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
Department of Medicine, Vanderbilt University Medical Center, Nashville, TennesseeVanderbilt-Ingram Cancer Center, Vanderbilt University Medical Center, Nashville, Tennessee
Department of Medicine, Vanderbilt University Medical Center, Nashville, TennesseeVanderbilt-Ingram Cancer Center, Vanderbilt University Medical Center, Nashville, Tennessee
Corresponding author. Address for correspondence: Christine M. Lovly, MD, PhD, Vanderbilt-Ingram Cancer Center, 2220 Pierce Ave., 777 Preston Research Building, Nashville, TN 37232-6307.
Department of Medicine, Vanderbilt University Medical Center, Nashville, TennesseeDepartment of Cancer Biology, Vanderbilt University Medical Center, Nashville, TennesseeVanderbilt-Ingram Cancer Center, Vanderbilt University Medical Center, Nashville, Tennessee
Patients with SCLC have a poor prognosis and limited treatment options. Because access to longitudinal tumor samples is very limited in patients with this disease, we chose to focus our studies on the characterization of plasma cell-free DNA (cfDNA) for rapid, noninvasive monitoring of disease burden.
Methods
We developed a liquid biopsy assay that quantifies somatic variants in cfDNA. The assay detects single nucleotide variants, copy number alterations, and insertions or deletions in 14 genes that are frequently mutated in SCLC, including tumor protein p53 gene (TP53), retinoblastoma 1 gene (RB1), BRAF, KIT proto-oncogene receptor tyrosine kinase gene (KIT), notch 1 gene (NOTCH1), notch 2 gene (NOTCH2), notch 3 gene (NOTCH3), notch 4 gene (NOTCH4), phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha gene (PIK3CA), phosphatase and tensin homolog gene (PTEN), fibroblast growth factor receptor 1 gene (FGFR1), v-myc avian myelocytomatosis viral oncogene homolog gene (MYC), v-myc avian myelocytomatosis viral oncogene lung carcinoma derived homolog gene (MYCL1), and v-myc avian myelocytomatosis viral oncogene neuroblastoma derived homolog gene (MYCN).
Results
Over the course of 26 months of peripheral blood collection, we examined 140 plasma samples from 27 patients. We detected disease-associated mutations in 85% of patient samples with mutant allele frequencies ranging from 0.1% to 87%. In our cohort, 59% of the patients had extensive-stage disease, and the most common mutations occurred in TP53 (70%) and RB1 (52%). In addition to mutations in TP53 and RB1, we detected alterations in 10 additional genes in our patient population (PTEN, NOTCH1, NOTCH2, NOTCH3, NOTCH4, MYC, MYCL1, PIK3CA, KIT, and BRAF). The observed allele frequencies and copy number alterations tracked closely with treatment responses. Notably, in several cases analysis of cfDNA provided evidence of disease relapse before conventional imaging.
Conclusions
These results suggest that liquid biopsies are readily applicable in patients with SCLC and can potentially provide improved monitoring of disease burden, depth of response to treatment, and timely warning of disease relapse in patients with this disease.
SCLC is an aggressive lung cancer of neuroendocrine origin, with a propensity for early and extensive metastatic dissemination. SCLC accounts for approximately 15% of all lung cancers and approximately 30,000 deaths in the United States annually.
The median overall survival (OS) for patients with SCLC is approximately 8 to 12 months for patients with extensive-stage (ES) disease and 12 to 20 months for patients with limited-stage (LS) disease.
The combination of elusive pathophysiology, poor prognosis, and a lack of therapeutic improvement for several decades has led the National Cancer Institute to designate SCLC a recalcitrant cancer.
Biallelic inactivation of the tumor suppressors tumor protein p53 gene (TP53) and retinoblastoma 1 gene (RB1) are detected in most SCLC tumors. In addition to these molecular “hallmarks,” other genomic variants have been detected at lower frequencies, including the following: amplification of v-myc avian myelocytomatosis viral oncogene homolog gene (MYC), v-myc avian myelocytomatosis viral oncogene lung carcinoma derived homolog gene (MYCL1), v-myc avian myelocytomatosis viral oncogene neuroblastoma derived homolog gene (MYCN), and fibroblast growth factor receptor 1 gene (FGFR); phosphatase and tensin homolog gene (PTEN) loss; and activating mutations in phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha gene (PIK3CA), BRAF and the four NOTCH gene paralogs (notch 1 gene [NOTCH1], notch 2 gene [NOTCH2], notch 3 gene [NOTCH3], and notch 4 gene [NOTCH4]).
Although many of these genomic alterations could be viewed as potential therapeutic targets, tumor molecular profiling and personalized treatments directed toward oncogenic driver mutations have not yet proved effective in SCLC.
