse events) as they develop.
In one embodiment, the present methods can also be used in adjuvant immune checkpoint blockade in earlier stage (3 or 2) melanoma as being able to identify patients at risk of severe toxicity would be especially beneficial in the adjuvant setting, where there is less tolerance for severe toxicity. If an indication of likelihood of toxicity is observed, then steps can be taken to mitigate the toxicity, or the treatment regimen of anti-CTLA-4 and/or anti-PD-1 can be interrupted or the dose reduced. For mitigating toxicity, corticosteroid treatment may be administered. For example, prednisone may be administered orally or via i.v. For skin rashes, topical corticosteroids may be used. Another approach is to administer a tumor necrosis factor-alpha (TNF-α) inhibitor prior to or concurrent with one or a combination of immune checkpoint inhibitors. A non-limiting embodiment of a suitable TNF-α inhibitor is infliximab, but other TNF-α inhibitors may also be used. Non-limiting examples of other suitable TNF-α inhibitors include Infliximab-abda, Infliximab-dyyb. Adalimumab, Adalimumab-adaz, Adalimumab-atto, Certolizumab pegol, Etanercept, Etanercept-SZZS, and Golimumab. Other treatments for steroid-refractory irAEs—typically colitis—include: mycophenolic acid, or tacrolimus.
When recurrence of cancer is predicted, another anti-cancer agent can be used, e.g., any other anti-cancer agent that is not the immune checkpoint inhibitor or combination of immune checkpoint inhibitors for which recurrence is predicted after treatment. Alternatively dosing of the one or more checkpoint inhibitors can be change. In embodiments, a different checkpoint inhibitor or combination of checkpoint inhibitors is used.
With respect to the foregoing description, given the overlap in the clinical presentation of irAEs and conventional autoimmune disorders, the disclosure relates to the discovery that certain individuals possess a subclinical predisposition for ICB toxicity, which is characterized by the presence of preexisting autoantibodies (autoAbs) and does not necessarily manifest spontaneously but can be unmasked following checkpoint blockade. Without intending to be bound by any particular theory, it is considered that prior to the present disclosure, there has been no clinically validated and accurate tool for predicting immunotherapy efficacy and/or immune-related toxicity in melanoma patients.
The following Examples are intended to illustrate but not limit the disclosure.
Methods
Patient Population
This disclosure included 950 patients from two Phase 3, randomized, double-blind trials of adjuvant immunotherapy in resected melanoma. The first dataset was composed of 565 patients who were enrolled in CheckMate-238 (NCT02388906), which was an investigation of adjuvant nivolumab versus ipilimumab in patients with high risk, completely resected stage IIIB, IIIC, or IV melanoma, as defined by the American Joint Committee on Cancer (AJCC) 7th Edition. The cohort was retrospectively chosen by the trial sponsor (Bristol Myers Squibb), was enriched for immune-related adverse events, and included 408 patients who received ipilimumab and 157 patients treated with nivolumab.
The second dataset consisted of 385 patients who were enrolled in CheckMate-915 (NCT03068455), which evaluated adjuvant nivolumab plus ipilimumab versus nivolumab monotherapy in patients who underwent complete resection of AJCC 8th Edition stage IIIB, IIIC, IIID, or IV melanoma. The patients were chosen at random by the trial sponsor (Bristol Myers Squibb) who was blinded to their recurrence and toxicity outcomes at the time of selection. In total, there were 190 patients who received nivolumab and 195 patients who received ipilimumab plus nivolumab. For both the CheckMate 238 and CheckMate 915 patients, all clinical and demographic data were obtained directly from the trial sponsor
Clinical Outcomes
The outcomes of interest were: (1) disease recurrence versus no recurrence, and (2) severe (Grade 3-5) irAEs versus no or mild (Grade 1-2) irAEs. Briefly, recurrence-free survival (RFS) was defined as the time from randomization until the date of first recurrence, new primary melanoma, death from any cause, or last follow-up if none of the above occurred. Toxicity events were classified as either “related” or “not related” to the study drug and events categorized as “not related” were excluded from the current disclosure, as were events determined not to be immune-related. Treatment-related events with a potential immunologic etiology were identified using a list of prespecified terms from the Medical Dictionary for Regulatory Activities. For the present disclosure, the events were then categorized based on the body or organ system of origin. The severity of irAEs was graded using the National Cancer Institute (NCI) Common Terminology Criteria for Adverse Events (CTCAE), version 4.0. Grade 3, 4, and 5 irAEs were classified as severe and grade 1 and 2 irAEs were classified as mild.
