If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Precision Cancer Therapeutics Program, Center for Individualized Medicine, Mayo Clinic, Rochester, MinnesotaDepartment of Medical Oncology, Mayo Clinic, Phoenix, Arizona
Precision Cancer Therapeutics Program, Center for Individualized Medicine, Mayo Clinic, Rochester, MinnesotaDepartment of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, Minnesota
Corresponding author. Address for correspondence: Aaron S. Mansfield, MD, Division of Medical Oncology, Mayo Clinic, 200 First Street Southwest, Rochester, MN 55905.
Precision Cancer Therapeutics Program, Center for Individualized Medicine, Mayo Clinic, Rochester, MinnesotaDivision of Medical Oncology, Mayo Clinic, Rochester, Minnesota
The favorable outcomes with immunotherapy for mesothelioma were somewhat unexpected because this tumor has a low tumor mutation burden which has been associated with benefit in other cancers. Because chromosomal rearrangements are common in mesothelioma and have neoantigenic potential, we sought to determine whether they are associated with survival in patients treated with immunotherapy.
Methods
Pleural biopsies of mesothelioma after at least one line of therapy were obtained from patients (n = 44) before treatment with nivolumab alone (NCT29908324) or in combination with ipilimumab (NCT30660511). RNA and whole-genome sequencing were performed to identify the junctions resulting from chromosomal rearrangements and antigen processing and presentation gene set expression. Associations with overall survival (OS) were estimated using Cox models. An OS cutoff of 1.5 years was used to distinguish patients with and without durable benefit for use in receiving operating characteristic curves.
Results
Although tumor junction burdens were not predictive of OS, we identified significant interactions between the junction burdens and multiple antigen processing and presentation gene sets. The “regulation of antigen processing and presentation of peptide antigen” gene set revealed an interaction with tumor junction burden and was predictive of OS. This interaction also predicted 1.5-year or greater survival with an area under the receiving operating characteristic curve of 0.83. This interaction was not predictive of survival in a separate cohort of patients with mesothelioma who did not receive immune checkpoint inhibitors.
Conclusions
Analysis of structural variants and antigen presentation gene set expression may facilitate patient selection for immune checkpoint inhibitors.
Given the mixed results observed with immune checkpoint inhibitors for the treatment of mesothelioma, it is more important than ever to identify biomarkers that may predict outcomes and guide the use of these therapies. Unlike other tumor types with high tumor mutation burdens where clear survival benefits have been found with immune checkpoint inhibitors, mesothelioma has a very low mutation burden. Mesothelioma primarily arises as a result of the exposure to the carcinogen asbestos, although some cases develop after therapeutic radiation or are inherited owing to loss of function mutations in BAP1.
This finding was unexpected because other tumors associated with carcinogenic exposures, such as malignant melanoma and SCLC and NSCLC, typically have a high TMB from ultraviolet radiation and tobacco exposure, respectively.
High TMBs are thought to be a surrogate for an increase in neoantigens that can be recognized by the adaptive immune system and facilitate tumor elimination. Despite the reportedly low TMB in mesothelioma, the combination of the programmed cell death protein-1 (PD-1) inhibitor nivolumab and the CTLA-4 inhibitor ipilimumab was found to be superior to treatment with cisplatin and pemetrexed chemotherapy in patients with unresectable mesothelioma and is now approved by the U.S. Food and Drug Administration for frontline use.
Current clinically available NGS approaches do not fully characterize the genomic complexity of tumors. Cytogenetic studies have identified recurrent, structural chromosomal abnormalities in mesothelioma,
For this reason, in previous work, we used a sequencing approach that tiles the whole genome with large DNA fragments (2–5 kb compared with standard 200–500 base pairs) to improve the detection of structural variants, such as insertions, deletions, and translocations. Chromosomal rearrangements disrupt gene regions generating truncations or fusion transcript reading into normally distal gene regions or noncoding DNA. We previously found multiple chromosomal rearrangements that resulted in discordant DNA junctions with the potential for novel fusions in mesothelioma.