There is an acute need for advances in effective care and treatment of patients with SCLC. Recently, clinical studies using novel therapeutic approaches for SCLC have been reported.
In the setting of either conventional treatment or clinical trial research, patients may benefit from access to “real-time,” highly sensitive monitoring of disease burden. Such information can potentially provide early insight into treatment efficacy, the likelihood of benefit of life-extending procedures, recurrence of disease, and end-of-life decision making. The current standard of care is cross-sectional imaging. Although imaging is an essential clinical tool, it cannot detect occult disease or treatment-induced changes in tumor genotypes.
A significant obstacle to advancing translational SCLC research has been the difficulty in obtaining tumor material. Surgical resections are rare in this disease, and tumor biopsy specimens are often of small size and poor quality.
Molecular analysis of circulating tumor cells identifies distinct copy-number profiles in patients with chemosensitive and chemorefractory small-cell lung cancer.
In addition, unlike in recurrent NSCLC, biopsies of recurrent SCLC are rarely performed, which makes serial analysis of tumor evolution throughout chemotherapy difficult. Therefore, there is an urgent need to develop novel methods to obtain SCLC samples to fully assess genomic changes in this malignancy.
Here, we describe the development of a liquid biopsy assay for serial monitoring of SCLC-derived cell-free DNA (cfDNA). Implementation of the assay across a cohort of 27 patients with SCLC revealed a diversity of disease genotypes and dynamic changes in cfDNA allele fraction over the course of treatment.
Methods
Study Design and Patients
An institutional review board–approved protocol for collection of blood plus clinical information and treatment history was used to prospectively identify patients with SCLC and obtain consent from them. All samples were de-identified, and protected health information was reviewed according to the Health Insurance Portability and Accountability Act guidelines.
Blood Samples and Isolation of cfDNA
Blood samples (10–20 mL) were collected in Streck tubes (Streck, Omaha, NE) at various points before, during, and after therapy. Blood was centrifuged at 1200 g for 30 minutes. Plasma was removed and recentrifuged at 500 g for 30 minutes and immediately aliquoted and stored at −80°C. DNA was extracted from patient plasma samples with a Circulating Nucleic Acids Extraction kit following the manufacturer’s instruction, except that samples were incubated with proteinase K for 1 hour rather than 30 minutes (Qiagen, Hilden, Germany). The yield of double-stranded DNA was quantified using a Qubit fluorometer (Thermo Fisher, Waltham, MA) and the corresponding high-sensitivity DNA quantification kit. Approximately 40 to 100 ng of cfDNA, depending on the yield of cfDNA from the sample, was used for library construction.
Targeted Next-Generation Sequencing
A detailed description of the platform utilized for cfDNA sequencing is described in the Supplementary Methods. Briefly, the panel (Fig. 1A) contains 1608 probes that target all coding exons of BRAF, KIT proto-oncogene receptor tyrosine kinase gene (KIT), NOTCH1, NOTCH2, NOTCH3, NOTCH4, PIK3CA, PTEN, RB1, and TP53. The panel also contains probes for detection of copy variation in the genes FGFR1, MYC, MYCL1, and MYCN and control probes that target select regions in all 22 autosomes.
Figure 1Mutational analysis in plasma cell-free DNA (cfDNA) from 27 patients using next-generation sequencing. (A) Study overview. (B) Summary of mutations identified by individual patient at any time point. Patients are separated by stage of disease at diagnosis (L [limited stage] versus E [extensive stage]). Alterations are color-coded per the figure legend below the image. The mutation frequencies for each gene are plotted on the right panel. TP53, tumor protein p53 gene; RB1, retinoblastoma 1 gene; PTEN, phosphatase and tensin homolog gene; NOTCH1, notch 1 gene; NOTCH2, notch 2 gene; NOTCH3, notch 3 gene; NOTCH4, notch 4 gene; MYC, v-myc avian myelocytomatosis viral oncogene homolog gene; MYCL1, v-myc avian myelocytomatosis viral oncogene lung carcinoma derived homolog gene; MYCN, v-myc avian myelocytomatosis viral oncogene neuroblastoma derived homolog gene; PIK3CA, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha gene; KIT, KIT proto-oncogene receptor tyrosine kinase gene; FGFR1, fibroblast growth factor receptor 1 gene.
OS was defined from the date of diagnosis to the date of death from any cause or the last follow-up date (data cutoff date February 8, 2017). Cox proportional hazard models were used to assess the association between OS and cfDNA levels, which were measured as genomic equivalents in plasma. For each patient, multiple cfDNA levels were measured over time and included as a time-dependent covariate in all models. One-year survival was estimated from the univariate model. Hazard ratios (HRs) for every onefold increase in cfDNA genomic equivalents with 95% confidence intervals (CI) were reported for both univariate and multivariable models. Disease stage and treatment status were selected a priori on the basis of clinical knowledge to be controlled in the multivariate models.