Serum Preparation and Processing
Peripheral blood samples were prospectively collected within 72-hours of administering the first dose of study medication (day 1 of week 1). The whole blood was allowed to clot at room temperature. The samples were then centrifuged at room temperature at 1100-1300×g for 10 minutes (swing out) or 15 minutes (fixed) until the clot and serum were separated. The serum was next transferred into separate polypropylene tubes and ultimately stored at −70° C.
Serum Autoantibody Profiling
To profile serum autoantibodies, we utilized the HuProt Human Proteome Microarray v4.0 (CDI Laboratories, Mayaguez, PR), which contains over 21,000 unique, individually purified full-length human proteins in duplicate, covering more than 81% of the proteome (Jeong et al., 2012). The HuProt arrays were first blocked with blocking solution (5% BSA/1×TBS-T) at room temperature for 1 hour and then probed with serum samples (diluted 1:1000) at 4° C. overnight. Next, the arrays were washed with 1×TBS-T a total of 3 times (10 minutes per wash). The arrays were then probed with Alexa-647 labeled anti-human IgG Fc gamma fragment specific secondary antibody (Jackson ImmunoResearch, West Grove, Pa.) at room temperature for 1 hour. This was followed by three washes of 1×TBS-T (10 minutes per wash). The arrays were subsequently dried with an air duster and scanned using a GenePix 4000B instrument (Molecular Dynamics, Sunnyvale, Calif.). GenePix Pro (v7.2.30) software was used to measure the signal intensities for IgG binding to array features as well as any background signal present. Array signal intensities data were quality controlled for successful printing, staining, and scanner alignment using internal software tools (CDI Laboratories, Mayaguez, PR), which were used to confirm that duplicate spots retained >0.95 R2 signal intensity correlation across each array. The net signal was the background-subtracted median intensity of each antibody spot. This background-subtracted signal intensity was log 2 transformed and normalized to the median of the total signal intensities on the array for each subject. This was done as an internal control to normalize for the various overall levels of autoAb expressions among subjects. We then standardized each autoAb intensity by its mean and standard deviation across all subjects.
Prediction of Disease Recurrence and Severe Toxicity
Patients who received nivolumab from the CheckMate 238 cohort were randomly divided into discovery and test sets in a ratio of 75% to 25%. The nivolumab arm of the CheckMate 915 cohort was used as an independent validation dataset. Patients who received ipilimumab from CheckMate 238 and patients who received ipilimumab plus nivolumab from CheckMate 915 were each randomly split into discovery and test sets in a ratio of 75% to 25%. Descriptive group comparisons were performed using two-sample t-tests and chi-square tests for continuous and categorical variables, respectively.
The signatures for predicting recurrence and severe toxicity were independently derived from the discovery datasets for each treatment regimen but followed a parallel process. We first employed univariable two-sample t-tests to assess the association between every autoAb candidate and the outcome of interest. We categorized autoAbs as differentially expressed autoantibodies (DEA) if they met 2 criteria: (1) they had higher intensities in the no recurrence group compared to the recurrence group or in the severe toxicity compared to the no severe toxicity group, and (2) the univariate P value was less than 0.05. We then performed stability selection in combination with least absolute shrinkage and selection operator (LASSO) regression on the complete set of DEA to identify a parsimonious subset of autoAbs that would constitute the final signature (Tibshirani et al., 1996; Meinshausen et al., 2010). The cut-off value for stable selection was set at 75% (the percentage of times a variable was selected into a model).