after previous treatment with platinum-based chemotherapy. DNA and RNA were purified using the AllPrep DNA, RNA, micro RNA universal kit (Qiagen, #80224) after the instructions provided by the manufacturer. The buffer included β-mercaptoethanol for the specimens obtained from NCT02497508 and dithiothreitol for the ones obtained from NCT03048474. Otherwise, there were no differences in the handling of the specimens or nucleic acid purification. The clinical trials and translational studies were approved by the local institutional ethics committees. Characteristics of the patients included in our analysis were compared with those of patients who were excluded owing to insufficient materials using the Fisher’s exact test for categorical variables and the Mann-Whitney U test for continuous variables. Survival between these groups was compared using the R packages “survival” and “survminer.”
Determination of Tumor Junction Burdens
Chromosomal rearrangements were reported by sequencing DNA prepared according to the mate-pair whole-genome library protocol (Nextera Library Prep Protocol). Sequencing results were mapped by BIMA, and the junctions of the chromosomal rearrangements were called by SVAtools. BIMA and SVAtools are Mayo Clinic in-house informatic pipelines.
The junctions of the chromosomal rearrangements were annotated with (1) the position of the junction with a resolution of 200 to 500 base pairs, (2) direction of the chromosomal rearrangement, and (3) genes at the junction using the National Center for Biotechnology Information RefSeq genes for GRCh38. The number of chromosomal rearrangements per sample was evaluated by counting the number of unique genes hit by all junctions in the sample. All specimens had 60× or greater bridged coverage for the detection of junctions, except one which had 40× bridged coverage. Chromosomal rearrangements may refer to insertions, deletions, translocations, and inversions. Junctions are the locations of the breaks of these chromosomal rearrangements. There may be one junction (deletion, insertion, translocation), two junctions (inversion, balanced translocations), or multiple junctions (three-way, four-way, etc., translocation) involved with each chromosomal rearrangement.
RNA-Sequencing Analyses
Mapping of the RNA-sequencing (RNA-seq) data and estimations of gene expression counts in each sample were performed by MAP-RSeq pipeline developed previously by the Mayo Bioinformatics Core.
Raw “count” files were processed by the “edgeR” package to generate log 2-normalized gene expression values.
Antigen Processing and Presentation
The Biological Processes Gene Ontology data set in the Molecular Signature database was searched for gene sets with names that included “antigen” and “presentation.” Of the 21 found hits, nine were eliminated for processes involving lipid, polysaccharide, and exogenous antigens or processes representing dendritic cell or T-cell antigen processing and presentation (APP). Single sample enrichment scores in the remaining 12 gene sets were calculated by using the “ssGSEA” (single sample gene set enrichment analysis) algorithm in the “GSVA” package.
Survival and Immune Checkpoint Inhibitor Survival Analyses
A statistical interaction is present when the association between two variables depends on a third variable. In our case, we hypothesized that the associations between tumor junction burden and survival (in terms of either hazard ratio (HR) or OR in Cox or logistic regression (LR) models, respectively) depended on the APP capabilities of tumors. Therefore, we tested the statistical significance of APP and tumor junction burden interactions in predicting overall survival (OS) or S1.5 yr. Associations of interactions between gene sets and log2-transformed junction burden (APP × log2[junction burden]) with OS were found by using the “coxph” (Cox proportional hazard) program in the “survival” package. Associations of these interactions with response to immune checkpoint inhibitors in terms of survival at 1.5 years (S1.5 yr) were calculated by LR using the “glm” (generalized linear model) package. APP and junction burden interactions were considered significant when either or both of the following conditions were met: (1) log-rank p values and the interaction terms in the OS models were significant (p < 0.05) or (2) the interaction terms in LR analysis were significant and the LR model had an accuracy on the basis of area under the curve (AUC) greater than 0.7. To create the Kaplan–Meier plot representing an individual gene set interaction with junction burden, samples were categorized as either “high” or “low” by using the median multiplication product of gene set scores and log2(junction burden) as the threshold. Reported p values in the plot are associations of the interaction and the model (log-rank test) with OS by “coxph” program.
