Longitudinal Molecular Analysis of Tumor Exome and Transcriptome to Evaluate Clonal Evolution and Identify Novel Therapeutic Targets in Thymoma

The rarity of thymic epithelial tumors and limited preclinical modeling constrain our understanding of disease biology and the advancement of rational treatment strategies. Tumor mutational profiling has demonstrated sporadic activating mutations in targetable tyrosine kinases that have not been successfully targeted outside of individual case reports.¹ Additionally, most molecular studies to date have focused on profiling treatment-naïve tumors, failing to capture potential changes that accompany recurrence and resistance to therapy.² To profile evolutionary changes through recurrence and identify novel treatment options in advanced thymoma, we present the molecular analysis of exome and bulk RNA sequences from a primary thymic tumor and multiple resected metachronous tumors from a single patient.

The patient initially presented because of bulbar symptoms of myasthenia gravis at age 25 years. Computed tomography imaging at that time demonstrated an anterior mediastinal mass without any pleural deposits, pulmonary nodules, or other findings concerning for distant metastatic disease. Median sternotomy, radical thymectomy, and partial pericardiectomy were performed. Pathologic analysis demonstrated a mixed B1/B2/B3 thymoma with two margins that were microscopically positive for disease. After R1 resection, the patient received adjuvant mediastinal radiation to reduce the risk of disease recurrence. Five years later, the patient underwent left-sided thoracotomy, total parietal pleurectomy, and diaphragmatic tumor resection with diaphragm reconstruction for recurrence in the pleural space, persistently progressive over 2 years. WHO subtype B3 histology was predominant, with B2 foci present. Six years later, the patient had another recurrence in the pleural space observed over 18 months, managed with redo thoracotomy, lysis of adhesions, and resection of implants on parietal pleura, pericardium, and diaphragm, with left lower lobe lung wedge resection for visceral implants. Histologic subtype was unchanged. One year later, the patient experienced another left-sided recurrence in the pleural space along with lateral diaphragm lesion with symptomatic chest wall involvement necessitating palliative intervention. Debulking was attempted with two cycles of CAP (cisplatin 50 mg/m², doxorubicin 50 mg/m², cyclophosphamide 500 mg/m² once every 3 weeks), but restaging scans demonstrated no radiographic response and the patient was taken for a repeat thoracotomy with pleural metastasis resection, parietal diaphragm resection, and primary diaphragm repair, along with retroperitoneal implant resection, again with unchanged tumor subtype. In total, the patient underwent a total of four surgeries for tumor debulking yielding the primary tumor and metastases from three metachronous recurrences, all outside the adjuvant radiation field (Fig 1A). For all recurrences, surgical exploration demonstrated additional subcentimeter deposits on visceral and parietal pleura below the detection limits of computed tomography imaging.

FIG 1.

Somatic mutational alterations associated with presentation and sequential metastatic recurrences. (A) Timeline of clinical history. (B) Plot of somatic point mutations with subsequent samples. Shared mutations are outlined in red. (C) CNAs for metastatic samples. (D) Heatmap with CNA concordance of subsequent samples. (E) Venn diagram with intersecting somatic mutations of all surgical samples. CAP, cisplatin, doxorubicin, cyclophosphamide; CNA, copy number alteration; cnLOH, copy number loss of heterozygosity.

This study was performed in accordance with the precepts of the Helsinki Declaration. The patient provided informed consent to a research protocol (13-416) previously approved by the Dana-Farber/Harvard Cancer Center Institutional Review Board, with Massachusetts General Hospital as the study site. The study was specifically performed after the patient provided informed consent. All protected health information has been removed to protect the patient's identity. The patient also provided informed consent for publication of information resulting from this study.