Results
Disease-Associated Variants Are Detected at High Frequency in SCLC Plasma Samples
Over the course of approximately 26 months of peripheral blood collection, we obtained 140 samples from 27 patients with SCLC. Most patients (25 of 27 [93%]) had multiple samples collected, with a range of one to 12 collections per patient. Most of the 27 patients (59%) had ES SCLC, their median age was 66 years, and 52% were women (Table 1). Individual patient information with treatment details are listed in Supplementary Table 1.
Table 1Baseline Demographics of Patients in This Study (N = 27)
For serial monitoring of circulating tumor burden, we applied a hybrid capture next-generation sequencing technology specifically designed for quantitative analysis of somatic alterations in cfDNA (Supplementary Methods). We developed a custom capture panel that targets 14 genes known to be frequently mutated in SCLC tumors
(see Fig. 1A). We detected disease-associated mutations in peripheral blood from 23 patients (85%) (Fig. 1B), with an allele frequency (AF) range of 0.1% to 87%. Six or more genomic alterations were detected in eight of 27 of patients (30%). Of 140 plasma samples analyzed, 82 (59%) had detectable somatic mutations. Significantly, 35 of the 58 mutation-negative samples (60%) were from patients whose somatic alterations dropped to undetectable levels in response to therapy (Supplementary Tables 1 and 2). Notably, in cases in which pretreatment plasma samples were available, the cfDNA somatic mutation signature in patients who experienced disease relapse was identical to the pretreatment profile.
The most common disease-associated mutations were detected in TP53 (19 of 27 patients [70%]) and RB1 (14 of 27 patients, [52%]). Fourteen patients (52%) had detectable mutations in both TP53 and RB1. In addition to mutations in TP53 and RB1, we detected alterations in 10 additional genes in our patient population (PTEN, NOTCH1–4, MYC, MYCL1, PIK3CA, KIT, and BRAF) (see Fig. 1B). Fourteen of 27 patients (52%) had inactivating mutations in NOTCH family genes, and four of 27 patients (15%) had genomic amplifications of a MYC family member. Copy number alterations included amplifications in NOTCH2, NOTCH3, MYC, MYCL1, PIK3CA, and BRAF and deletions in TP53, RB1, NOTCH1, and KIT. In total, 91 alterations in 12 genes, including 42 single-nucleotide variants, 39 copy number alterations, and 10 insertion/deletions (indels) were detected.
cfDNA tracking is now being used as the standard of care in management of certain genotypes of NSCLC.
we compared mutant AFs between these two types of lung cancer. We analyzed plasma samples from 43 patients with metastatic NSCLC with known TP53 mutations and compared the AFs of TP53 mutations with those of the patients with SCLC in our cohort. The observed AFs were significantly higher for SCLC (range 0.45%–70.4% in SCLC versus 0.4%–36.9% in NSCLC [ p = 0.006]) (Supplementary Fig. 1).
Longitudinal cfDNA Analysis Identifies Disease Recurrence before Radiographic Evidence of Tumor Progression
To determine whether cfDNA sequencing can be used to monitor a patient’s response to systemic therapy, we analyzed serial plasma samples from 25 patients before, during, and after therapy. We detected a marked increase in mutation abundance that preceded radiographic evidence of disease progression in nine patients (VSC-4, VSC-8, VSC-10, VSC-11, VSC-12, VSC-13, VSC-25, VSC-27, and VSC-29). Two representative examples of how changes detected in cfDNA preceded radiographic progression of disease during a patient’s clinical course are highlighted in the following paragraphs (Figs. 2 and 3), and the remaining case descriptions are included in Supplementary Figures 2 through 5 and Supplementary Table 2.
Figure 2Cell-free DNA detection precedes clinical or radiographic disease progression. (A) The time line for the clinical course of patient VSC-8 from diagnosis until date of death is shown. The light blue bar represents the treatment time frame, and the red dots indicate blood collection time points. Radiographic images were acquired at the time of diagnosis, on day 144, and on day 256. The red arrow in the first CT scan shows the right pleural–based primary tumor, which resolved on further imaging. Bone marrow biopsy was performed on day 289 and assessed for common SCLC markers, including CD56 (hematoxylin and eosin; original magnification, ×200 and CD56 stain; original magnification, ×200). (B) Percent mutant allelic frequency and copy number alterations for patient VSC-8 are shown. The light blue box indicates treatment time frame. A plus or minus symbol indicates presence/absence of the copy number alteration listed, and an asterisk indicates a stop. SNV, single-nucleotide variant; TP53, tumor protein p53 gene; PIK3CA, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha gene; RB1, retinoblastoma 1 gene; amp, amplification; del, deletion.