Forest Plots
Median enrichment scores in each of the APP gene sets were used to group samples into high and low APP categories. In each category, HRs representing associations between junction burdens and OS were calculated by “coxph” and plotted using the “forestplot” package.
were evaluated for predicting significant benefit (SB) and no SB (NSB). Log2-transformed gene expression data were normalized in each row by subtracting average values across all samples according to the authors’ instructions. Normalized expression values were input to the python program “tidepy” to estimate individual tumor scores in 14 models. LR analyses were then used to estimate the accuracy of models with the CD8 model having been found as the best performer. Finally, “pROC” program was used to plot the ROC curves for TIDE, IFNG, programmed death-ligand 1 (PD-L1), and CD8 models.
Immune Deconvolution
The immunedeconv package in R was used to evaluate the tumor microenvironment (TME). It contains six approaches (quantiseq, timer, cibersort_abs (and first-generation cibersort), mcp_counter, xCell, and epic) to estimate the abundance scores of multiple cell types, including adaptive and innate immune cells, on the basis of ssGSEA data. Statistical significance of differential cell-type enrichment between cohorts of patients with high or low “REGULATION OF APP OF PEPTIDE ANTIGEN” gene set expression was compared using the t test.
Results
A total of 68 patients with pleural mesothelioma were treated with the PD-1 inhibitor nivolumab alone or in combination with the CTLA-4 inhibitor ipilimumab on the NivoMes (n = 34) and INITIATE (n = 34) clinical trials, respectively
(Supplementary Table 1). These patients had received at least one previous line of platinum-containing therapy. Biopsies were obtained on 65 of these patients just before the start of treatment with an immune checkpoint inhibitor(s), and 44 of these specimens had sufficient DNA and RNA content for analysis. There were no significant differences between the characteristics of the patients included in this analysis and those excluded on the basis of sample insufficiency, including sex, trial treatment, performance status, line of therapy, age, or OS. Despite the historic median survivals of less than 6 months with second or later line of therapy in mesothelioma,
there was a separation in OS at 1.5 years (S1.5 yr) from start of treatment on trial which we selected to group patients into categories of SB (>S1.5 yr) and NSB (≤S1.5 yr) (Fig. 1A). There were no differences in OS between those who receive nivolumab with or without ipilimumab (Supplementary Fig. 1). The biopsies obtained just before treatment were analyzed by mate-pair DNA sequencing and RNA-seq. There were many chromosomal rearrangements in each specimen (median = 130 junctions, range: 23–348), and a fraction of these involved unique genes (median = 18, range: 1–68). We selected the chromosomal rearrangements involving unique genes in each tumor for our analysis given their potential to be expressed and refer to them as the tumor junction burden from hereon.
Figure 1(A) Survival times of the study cohort. Red, blue, and green represent the best responses of PD, SD, and PR, respectively. Circles, triangles, and squares represent epithelioid, sarcomatoid (including mesenchymal), and mixed histologies, respectively. “+” designates alive at the last follow-up. (B) Heatmap representing survival times, junction burden, APP, and immune checkpoint markers. The lower bar represents best responses with PD, SD, and PR as per panel A. Orange arrows point to two cases with high junction burdens, short survival times, and low expression in genes involving APP. In contrary, green arrows point to two cases with moderate junction burdens, long survival times, and robust APP expression. APP, antigen processing and presentation; Epit, epithelioid; Mix, mixed; NSB, no significant benefit; PD, progression of disease; PD-1, programmed cell death protein-1; PD-L1, programmed death-ligand 1; PD-L2, programmed death-ligand 2; PR, partial response; Sar, sarcomatoid; SB, significant benefit; SD, stable disease.
Given our previous findings of the neoantigenic potential of chromosomal rearrangements, we sought to determine whether tumor junction burdens were associated with survival in patients with mesothelioma treated with immune checkpoint inhibitors. We did not find an association between tumor junction burden and OS (Cox model log-rank p > 0.5) (Supplementary Fig. 2A). Notably, two patients with the highest tumor junction burdens had very short survival times, whereas two other patients with moderate tumor junction burdens had a durable survival benefit (Fig. 1B). The two patients with the highest tumor junction burdens and poor survival had low expression of genes involved in APP. In contrast, patients with moderate tumor junction burdens and more durable survival had very robust expression of APP-associated genes.