To evaluate for somatic mutational changes at baseline and through subsequent recurrences, we performed whole-exome and bulk RNA transcriptome sequencing on formalin-fixed, paraffin-embedded tumor samples (samples 1, 2, and 3) and a fresh-frozen tumor sample (sample 4) using the BostonGene Tumor Portrait test. Anatomic site of origin for each analyzed metastasis is listed in Figure 1A. Peripheral blood mononuclear cells were used as a germline control for DNA exome sequencing. With respect to point mutations and short insertions/deletions, only a limited number of alterations were identified and shared between tumor samples (Fig 1B). All variants were of unknown significance, consistent with passenger events representative of subclonality. No driver alterations in GTF2I, TP53, HRAS, or NRAS were noted. The majority of mutations were missense in nature and exclusively confined to just one tumor sample. A limited number of missense mutations were shared between two tumor specimens, but no mutation was repeatedly seen in more than two samples: BDKRB2 p.R14L (11% variant allele fraction [VAF] in sample 2, 7% VAF in sample 3); OR4A15 pI124V (8% VAF in sample 2, 18.9% sample 4); SLC16A4 p.Q360E (12% VAF in sample 2, 10% VAF in sample 4); HELLS p.S304G (10% in sample 1, 8.3% in sample 3); NONGNB p.A186D (7.8% in sample 1, 7.7% in sample 3); and DYNC1I1 p.T544A (10% in sample 1 and 9.5% in sample 3). By contrast, large copy-number alterations (CNAs) were more recurrently observed across multiple samples (Fig 1C). Unfortunately, tumor purity of 20% limited the calling of any copy-number changes or loss of heterozygosity in the primary tumor sample. Instead, numerous copy-number changes were called throughout all metastatic specimens including 1q amplification, 6 loss, 7 gain, 9q gain, and 14 gain. The consistent representation of these findings in all metastatic recurrences suggests clonal founder effects, which is consistent with recurrent reporting of these alterations in previous studies.^2,3 These findings include gain of 7p11, where the EGFR locus is found, a feature commonly seen in aggressive types of thymoma. No fusion events were detected, including a KMT2A-MAML2 rearrangement. Heatmap concordance on the basis of copy-number changes demonstrated strong concordance between all three metastatic recurrences (Fig 1D). For single nucleotide polymorphisms and indels, there was limited overlap between identified mutational changes (Fig 1E), with no alteration seen in more than two samples as mentioned above. These findings suggest that CNAs constitute the main driver of mutational events in this patient's thymoma. As sequencing of the primary tumor could not support copy-number calls, we cannot comment on whether there are differences between primary tumor and metastatic recurrences. Noteworthy confounding factors that may explain the variance seen in primary tumor versus metastasis include limited tumor purity, sampling bias within a large heterogeneous primary tumor, and contribution of DNA from associated maturational lymphocytes.

With respect to the bulk RNA transcriptome, principal component analysis and uniform manifold approximation and projection showed that the patient's tumor samples generally clustered together with transcriptomes available through the The Cancer Genome Atlas (TCGA) cohort and did not demonstrate a unique subclustering on a transcriptional basis (Fig 2A). To further approximate the composition of immune cells, we deconvoluted immune cell subsets using the Kassandra algorithm⁴ (Fig 2B). Naïve CD8+ T-cell populations were the most predominant throughout all samples (ranging from 65.45% in initial to 45.02% in third metastatic sample), followed by naïve helper T cells. No major changes in population representation were seen throughout progression, with the slight exception of increased CD8+ central memory cells in the third metastatic recurrence (9.2%, up from 2.8% in initial sample). The cell type percentages were compared with respect to the thymic epithelial tumor cohort from TCGA² (Fig 2B). The enrichment for naïve CD8+ signatures was largely shared by AB, B1, and B2 subtypes, possibly a reflection of associated immature lymphocytes. Of the cell populations predicted by Kassandra, four were associated with statistically significant overall survival differences by Kaplan-Meier analysis (Fig 2C), notably with increased CD8+ enrichment associated with improved overall survival.

FIG 2.

The transcriptomic inference of immune cell composition of tumor and metastatic recurrences. (A) PCA and UMAP representations of patients and TCGA expressions colored by histology. (B) Cellular composition of the tumor microenvironment of the samples, represented as the percentage of different cell types in tumor microenvironment according to the deconvolution algorithm. (C) The Kaplan-Meier curve with the OS and PFS correlated with the particular cell type in the TCGA cohort (N = 105, thymic carcinoma excluded). CDC, conventional dendritic cell; OS, overall survival; PCA, principal component analysis; PDC, plasmacytoid dendritic cell; PFS, progression-free survival; TCGA, The Cancer Genome Atlas; UMAP, uniform manifold approximation and projection.