The capacity of the autoAb signatures to predict recurrence or severe toxicity in the test and validation datasets was evaluated using area under the curve (AUC) of the receiver operating characteristic (ROC) curve. In addition, we identified a cutoff prediction score from the discovery sets and used it to stratify patients from the test and validation sets into high- versus low-risk groups. We selected cut points that achieved a minimum negative predictive value (NPV) of 80% to assure high accuracies for the patients predicted to have severe toxicity or no recurrence. For the nivolumab cohorts, the same cutoff used for the test set was applied to the independent validation dataset to assess its reproducibility. The sensitivity, specificity, positive predictive value (PPV) and NPV were estimated at this threshold. Recurrence free survival of the predicted high- versus low-likelihood of recurrence groups was compared by Kaplan-Meier curves using the log-rank test. Waterfall plots were generated to illustrate the performance of the toxicity signatures. Finally, the predictive utility of the autoAb signatures was assessed by comparing their AUC, by Delong's Test, to that of a model containing multiple clinical covariates including age, gender, BRAF mutation status, disease stage at study entry, ECOG performance status, baseline LDH (U/L), and PD-L1 status at a cutoff of 5%, which was the threshold used in CheckMate 238 (DeLong et al., 1998). All statistical analyses were performed using R (version 4.1.0, The R Foundation for Statistical Computing).
Functional and Enrichment Analyses of Differentially Expressed Autoantibodies
We started by using Metascape to perform broad-based functional analyses of the differentially expressed autoAb targets (metascape.org). We then performed more targeted enrichment analysis on select pathways from the Kyoto Encyclopedia of Genes and Genomes (KEGG), WikiPathways, and Human Phenotype Ontology (HPO) databases. This included lists of key genes associated with conventional autoimmune disease pathogenesis, cutaneous melanoma, and immunotherapy-related pathways. In addition, we independently compiled a list of proteins reported in conventional autoimmune disorders and a list of known melanoma autoantigens such as members of the Melanoma Antigen Gene (MAGE) and B melanoma antigen (BAGE) families (Hodi, 2006; Tio et al., 2019). The statistical significance of autoAb enrichment was calculated using the hypergeometric test (Federico et al., 2020).
Results
Study Population
Tables 1a-c show that the baseline clinical and demographic characteristics were generally well-balanced between the discovery and test datasets for each of the three regimens. However, as shown in Table 2 and Supplemental Table 1, there were several differences between the patients from CheckMate 238 and CheckMate 915. As a result, there was corresponding discordance between the nivolumab discovery and test sets (from CheckMate 238) and the validation set (from CheckMate 915). For instance, among the former two, more individuals had a baseline ECOG performance status of 1 rather than 0 (10.3% vs. 12.5% vs. 3.7%; P=0.029). Nivolumab patients from the discovery and test sets also had longer RFS than those from the validation set (24.73 vs. 27.53 vs. 21.03 months; P=0.002), but there was no significant difference in the percentage of patients to suffer recurrence (P=0.36). Finally, compared to patients from the validation set, there were more individuals from the discovery and test sets who experienced severe toxicity (39.3% vs. 45.0% vs. 22.1%; P=0.001), which was expected due to the enrichment sampling of the CheckMate 238 analysis cohort.
Prediction of Disease Recurrence and Severe Toxicity
The experimental design and workflow of the experiments are shown in
The nivolumab recurrence signature includes a total of 29 autoAbs. It performed with AUC 0.84 (95% CI: 0.71-0.97) and NPV 0.81 (95% CI: 0.53-0.96) on the test set of 40 CheckMate 238 patients. The signature achieved AUC 0.82 (95% CI: 0.75-0.88) and NPV 0.80 (0.70-0.88) when applied to the independent validation dataset of 190 CheckMate 915 patients (Table 3). Kaplan-Meier analyses show that the patients assigned to the high efficacy group had significantly better RFS than patients predicted to experience low efficacy (for the test set, p=0.017; for the validation set, p<0.0001;
The ipilimumab recurrence signature consists of 55 autoAb and performed with AUC 0.76 (95% CI: 0.66-0.85) and NPV 0.80 (95% CI: 0.63-0.87) on the test set of 102 CheckMate-238 patients (Table 3). Kaplan-Meier analyses showed that patients predicted to have a low likelihood of recurrence had significantly better RFS (p=0.0013) [