Because the impact of tumor junction burdens seemed to be modulated by APP, we hypothesized that the neoantigenic potential of chromosomal rearrangements was dependent on the capability of cancer cells to present neoantigens to the immune system. To evaluate whether there was an interaction between APP gene sets and tumor junction burdens that affected outcomes, we selected 12 APP gene sets from the Gene Ontology-Biological Processes data set in the Molecular Signature Database and calculated their enrichment scores (Supplementary Table 2). We then used these scores to test for interactions between APP gene sets and junction burdens on survival and found significant interactions with six APP gene sets (Table 1). With these six APP gene sets, the HRs representing associations between tumor junction burdens and OS favored patients with high APP scores (all HRs < 1) more so than patients with low APP scores (all HRs > 1) (Fig. 2). There were no differences in survival between patients with high or low APP scores (Supplementary Fig. 2B). In patients with low APP scores, those with a high tumor junction burden were at increased risk of death compared with patients with low tumor junction burdens (Supplementary Fig. 2C). In contrast, in patients with high APP scores, those with high tumor junction burdens were at reduced risk of death compared with patients with low tumor junction burdens (Supplementary Fig. 2D).
Table 1Antigen Processing and Presentation Gene Set Analysis
Note: The statistical significance of interactions (P-IA-Cox) and log-rank (p-Log-Rank) in Cox models, interactions (P-IA-lr), and AUC in LR models are listed. The gene sets with significant interactions are in bold. APP and junction burden interactions were considered significant when either or both of the following conditions were met: (1) log-rank p values and the interaction terms in the OS models were significant (p < 0.05) or (2) the interaction terms in LR analysis were significant and the LR model had an accuracy on the basis of AUC greater than 0.7.
APP, antigen processing and presentation; AUC, area under the curve; LR, logistic regression; OS, overall survival.
Figure 2Forest plots displaying the hazard ratios for junction burdens and overall survival associations in samples with high and low APP gene set expression, respectively, in gene sets identified as significant. APP, antigen processing and presentation; HR, hazard ratio.
We further evaluated the interaction models that included the “REGULATION OF APP OF PEPTIDE ANTIGEN” gene set which included six genes (PYCARD, HFE, HLA-DOA, HLA-DOB, TREM2, and TAPBPL). Both the interaction parameter between this gene set and the tumor junction burden and the survival model were highly significant (Table 1 and Fig. 3A). Furthermore, this interaction was highly predictive of S1.5 yr with an AUC of 0.831 (Fig. 3B). For comparison, we tested several available gene models previously reported to associate with response to immune checkpoint inhibitors, including TIDE, IFNG, PD-L1, and CD8.
In our cohort, none of these other models performed, including the interaction of APP gene sets with tumor junction burdens in predicting S1.5 yr, but the most accurate of these gene models was the CD8 model with an AUC of 0.683 (Fig. 3C). On the basis of this observation, and to account for the role of antitumor lymphocytes in survival with immune checkpoint inhibition, we included CD8A in our prediction model. This addition increased the accuracy of the model from 0.831 to 0.890 (Fig. 3D).
Figure 3(A) The Kaplan–Meier curve representing a survival model on the basis of the interactions between “REGULATION OF APP OF PEPTIDE ANTIGEN” gene set and junction burdens is illustrated. Both the interaction terms and the log-rank test were significant. (B) The ROC curve representing APP and log2[junction burden] interactions (cyan) in predicting NSB and SB is illustrated for the REGULATION_OF_AP&P_OF_PEPTIDE_ANTIGEN gene set. (C) ROC curves representing the accuracy of TIDE (blue), IFNG (green), PD-L1 (dark red), and CD8 (orange) models in predicting NSB and SB. (D) ROC curves representing APP (purple), log2[junction burden] (lime green), CD8A (light salmon), and the final model including APP/log2[junction burden] interactions and CD8A (magenta). The inlet is a boxplot and individual patient prediction values by the final model in NSB and SB categories. Colors represent radiologic responses as defined in Figure 1. APP, antigen processing and presentation; AUC, area under the curve; Hi, high; Jn, junction burden; Lo, low; NSB, no significant benefit; PD-L1, programmed death-ligand 1; ROC, receiving operating characteristic; SB, significant benefit.