We also examined the expression of individual genes of interest at the RNA level, with reference to normalized expression of the TCGA cohort (Fig 3A). PBX1 expression by immunohistochemistry (IHC) has previously been demonstrated as a negative prognostic factor associated with decreased overall survival,⁵ potentially reflective of lineage-specific changes associated with tumor progression. PBX1 expression increased from the primary tumor sample to the metastatic recurrent tumor samples, with PBX1 transcripts at the 99th percentile of expression in the third metastatic recurrence. Given the persistent gains of chromosome seven throughout metastatic recurrence, we evaluated expression of the EGFR (7p11) and MET (7q31) oncogenes. Expression of both oncogenes progressively increased from the primary tumor through metastatic recurrence. By the third metastatic occurrence, the RNA level of EGFR and MET was at the 98th and 94th percentiles of the TCGA cohort, respectively. Within the TCGA cohort, EGFR expression is increased in B3 thymomas compared with B1/B2 thymoma and thymic carcinoma, although the same is not true for MET (Fig 3B). Fluorescence in situ hybridization confirmed low-level copy-number gain of the 7p11 locus (+1 gain) (data not shown), and IHC for EGFR demonstrated strong membranous expression throughout each metastatic recurrence sample (Fig 3C). By contrast, there was weak membranous staining of MET. Other notable RNA findings on potential therapeutic biomarkers include absent SSTR2/SSTR5 expression throughout all samples, absent TGFB1, and present VEGFA expression.

FIG 3.

Identification of EGFR as potential therapeutic target from transcriptomic analysis. (A) RNA expression of EGFR and other candidate biomarkers relative value compared with TCGA cohort. (B) Violin plot demonstrating TPM transcript reads on the basis of histology of respective WHO tumor types. Statistical testing performed by Mann-Whitney U test with reference to B3 cohort. (C) Hematoxylin and eosin and immunohistochemistry staining of select biomarkers. For each stain, 200× and 600× views are shown. TCGA, The Cancer Genome Atlas; TPM, transcripts per million.

For patients with advanced thymoma, there is an unmet need for novel treatment options. Standard options for patients include platinum-based chemotherapy regimens, typically in combination with taxane, anthracycline, or etoposide. Although patients frequently respond, remissions are often short-lived. Numerous efforts have been made at identifying targeted therapies; although sporadic activating point mutations have been identified in both EGFR and KIT in thymic epithelial tumors, these mutations have not been found recurrently and systemic efforts to target these oncogenes with selective small molecule inhibitors have not been successful.¹ Beyond activating point mutations, both of these oncogenes are variably expressed in thymic tumors at the protein level. In the setting of thymic carcinoma where there is frequently strong KIT expression, the multispecific kinase inhibitors sunitinib and lenvatinib have clinical efficacy.^6,7 For sunitinib, only minimal efficacy was seen in thymoma where there is markedly less KIT expression.⁶ Relevant to the 7p gain and EGFR expression seen in our patient, EGFR expression in thymoma has been previously reported.^8,9 On the basis of EGFR expression, the EGFR-targeted monoclonal antibody cetuximab has been evaluated in a limited number of patients with thymoma. In a case series of platinum pretreated patients with autoimmune disorders, five of seven experienced partial responses, although a full reporting of progression-free survival and overall survival was not reported.¹⁰ This contrasts with a different study of cetuximab, studied in combination with CAP chemotherapy in the neoadjuvant setting with the primary end point of pathologic complete response rate in treatment-naïve patients.¹¹ There, the primary end point was not met and cetuximab was not deemed to be effective. In our present case, EGFR expression increased from initial presentation through subsequent recurrences. These findings suggest evolutionary changes at the RNA and protein level not captured at the DNA level, although we cannot rule out bias introduced by limited sampling of an otherwise heterogeneous, mixed B-type primary tumor. It is possible that diminished expression in primary tumors and increased expression in metastatic recurrence may be seen in other patients, potentially consistent with the mixed results of cetuximab on the basis of previous treatment settings. These results argue for further study of cetuximab in patients with advanced disease. Also of interest was the increasing expression of MET throughout subsequent recurrences. Both EGFR and MET are targeted by the bispecific antibody amivantamab, which is currently US Food and Drug Administration approved for EGFR-mutant lung adenocarcinoma for tumors with exon 20 insertions.¹² Amivantamab remains under investigation for other EGFR and MET indications, and the co-occurring expression of MET in this case raises the question of whether this finding may be more generalizable and a basis for multispecific targeted therapies, although MET expression was diminished relative to EGFR.