The recurrence signature for ipilimumab plus nivolumab includes 40 autoAbs. It performed with AUC 0.92 (95% CI: 0.85-0.99) and NPV 0.82 (95% CI: 0.60-0.99) on the test set of 49 CheckMate-915 patients (Table 3). Kaplan Meier analyses showed that patients could be stratified into two groups (high versus low risk for recurrence) with significantly different RFS (p<0.0001) [
Comparison to a Model Composed of Clinical Covariates and Percent PD-L1 Expression
For disease recurrence, the autoAb signatures outperformed the clinical model composed of predictors including age, gender, BRAF mutation status, disease stage at study entry, ECOG performance status, baseline LDH (U/L), and PD-L1 status, on the nivolumab test set (AUC 0.84 vs. 0.66; P=0.124), the nivolumab validation set (AUC 0.82 vs. 0.56; P<0.001), ipilimumab test set (AUC 0.76 vs. 0.59; P=0.028), and ipilimumab plus nivolumab test set (AUC 0.92 vs. 0.51; P<0.0001). For toxicity, the autoAb signatures outperformed the clinical model on the nivolumab test set (AUC 0.78 vs. 0.56; P=0.106), the nivolumab validation set (AUC 0.75 vs. 0.56; P=0.006), ipilimumab test set (AUC 0.79 vs. 0.55; P=0.001), and ipilimumab plus nivolumab test set (AUC 0.87 vs. 0.51; P=0.006, Supplemental Table 2]. Decision curve analyses showed that using the autoAb signatures added net clinical benefit as compared to using the clinical models, or the strategies of treating all patients or no patients (
Functional and Enrichment Analyses of Differentially Expressed AutoAb (DEA)
We approached the analysis of the DEA by investigating: (1) what are the enriched biological and functional roles of the DEA targets, and (2) are the DEA targets also known to be autoantigens in conventional autoimmune diseases, or melanoma neoantigens, or elements of either the PD-1 or CTLA-4 immune checkpoint signaling cascades?
To understand the processes driving baseline humoral immune system activity predictive of ICB response, we first compared the biological and functional roles of the DEA autoantigen targets for the six treatment and outcome combinations (
There was no significant overlap in the autoAbs that constituted the final prediction signatures (P>0.05 for all comparisons;
Given the similar clinical manifestations of immunotherapy toxicity and conventional autoimmune disorders, we next investigated whether the same autoAbs are associated with both processes. The HuProt microarray includes several established markers of autoimmune disease including dsDNA, JO-1, SCL-70, Smith (Sm) antigen, Sm/RNP complex, SSA, Lupus La protein (SSB), and thyroid peroxidase (TPO). SSB was found at higher levels in patients who received nivolumab and developed severe toxicity (P=0.0005), but there was otherwise no significant differential expression of these antibodies for any of the treatment-outcome combinations. We developed a more comprehensive list of autoAb commonly found in autoimmune disease but found that there was no significant enrichment for these antibodies. In an additional analysis of genes related to autoimmune disease pathogenesis, we found that the list of DEA for patients who received ipilimumab and developed severe toxicity was enriched for antigens related to autoimmune diseases (P=0.013). However, there was no significant enrichment for any of the other treatment/outcome combinations (P>0.15 for all remaining) [Supplemental Table 3].
For each treatment-outcome combination, we explored whether the profile of baseline DEA was enriched for IgGs targeting proteins linked to melanoma pathogenesis. We included in this analysis an additional 47 known melanoma autoantigens. None of the treatment-outcome combinations were enriched for antibodies to this set of melanoma related proteins and neoantigens (P>0.20 for all).
Finally, we explored whether any of the DEA targets were known elements of PD-1 or CTLA-4 signaling cascades. The DEA for ipilimumab efficacy was enriched for self-antigens related to CTLA4 blockade [CD80, CD86, PIK3CA, and PTPN11; P=0.001). The list of DEA in nivolumab patients with severe toxicity was enriched for self-antigens related to PD1 blockade [HLA-A, HLA-DRB1, NFAT5, and STAT3; P=0.038]. AutoAb to LAG3 and CTLA4 were expressed at significantly higher levels in patients who received ipilimumab plus nivolumab and did not recur (P=0.013 and P=0.023, respectively). There was no significant differential expression of autoantibodies to either LAG3 or CTLA4 for the other treatment-outcome combinations. Similarly, there was no significant differential expression of autoantibodies to PD-1 or PD-L1 for any of the treatment-outcome combinations.