We sought to evaluate whether the interaction models were predictive of patient OS irrespective of treatment approach. To the best of our knowledge, the only available mesothelioma data set that includes both chromosomal rearrangements from whole-genome sequencing, and RNA-seq, is from our previous study of patients (n = 24) who provided biopsy or surgical specimens before any cytotoxic systemic therapy (Mayo_2019 cohort).
The patients in the Mayo_2019 cohort did not receive immune checkpoint inhibitors as these therapies were not available during their lifetimes. There was a break in OS at 1.5 years from diagnosis in this cohort that was used as the threshold for categorizing patients as NSB and SB (Supplementary Fig. 3). The Mayo_2019 cohort performed similar to other historic mesothelioma cohorts as a previously established mesothelioma survival signature gene set
Global gene expression profiling of pleural mesotheliomas: overexpression of aurora kinases and P16/CDKN2A deletion as prognostic factors and critical evaluation of microarray-based prognostic prediction.
had very high prognostic significance for both OS and S1.5 yr (Supplementary Fig. 4A and B). We did not find an interaction between the tumor junction burdens and any of the 12 APP gene sets on OS to be statistically significant (Supplementary Table 3). In further analysis, we noted that the tumors in the Mayo_2019 cohort had fewer junctions than the current cohort (Supplementary Fig. 5) which may have affected the predictive values of the interaction models.
Finally, we used RNA-seq for computational immune deconvolution to compare the TME in mesotheliomas with low and high expression of the “REGULATION OF APP OF PEPTIDE ANTIGEN” gene set. The “immunedeconv” package used for our analyses provides results from six different computational approaches (see the Methods section). In all approaches, we observed a lower concentration of immune cells suggesting a “cold” TME in tumors with low compared with high APP gene set expression (Supplementary Figs. 6–9). We found higher TME and immune scores (by xCell) and cytotoxicity score (by MCP-counter) and an enrichment of lymphocytes that are often associated with antitumor immunity, such as B, T, and natural killer cells and M1 macrophages in tumors with high APP.
Discussion
Genomic structural variants are common in mesothelioma. In this analysis, the tumor junction burdens resulting from chromosomal rearrangements were associated with improved survival outcomes in patients treated with immune checkpoint inhibitors in the presence of APP gene set expression. In contrast, tumor junction burdens in the absence of APP gene set expression were associated with reduced survival despite treatment with immune checkpoint inhibitors. Our model was further improved by the inclusion of CD8A, a marker of cytotoxic lymphocytes. We interpreted these observations to be consistent with our understanding of the mechanisms of adaptive antitumor immunity where antigen-specific T-cell responses that are restored or generated by PD-1 and CTLA-4 inhibition require tumor cell presentation of neoantigens. Because the interaction signature between the tumor junction burdens and APP gene sets did not favorably affect OS in a separate cohort of patients who did not receive immune checkpoint inhibitors, this signature is not likely to be predictive in settings outside of treatment with immunotherapy. Chromothripsis represents a complex pattern of multiple chromosomal rearrangements typically on a single chromosome. We previously identified that higher numbers of chromothripsis-like patterns detected from copy number segmentation data were a negative prognostic factor in mesothelioma,
Despite the negative prognostic significance that has been attributed to increases in these complex patterns of chromosomal rearrangements in mesothelioma and other tumors, tumor junction burdens were associated with improved survival in the context of APP gene set expression in this cohort of patients with mesothelioma treated with immune checkpoint inhibitors.