Figure 3Cell-free DNA detection can clarify mixed response on imaging. (A) The time line for the clinical course of patient VSC-10 from diagnosis until last follow-up date with radiographic images is shown. The colored bars represent the active treatment time frames, including radiotherapy (RT) (as indicated by the gray bar). The purple bar indicates prophylactic cranial irradiation (PCI), and the blue shaded bar represents concurrent chemoradiation (chemorad) with weekly carboplatin and paclitaxel. The red dots indicate blood collection time points. Radiographic images were obtained at time of progression after first-line therapy (day 223) and on days 326, 417, and 494. The red arrow at day 223 indicates initially multifocal pulmonary parenchymal disease which resolved after second-line paclitaxel (day 326). The subsequent image at day 417 shows a 0.9-cm retroperitoneal lymph node of uncertain etiology. On the day 494 images, the retroperitoneal lymph node enlarged to 3.0 x 2.7 cm. (B) Percent mutant allelic frequencies and copy number alterations for patient VSC-10 are shown. The light red and blue boxes indicate treatment time periods corresponding to labeling on the time line above. The dotted line indicates the time of radiologic recurrence. A plus or minus symbol indicates presence/absence of the copy number alteration listed. CT, computed tomography; TP53, tumor protein p53 gene; NOTCH3, notch 3 gene; del, deletion.
LS SCLC was diagnosed in patient VSC-8 (see Fig. 2A). Before initiation of first-line chemoradiotherapy with carboplatin and etoposide (day 32), we detected a TP53 (R65*) single-nucleotide variant at 32.2% AF, a PIK3CA amplification (4.0 copies), and copy number loss in both TP53 and RB1 (Fig. 2B). After one cycle (day 60), the TP53 (R65*) mutation AF decreased to 4.7% and the copy number changes in PIK3CA, TP53 and RB1 were no longer detectable. Repeat cfDNA assessments after the third and fourth cycles (day 109 and 144, respectively) of first-line therapy detected no somatic mutations (see Fig. 2B), and imaging after completion of the fourth cycle showed a partial response (see the day 144 image in Fig. 2A). Eight weeks after completion of first-line therapy (day 200), cfDNA analysis showed recurrence of the TP53 (R65*) mutation at an AF of 1.3%, but imaging showed an excellent response in pulmonary parenchymal disease and hilar adenopathy with ambiguous, subcentimeter adrenal nodularity and no clear radiographic progression (data not shown). At day 256 after diagnosis, the AF of the TP53 (R65*) mutation increased to 74% (>twofold higher than at initial diagnosis), the PIK3CA amplification (3.0 copies) reappeared, and deletions in TP53 and RB1 were again detected. Imaging continued to show ambiguous findings in the postradiation field (see the day 256 image in Fig. 2A), and 33 days later, with clinical worsening and the development of pancytopenia, progression become apparent when a bone marrow biopsy showed involvement by SCLC (see the images at day 289 in Fig. 2A). This patient did not have a pretreatment bone marrow biopsy, as he had mild anemia but no other cytopenia at the time of diagnosis. In this case, cfDNA testing captured disease progression that was not seen on computed tomography (CT) imaging.
cfDNA Testing May Clarify Ambiguous Radiographic Findings and Detect Occult Disease
Although the previous case highlights disease progression in an occult anatomic location detected by peripheral blood cfDNA, many times disease progression on imaging is difficult to discern from treatment effects, infection, or inflammatory changes. In these cases, more information is needed before patients are subjected to changes in therapy. Currently, additional time with repeat imaging is the standard approach, but cfDNA monitoring has the potential to add valuable insight in these cases, with ambiguous imaging findings as noted in the ensuing case.