Beyond the immediate therapeutic implications, this analysis provides additional insights into the clonal evolution of thymic epithelial tumors. To date, few studies have evaluated clonal evolution within an individual patient. As part of the rapid autopsy program at the National Institutes of Health, analysis of a patient with thymic carcinoma demonstrated few conserved single-nucleotide polymorphisms and insertions/deletions, but rather clonal and subclonal changes with respect to CNAs.¹³ Although that case examined metastases evaluated at a synchronous time point, our case evaluates a primary tumor and three metachronous metastatic recurrences. Similarly, we identified CNAs as the conserved finding across patient tumors, suggestive of foundational clonal effects for the patient's metastatic disease. Unfortunately, tumor purity limited the ability to call CNAs in the primary tumor. This is a recurrent challenge in B1 and B2 subtypes because of contaminating nucleic acids from immature lymphocytes. Although purity can be enriched through techniques such as microdissection of B3 components, this fails to capture the mutational processes in remaining B1/B2 components. Fluorescence in situ hybridization may add value to future studies as a method less affected by tumor purity. However, this method carries its own set of limitations, namely limited detection for shallow copy-number gains.

CNAs have been described recurrently in B-type thymoma and thymic carcinoma,^1-3 consistent with a conserved role in tumorigenesis for both of these thymic epithelial tumors. The promise offered by the notion of targeted therapies is that recurrent somatic mutations can serve as the basis for a therapeutic intervention. The current druggable space of targeted therapies is confined to activating point mutations in tyrosine kinases where genotypic changes have a strong correlation with phenotype through activation of oncogenic signaling. Recent description of a focused metastatic cohort (N = 49) identified sporadic point mutations in thymoma, most in TP53 (31%).¹⁴ Another recent study of 90 thymomas profiled by NGS described infrequent point mutations among the genomic alterations associated with B3 and non-B3 thymoma.¹⁵ The findings of our case are consistent with these two studies and highlight the importance of attention to CNAs as a potentially foundational aspect of tumor biology. In oncology, the molecular adaptations associated with metastatic recurrence and progression have garnered increased attention for therapeutic implications. For tumors such as thymoma where the types of recurrent somatic alterations exist outside of the currently targetable space, it is critical that mutational testing captures the pertinent molecular alterations. Without the ability to detect and track relevant CNAs, the field remains limited in clarifying biologic, prognostic, and therapeutic implications. Improved annotation of the molecular events that accompany tumor evolution will inform the requisite preclinical testing needed to define dependencies for tumor growth and survival, advancement of new targets for clinical investigation, and improvement in patient outcomes.

AUTHOR CONTRIBUTIONS

Conception and design: Christopher S. Nabel, Alexander Bagaev, Abner Louissaint Jr

Financial support: Abner Louissaint Jr

Administrative support: Alexander Bagaev

Provision of study materials or patients: Cameron D. Wright, Abner Louissaint Jr

Collection and assembly of data: Christopher S. Nabel, Yin P. Hung, Aleksandra Lopareva, Dora Dias-Santagata, Nikita Batashkov

Data analysis and interpretation: Christopher S. Nabel, Anna Kurilovich, Aleksandra Lopareva, Dora Dias-Santagata, Nikita Batashkov, Dmitry Tabakov, Maria Sorokina, Aleksey Makarov, Georgy Sagaradze, Anna Butusova, Olga Kudryashova, Lev Bedniagin, Cameron D. Wright, Nara Shin, Ekaterina Postovalova

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).