In view of the foregoing, it will be recognized that the present disclosure provides a composite panel of baseline serum autoantibodies that predict disease recurrence and severe toxicity in patients with resected stage III or IV melanoma receiving adjuvant nivolumab, ipilimumab, or combination ipilimumab plus nivolumab. For each treatment regimen, the autoAb signatures could be used together to accurately stratify patients based on their projected likelihood of suffering disease recurrence and developing severe irAEs. This is in contrast with most biomarkers of immunotherapy response under active investigation, which aim to predict either treatment efficacy or toxicity. The ability to simultaneously forecast both outcomes would enable providers and patients to view the possibility of clinical benefit in the context of potential toxicity, and ultimately help optimize treatment regimens while minimizing exposure to severe irAEs.
Prior to the present disclosure, there were no robust tools for identifying melanoma patients at risk for disease recurrence or immunotoxicity after treatment with immune checkpoint inhibitors (Patil et al., 2018; von Itzstein et al., 2020). Pretreatment tumor cell expression of PD-L1 was one of the first and most extensively studied candidate biomarkers of ICB response (Gibney et al., 2016; Topalian et al., 2012). Although some data suggest that higher PD-L1 expression is associated with better treatment outcomes, PD-L1 status has limited predictive utility and many patients classified as negative still respond to anti-PD-1 therapy (Daud et al., 2016). We found that autoAb panels outperformed a model composed of clinical covariates including PD-L1 status. Compared to the clinical model, the parsimonious autoAb signatures predicted recurrence more accurately and with demonstrated net clinical benefit for all three treatment regimens.
The autoAb signatures identified in this disclosure offer several other advantages including: (1) they can be combined into a single assay that simultaneously yields recurrence and severe toxicity predictions for multiple treatments; (2) they can reliably predict recurrence for therapies targeting a wide spectrum of immune checkpoints, which stands in contrast to percent PD-L1 expression, whose use is limited to anti-PD-1 therapies (Postow et al., 2015); and (3) from a technical standpoint, the autoAb readouts and the signature algorithms can be applied to each patient using their own internal control, which is the median overall signal intensity on that individual's panel. This obviates the need to use external controls or normalizing processes when testing serum from new patients.
Certain aspects of the disclosure involve using only the predictive power of antibodies expressed at higher levels in patients without recurrence or with severe toxicity. This is because in part the production and presence of autoAbs indicates subclinical immune system activity, which in turn foreshadows the immune system activation seen in patients who respond to treatment or develop severe immunotoxicity. For the same reason, in embodiments, the disclosure does not limit the signature building process to autoAbs whose expression exceeded a select number of standard deviations above that of the healthy controls, which is frequently done in autoimmune disease antibody research (de Moel, Derksen, et al., 2019; Pardos-Gea et al., 2017).
In the present disclosure, the datasets come from two separate Phase III, randomized, controlled studies so the serum collection, treatment schedules, and outcomes assessments were all standardized. Although the cohort of patients from CheckMate 238 was intentionally chosen to be enriched for irAEs, the patients and their clinical outcomes from both CheckMate 238 and CheckMate 915 are otherwise representative of the overall trial populations. The disclosure also demonstrates that the autoAb signatures consistently outperform a model that includes percent PD-L1 expression, which is the current clinical benchmark for identifying patients likely to respond to immunotherapy. Further, since both CheckMate 238 and CheckMate 915 had an arm of patients treated with nivolumab, the disclosure includes validation of the chosen cutoffs for the recurrence and toxicity signatures on a held out independent validation dataset composed of all the patients from CheckMate 915.
The following reference list is not an indication that any of the references are material to patentability:
This application claims priority to U.S. provisional patent application No. 63/277,279, filed Nov. 9, 2021, and U.S. provisional patent application No. 63/277,336, filed Nov. 9, 2021, the disclosures of each of which are incorporated herein by reference.
This invention was made with government support under R01 CA231295 awarded by the National Institutes of Health. The government has certain rights in the invention.
| Number | Date | Country | |
|---|---|---|---|
| 63277279 | Nov 2021 | US | |
| 63277336 | Nov 2021 | US |