LS SCLC was initially diagnosed in patient VSC-10, who received first-line chemoradiotherapy (cisplatin and etoposide) with a partial response. At day 223 after diagnosis, local recurrence of the tumor was noted on CT scan (see the day 223 image in Fig. 3A) and confirmed with supraclavicular lymph node biopsy; second-line therapy with paclitaxel was initiated. One week after initiation of second-line treatment (day 251), we detected alterations in TP53 (E285K) and NOTCH3 (G1551C) at rates of 6.7% and 2.4%, respectively (see Fig. 3B). The patient achieved a complete response with second-line therapy (see the day 326 image in Fig. 3A), and cfDNA analysis 6 weeks (day 284) and 13 weeks (day 326) after initiation of second-line therapy detected no tumor associated alterations. However, nearly 6 months after initiation of second-line therapy (day 417) the patient’s same disease-associated mutations reappeared (TP53 E285K at 3.9% and NOTCH3 G1551C at 3.7%). Cross-sectional imaging at this time (see the day 417 image in Fig. 3A) showed no demonstrable tumor recurrence, and in fact, there was an interval decreased size of the lung nodules versus on the previous scan and a small (0.9-cm) left retroperitoneal lymph node that was of uncertain etiology. Repeat cfDNA analysis on day 494 revealed a more than 20-fold increase in TP53-mutant AF (E285K at 82.6%) and a more than 10-fold increase in NOTCH3-mutant AF (G1551C at 39.9%). Surveillance imaging on day 494 revealed massive, intraabdominal disease (see the day 494 image in Fig. 3A). The left retroperitoneal lymph node now measured 3.0 × 2.7 cm. There were also a new 5.9 × 4.5 × 4.1-cm hepatic metastasis, diffuse intraabdominal lymphadenopathy, and new pleural-based metastases (data not shown). This case illustrates an example in which liquid biopsy testing was a valuable diagnostic companion to imaging in surveillance for disease recurrence. Similar observations were made in four additional cases (VSC-1, VSC-7, VSC-12, and VSC-13 [see Supplementary Figs. 4, 6, and 7 and Supplementary Table 2, respectively]).
Serial cfDNA Analysis Could Be Used to Refine Treatment Intensity
Prophylactic cranial irradiation (PCI) after first-line chemotherapy is the standard of care for patients with LS SCLC.
Prophylactic cranial irradiation for patients with small-cell lung cancer in complete remission. Prophylactic Cranial Irradiation Overview Collaborative Group.
Detection of tumor-associated mutations in cfDNA could potentially be used to identify patients most likely to benefit from PCI while sparing other patients from the toxicities associated with whole brain radiotherapy. The case of patient VSC-9, a 40-year-old woman with LS SCLC, illustrates this point (Fig. 4A). Profiling of the patient’s cfDNA before first-line chemoradiotherapy revealed a MYCL1 amplification (15.5 copies [Fig. 4B]). After treatment with cisplatin and etoposide, this amplification fell to undetectable levels in all serial samples collected up to 413 days after diagnosis (see Fig. 4B), suggesting that the patient had platinum-sensitive disease. CT imaging after treatment completion (see the day 93 image in Fig. 4A) and 15 months after completion of therapy (see the day 539 image in Fig. 4A) also showed no evidence of disease. Although both CT imaging and cfDNA testing showed complete response to chemotherapy, the patient still received PCI (days 112–126 after diagnosis [see Fig. 4A]) as the standard of care in an attempt to prevent brain metastases. In such cases, the absence of detectable ctDNA could add a piece of information that would allow discussion regarding the likelihood of central nervous system recurrence and potential impact of PCI. Two other patients in our cohort had similar clinical scenarios, with persistent clearance of tumor-associated variants after treatment (Supplementary Figs. 8 and 9). None of these patients had disease recurrence at the most recent follow-up (at 13, 20, and 21 months after diagnosis). At this time, we do not have sufficient data to confidently state the concordance between systemic disease control and intracranial disease control in patients with SCLC when the results of cfDNA analysis are negative.
Figure 4Cell-free DNA changes correspond to remission. (A) The time line for the clinical course of patient VSC-9 from diagnosis until last follow-up date with radiographic images is shown. The colored bars represent the treatment time frames. The gray bar indicates radiation therapy (RT) and the purple bar indicates prophylactic cranial irradiation (PCI). Radiographic images were obtained at time of diagnosis, on day 93, and at last follow-up date (day 539). The red arrow shows primary mediastinal disease, which resolved on subsequent imaging. (B) Copy number alterations for patient VSC-9 are shown. The blue box indicates the time period during which cisplatin plus etoposide was administered, and the purple box indicates PCI treatment. A plus or minus symbol indicates presence/absence of the copy number alteration listed. CT, computed tomography; MYCL1, v-myc avian myelocytomatosis viral oncogene lung carcinoma derived homolog gene; amp, amplification.