Given the marked differences between the TME in tumors with and without high APP gene set expression, we speculate that methods to manipulate the TME might be beneficial for these patients. Recently, it was found that low-dose radiotherapy in murine models promotes T-cell infiltration, enabling response to combination immunotherapy.
A clinical trial has recently activated to test this approach in mesothelioma (NCT04926948). Other work has suggested that oncolytic virotherapy may reprogram the TME to enable responses to immune checkpoint inhibitors.
It is a major initiative across tumor types to identify means of converting tumors to be responsive to immune checkpoint inhibitors.
There have been inconsistent results with the use of immune checkpoint inhibitors for the treatment of mesothelioma. On the basis of the CheckMate 743 trial, the frontline use of ipilimumab and nivolumab clearly benefits patients with nonepithelioid mesothelioma, partially because chemotherapy is so ineffective for this group.
The same degree of benefit was not observed in the epithelioid group, as chemotherapy is more effective for patients with that variant of disease. Because the survival analysis of all randomized patients was positive in the CheckMate 743 trial with a stratified HR of 0.74 (96.6% confidence interval: 0.60–0.91; p = 0.0020), ipilimumab and nivolumab were approved by the U.S. Food and Drug Administration for frontline treatment of unresectable pleural mesothelioma regardless of histologic subtype. In second or later lines of treatment, single-agent PD-1 inhibitors have been found to be superior to placebo in the CONFIRM
A multicentre randomised phase III trial comparing pembrolizumab versus single-agent chemotherapy for advanced pre-treated malignant pleural mesothelioma: the European Thoracic Oncology Platform (ETOP 9–15) PROMISE-meso trial.
; however, these studies both reported that there are responses with immune checkpoint inhibitors in patients with epithelioid disease. Surprisingly, the overall response rate with the PD-1 inhibitor pembrolizumab was higher than that observed with chemotherapy (22% versus 6%) in the PROMISE-meso trial, although this difference did not translate into a survival benefit. PD-L1 expression was not able to discriminate benefit in the CONFIRM
A multicentre randomised phase III trial comparing pembrolizumab versus single-agent chemotherapy for advanced pre-treated malignant pleural mesothelioma: the European Thoracic Oncology Platform (ETOP 9–15) PROMISE-meso trial.
or in our cohort. Given the discrepancies with survival outcomes between these clinical trials, it is critical to develop better predictive biomarkers, especially for patients with epithelioid disease where benefit with immune checkpoint inhibitors is less certain.
There have been multiple efforts to identify predictors of benefit with immune checkpoint inhibitors.
TMB has also been proposed as a surrogate of neoantigens that can be recognized by the adaptive immune system for elimination. Recently, a PD-1 inhibitor has been approved for solid tumors with a TMB greater than or equal to 10 mutations per megabase
Association of tumour mutational burden with outcomes in patients with advanced solid tumours treated with pembrolizumab: prospective biomarker analysis of the multicohort, open-label, phase 2 KEYNOTE-158 study.
There is significant heterogeneity in the approaches used to determine TMB, and use of population germline variant databases to filter calls can inflate scores and introduce racial bias.
TMBs frequently do not evaluate or include structural variants or junction burdens. In addition, TMB fails to incorporate the full complexity of an adaptive, antitumor immune response.
Immunograms may provide better predictors of response to immune checkpoint inhibitors as these would incorporate tumor foreignness (using comprehensive mutation burdens), the ability of tumors to present neoantigens with MHC proteins (APP), lymphocytes, and their ability to traffic to tumors, and the expression of immune checkpoints and other regulatory signals.
Our findings represent one step toward adopting an immunogram to predict survival with immune checkpoint inhibitors in mesothelioma by incorporating APP gene set expression in our analysis. These results also suggest that genomic approaches that identify and incorporate junction burdens can improve the determination of TMB, especially in tumors such as mesothelioma that have relatively few single-nucleotide mutations.