In several patients (VSC-1, VSC-12, VSC-14, VSC-18, and VSC-27 [Fig. 5A and B and Supplementary Figs. 2, 4, 7, and 10]), cfDNA mutation tracking provided early evidence of resistance to therapy. In all these cases, we noted an obvious rise in mutation abundance during treatment. For instance, patient VSC-14 (with ES SCLC) enrolled before first-line therapy with carboplatin and etoposide (see Fig. 5A). We detected two somatic mutations in TP53 (R158L) and RB1 (Q850*) at AFs of 63.1% and 64.9%, respectively; deletions in KIT, TP53, RB1, and NOTCH1; and a PIK3CA amplification (see Fig. 5B). The patient did not tolerate the first cycle of carboplatin/etoposide, and repeat cfDNA assessment 6 weeks after treatment initiation (day 49) revealed the same genomic alterations with TP53 (R158L) and RB1 (Q850*) mutations at AFs of 54.5% and 57.4%, respectively. Imaging revealed ongoing progression of disease at this time, with abdominal adenopathy, pleural, and adrenal metastases slightly increasing in size (see the day 49 image in Fig. 5A). One week later, the patient began taking nivolumab, and cfDNA revealed TP53 (R158L) and RB1 (Q850*) mutations at AFs of 51.0% and 52.9%, respectively. One month after initiation of nivolumab (day 82), cfDNA revealed AFs of TP53 (R158L) at 73.0% and RB1 (Q850*) at 71.9% before response imaging was obtained. Eight weeks after initiation of nivolumab (day 111), AFs continued to increase to 81.6% for TP53 (R158L) and 81.4% for RB1 (Q850*) and imaging showed diffuse, progressive disease throughout the thorax and abdomen (see the day 111 image in Fig. 5A). The patient started receiving third-line therapy with paclitaxel, and 1 month after therapy initiation (day 138), cfDNA revealed AFs of TP53 (R158L) at 77.5% and RB1 (Q850*) at 77.3%. Imaging 2 weeks later showed that the patient had progressed (see the day 152 image in Fig. 5B); no cfDNA was assessed at that time. The patient declined clinically and died 7 weeks later. In this patient’s case, the peripheral blood cfDNA AFs continued to increase during nivolumab and paclitaxel therapy, in both cases before repeat imaging. Early identification of nonresponders with cfDNA monitoring could alleviate unnecessary toxicity and promote timely changes in therapy.
Figure 5Cell-free DNA sequencing enables early identification of treatment refractory disease. (A) The time line for the clinical course of patient VSC-14 from diagnosis until date of death is shown with radiographic images from diagnosis and on days 49, 111, and 152, demonstrating slow progression in intrathoracic disease despite all therapy. The red arrow indicates primary mediastinal disease. The color bars represent treatment time frames. The blue bar represents one cycle of carboplatin and etoposide treatment, the light yellow box represents nivolumab treatment, and the pink box indicates paclitaxel treatment. (B) Percent mutant allelic frequencies and copy number alterations for patient VSC-14 are shown. The light blue, yellow, and pink boxes indicate treatment periods corresponding to labeling on the time line above. The dotted line indicates the time of radiologic recurrence. A plus or minus symbol indicates presence/absence of the copy number alteration listed, and the asterisk indicates stop. SNV, single-nucleotide variant; TP53, tumor protein p53 gene; RB1, retinoblastoma 1 gene; PIK3CA, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha gene; KIT, KIT proto-oncogene receptor tyrosine kinase gene; amp, amplification; del, deletion; NOTCH1, notch1 gene.
We assessed whether cfDNA levels, measured as genomic equivalents (GEs) in plasma, had prognostic significance. Increased cfDNA GEs were associated with worse OS in both univariate analysis (HR = 2.65, 95% CI: 1.41–4.98, p = 0.0024) and multivariate analysis (HR = 2.73, 95% CI: 1.27–5.86, p = 0.0099) (Supplementary Table 3). The predicted 1-year survival followed a stepwise pattern on the basis of the GEs detected in our cohort. For patients with cfDNA GEs of 2000 (median for the overall cohort 2044), the 1-year survival was 90%. For patients with cfDNA GEs of 4000, 6000, 8000, 12,000, and 16,000, the 1-year survival probabilities were 75%, 60%, 47%, 26%, and 13%, respectively. The genomic equivalents and cfDNA concentrations for all patients in our cohort across all collection dates are provided in Supplementary Table 4.
Discussion
The results from our study confirm that SCLC-associated cfDNA is detectable in peripheral blood in more than 80% of patients when our custom SCLC-specific gene panel is used. This rate of detection of tumor-associated peripheral blood biomarker is analogous to that of the more labor-intensive strategy of isolating circulating tumor cells in patients with SCLC.
Clinical significance and molecular characteristics of circulating tumor cells and circulating tumor microemboli in patients with small-cell lung cancer.
Furthermore, cfDNA appears to have a higher sensitivity than circulating tumor cells when compared across multiple histologic types, a finding suggesting that these markers may have unique pathologic significance.
Moreover, we demonstrate that cfDNA monitoring in patients with SCLC has the potential to be a useful clinical tool. Specifically, we have shown that tracking of cfDNA mutation abundance is able to detect disease recurrence and occult disease that were not evident with radiographic imaging. Earlier detection may provide relatively fit patients with secondary treatment options. We found that some patients had lasting responses, and these patients may benefit from forestalling potentially toxic procedures such as PCI. Other patients showed little response to therapy, and in these cases palliative care may be preferable to the possible harm of ineffective treatments. Finally, as new therapeutic options are tested in clinical trials, we believe that cfDNA assessment can provide early insight into treatment efficacy.