We tested the tumors of patients who had received previous platinum-based chemotherapy. Because the numbers of junctions were slightly higher in the current cohort than a separate cohort of patients who had not received previous platinum-based chemotherapy, it is possible that cytotoxic therapy introduced structural variants. Along these lines, our findings will need to be validated in a cohort of treatment-naive patients. In addition, given the DNA sample requirements to perform our analysis of structural variants, we did not have sufficient materials to perform traditional sequencing approaches to evaluate single-nucleotide mutations. Given the reportedly low TMB in mesothelioma and our previous findings of large, complex rearrangements in this malignancy, we felt that it was reasonable to focus our efforts on these structural variants. Finally, efforts are underway to develop chemoimmunotherapy regimens for mesothelioma. We are not certain whether structural variants would retain their association with survival outcomes in the setting of combination cytotoxic and immunotherapy.
In conclusion, in the context of APP gene set expression, tumor junction burdens were associated with improved survival in patients with mesothelioma treated with immune checkpoint inhibitors. In contrast, in the absence of APP, tumor junction burdens were associated with poor survival. The inclusion of genomic approaches that can detect structural variants, and transcriptomics to evaluate APP, may help refine the selection of patients to receive immune checkpoint inhibitors, especially for patients with mesothelioma.
Maria Disselhorst: Conceptualization, Investigation, Resources, Writing—original draft, Writing—review and editing.
Jun Yin: Methodology, Formal analysis, Investigation, Writing—review and editing.
Tobias Peikert, John Cheville: Conceptualization, Writing—review and editing.
Julia Udell: Formal analysis, Investigation, Writing—review and editing.
Sarah Johnson, James Smadbeck, Stephen Murphy: Investigation, Data curation, Writing—review and editing.
Alexa McCune, Giannoula Karagouga, Aakash Desai: Investigation, Writing—review and editing.
Janet Schaefer-Klein: Supervision, Funding acquisition, Writing—review and editing, Project administration.
Mitesh J. Borad: Supervision, Funding acquisition, Writing—review and editing.
George Vasmatzis: Conceptualization, Supervision, Writing—review and editing.
Paul Baas: Conceptualization, Resources, Supervision, Writing—review and editing.
Aaron S. Mansfield: Conceptualization, Methodology, Formal analysis, Investigation, Writing—original draft, Writing—review and editing, Visualization, Funding acquisition.
Acknowledgments
This work was financially supported by Mark Foundation ASPIRE Award, NCI R21 CA251923, the Barry Family, and Mayo Clinic’s Center for Individualized Medicine.
Global gene expression profiling of pleural mesotheliomas: overexpression of aurora kinases and P16/CDKN2A deletion as prognostic factors and critical evaluation of microarray-based prognostic prediction.
A multicentre randomised phase III trial comparing pembrolizumab versus single-agent chemotherapy for advanced pre-treated malignant pleural mesothelioma: the European Thoracic Oncology Platform (ETOP 9–15) PROMISE-meso trial.
Association of tumour mutational burden with outcomes in patients with advanced solid tumours treated with pembrolizumab: prospective biomarker analysis of the multicohort, open-label, phase 2 KEYNOTE-158 study.
Drs. Kosari and Disselhorst contributed equally as co-primary authors.
Disclosure: Dr. Disselhorst reports receiving grants to her institution from and having an advisory role for Bristol-Myers Squibb. Dr. Peikert reports having advisory and consultant work for AstraZeneca with all fees being paid to Mayo Clinic. Dr. Baas reports having advisory and consultancy work for Bristol-Myers Squibb, Merck, Pfizer, Beigene, and Trizell and receiving research grant from Bristol-Myers Squibb and Merck (all payments to the hospital). Dr. Mansfield reports receiving research support from Novartis and Verily; receiving remuneration to his institution for participation on advisory boards for AbbVie, AstraZeneca, Bristol-Myers Squibb, Genentech, and Janssen; receiving travel support and payment from Shanghai Roche Pharmaceuticals Ltd.; and serving as a nonremunerated director of the Mesothelioma Applied Research Foundation. A patent has been filed on the basis of these findings. The remaining authors declare no conflict of interest.
03319-0&id=gr1.jpg){kind=link}
03319-0&id=gr2.jpg){kind=link}
03319-0&id=gr3.jpg){kind=link}