Broad clinical implementation of liquid biopsies in the care of patients with SCLC is a realistic goal. The gene panel used in this study had 14 genes that have been found to be frequently mutated in SCLC tumors.
It was designed to monitor disease in the circulating DNA of patients with SCLC by quantifying SCLC-associated somatic variants that are at potentially low AFs through sequencing of high-coverage genomic libraries. In fact, the AFs of circulating somatic mutations in patients were much higher than in other forms of lung cancer (see Supplementary Fig. 1). This finding may indicate increased tumor shed in SCLC compared with in NSCLC, although other factors—specifically, the extent of metastatic disease—need to be considered. The high signal-to-noise ratio observed in this study suggests that routine cfDNA monitoring in this disease setting is technically feasible. The modest breadth of the gene panel translates into a sequencing cost that may be economically feasible as well. Our immediate objective is to evaluate potential benefits to patients with SCLC in prospective studies in which molecular analysis is included in patient care.
The canonical mutations in patients with SCLC are alterations in TP53 and RB1.
In studies of tumor tissue predominantly from patients with LS SCLC before first-line chemotherapy, TP53 mutations were identified in 80% to 95% of patients and RB1 mutations were identified in 35% to 70%.
In our study of cfDNA predominantly in patients with ES SCLC after initiation of systemic chemotherapy, the rates of detection of TP53 and RB1 mutations were 70% and 52%, respectively. These rates are similar to the previously presented rates of 70% and 32%, respectively, that were detected in peripheral blood in a cohort of patients with SCLC.
The differences in TP53 and RB1 mutation rates may relate to differences in extent of disease, timing of recent therapy, and/or partial shedding of tumor cfDNA into circulation. Also, RB1 mutations are likely underreported in our analysis because RB1 is frequently deleted in SCLC
and somatic deletion events are difficult to detect at low AFs.
Despite the fact that our monitoring panel was not designed to discover novel mutational profiles, we observed a higher than anticipated mutation rate in NOTCH genes.
The broad spectrum of mutational events in the NOTCH family is consistent with the proposed tumor-suppressive role of NOTCH gene function in this disease.
The discrepancy may be due to the fact that the NOTCH coding regions are guanine-cytosine–rich sequences that can be difficult to sequence. Consistent with this idea, the same study that showed a 25% mutation rate in NOTCH genes also showed that 77% of tumors had an expression profile consistent with NOTCH inactivation.
Future studies are likely to focus on clinically actionable aspects of NOTCH gene loss in SCLC.
The clinical implications of co-occurrence of the tumorigenic mutations we identified necessitate further study. For example, murine models of SCLC have shown that co-occurrence of TP53 and RB1 mutations with MYC amplification may predispose to sensitivity to aurora kinase inhibition.
Further connections between the hallmark genomic changes in SCLC and additional pathogenic mutations identified by us must be carefully recorded to identify molecular patterns that may prove particularly susceptible to novel therapies.
There are several limitations to this work. First, our study included a relatively small number of peripheral blood samples from patients at the time of initial diagnosis (that is, treatment-naive patients).
This may be partially accounted for by the fact that some patients receive first-line therapy in the community and are then referred to an academic medical center for second-line therapy and beyond. Second, we were not able to obtain blood samples from every patient before cross-sectional imaging, which limited our ability to detect occult disease. Some patients did not have blood draws between imaging studies, and because we obtained research samples only when blood was being drawn as part of the standard of care, this may explain why we were not able to capture blood between imaging studies for all patients in our cohort. Although the variable blood collection timing and absence of a control group limit the rigor of the current analysis, we believe that the data are of importance for the burgeoning field of cfDNA analysis in patients with SCLC. Third, the sample size of our patient cohort precluded statistical subgroup analyses. Finally, it must be noted that the peripheral blood cfDNA assay we used is one of several commercially available options, all of which continue to necessitate prospective investigation regarding their ability to improve patient outcomes.
Although cfDNA identification technique and cutoff values for statistical analysis have varied widely in patients with NSCLC, this is the only recent cfDNA analysis to be linked to clinical outcomes in patients with SCLC,
allowing for a more uniform standard to be applied in replicating this observation in future studies. The uniquely longitudinal data provided in our analysis sets the stage for future studies of peripheral blood cfDNA as an early predictor of disease progression in patients with SCLC, similar to the conceptual application of cfDNA in patients with breast cancer,
Sensitivity of plasma BRAFmutant and NRASmutant cell-free DNA assays to detect metastatic melanoma in patients with low RECIST scores and non-RECIST disease progression.
The optimal frequency of peripheral blood cfDNA monitoring in patients with SCLC (every 2 weeks, 4 weeks, etc.) needs validation in future prospective studies. We propose a randomized controlled trial comparing standard of care management of patients with SCLC to standard of care plus peripheral blood cfDNA monitoring to detect early relapse and the need for therapy reinitiation or change. Because of the aggressive nature of SCLC and current limitations in treatment options for this patient population, better lead time information predicting progression is of unclear clinical benefit, but we hypothesize that clinical outcomes in patients with SCLC may be improved by detecting and treating disease relapse sooner than is currently possible with conventional imaging. We do not advocate monitoring cfDNA in all patients with SCLC as standard of care at this time; rather, we believe that the additional prospective analyses we have proposed are still required to ensure this novel technology is best applied to clinical decision making.
In conclusion, our study demonstrates that quantitative changes in cfDNA levels correlated with responses to therapy and relapse of disease in patients with SCLC. The hope is that prospective application of this technology will translate into improved outcomes for patients afflicted with this dreadful disease.
Acknowledgments
This study was supported in part by a Vanderbilt Ingram Cancer Center Young Ambassadors Award and by the National Institutes of Health and National Cancer Institute R01CA121210 (to Dr. Lovly). Dr. Lovly was also supported by a Damon Runyon Clinical Investigator Award, a LUNGevity Career Development Award, a V Foundation Scholar-in-Training Award, an American Association for Cancer Research–Genentech Career Development Award, and grant U10CA180864. Dr. Almodovar was supported by a Ruth L. Kirschstein National Research Service Award Fellowship (T32HL094296). Mr. Zhao was supported in part by National Cancer Institute/National Institutes of Health Cancer Center Support Grant 2P30CA068485-19. We first and foremost would like to thank the patients and their families. We are extremely grateful to the Vanderbilt Ingram Cancer Center Young Ambassadors Award for their generous support of this pilot project; to Anel Muterspaugh, Hina Chowdhry, and Brandon Winston for their assistance in obtaining consent from patients and collecting patient samples; to Dr. Adam Seegmiller for providing the bone marrow biopsy images, and to the entire Lovly laboratory and Darren Tyson for their thoughtful and critical review of the manuscript. Drs. Almodovar, Iams, Lim, Raymond, and Lovly designed the experiments. Drs. Almodovar, Yan, Hernandez, Lim, and Raymond performed the experiments. Drs. Almodovar, Iams, Meador, Hernandez, Lim, Raymond, Lovly generated and analyzed data. Drs. Horn, York provided direct patient care. Drs. Almodovar, Iams, and Lovly wrote the manuscript. Drs. Zhao, Chen, Shyr performed the statistical analysis. Drs. Almodovar, Iams, Meador, Zhao, Horn, Lim, Raymond, and Lovly reviewed the data and the final manuscript.
Molecular analysis of circulating tumor cells identifies distinct copy-number profiles in patients with chemosensitive and chemorefractory small-cell lung cancer.
Prophylactic cranial irradiation for patients with small-cell lung cancer in complete remission. Prophylactic Cranial Irradiation Overview Collaborative Group.
Clinical significance and molecular characteristics of circulating tumor cells and circulating tumor microemboli in patients with small-cell lung cancer.
Sensitivity of plasma BRAFmutant and NRASmutant cell-free DNA assays to detect metastatic melanoma in patients with low RECIST scores and non-RECIST disease progression.
Disclosure: Dr. Lovly has served as a consultant for Pfizer, Novartis, AstraZeneca, Genoptix, Sequenom, and Ariad and has been an invited speaker for Abbott and Qiagen. Dr. Horn is a consultant for Abbvie, Bayer, Bristol-Myers Squibb, Eli Lilly, Merck, Roche, and Xcovery. Ms. Hernandez, and Drs. Lim, Raymond have an ownership stake in and are employees of Resolution Biosciences. The remaining authors declare no conflict of interest.
SCLC accounts for about 15% of lung cancers. Although SCLC is initially sensitive to chemotherapy, it becomes aggressive and resistance to chemotherapy and remains a dismal disease, with a 20% to 30% and 1% to 3% 5-year survival rates for limited disease and extensive disease, respectively. In contrast to the situation in NSCLC, druggable targets are rare in SCLC. Despite the rarity of surgical resection and poor quality with the small size of SCLC tumor biopsy specimens, several landmark studies have demonstrated a high incidence of genomic alterations, including inactivating mutations of the tumor protein p53 gene (TP53) and retinoblastoma 1 gene (RB1) found in more than 90% of SCLCs.
32705-3&id=gr1.jpg){kind=link}
32705-3&id=gr2.jpg){kind=link}
32705-3&id=gr3.jpg){kind=link}
32705-3&id=gr4.jpg){kind=link}
32705-3&id=gr5.jpg){kind=link}