Most patients whose large bowel cancer has spread to other organs do not respond to immune therapy. We detected a rare gene mutation, termed 9p24.1 copy-number gain (CNG), in an otherwise incurable colorectal cancer that provoked an immune therapy response. We identified this gene mutation by gene-panel sequencing of DNA from a liver metastasis biopsy from a patient who had disease refractory to standard therapies. Following immune checkpoint blockade (ICB) with pembrolizumab (anti–PD-1), the patient experienced conversion of the tumor phenotype from one with epithelial features to that of an inflamed microenvironment, detected by high-resolution RNA sequencing. Circulating tumor DNA disappeared over the first weeks of therapy. As assessed by standard radiographic measurement, the patient had a partial response that was durable. This patient's response may support the use of histology-agnostic ICB in solid tumors that carry the rare 9p24.1 CNG.
The immune context of the tumor microenvironment (TME) determines the response to immune checkpoint blockade (ICB) in colorectal cancer (1, 2). Studies have demonstrated favorable survival following ICB in metastatic disease from the minority of cases with primary colorectal cancer that is deficient in the DNA-mismatch repair (MMR) function reflected by either high tumor mutational burden or positive microsatellite-instability (MSI) status (3, 4). Unlike for non–small-cell lung cancer (NSCLC), tumor-cell expression of the ligand of the programmed death receptor (PD-1), PD-L1, has not been used as a predictive biomarker of ICB response in colorectal cancer.
Despite the extensive immune-cell infiltrate recruited by the malignant Reed–Sternberg cells in classic Hodgkin lymphoma, the cytotoxic immune cells are anergic. This is frequently caused by Reed–Sternberg cell copy-number gain (CNG) of chromosome 9p24.1, which includes loci for CD274, PDCD1LG2, and JAK2. This CNG leads to the enhanced expression of PD-L1, PD-1 ligand PD-L2, and Janus kinase-2 (JAK2), and thereby maintains Reed–Sternberg cell survival signaling, unless evasion of T-cell cytotoxicity is revoked by ICB (5–8). The CNG of the 9p24.1 locus, associated with susceptibility to ICB, has been observed in rare lymphoma entities, and its prevalence in solid tumors is low, estimated to only 0.2% of colorectal cancer cases (9).
The “Actionable Targets in Cancer Metastasis” (MetAction) study (ClinicalTrials NCT02142036) established expedited and safe mutation profiling of individual patients' progressing metastatic cancer following failure of systemic therapies, to offer histology-agnostic, molecularly matched medication for end-stage cancer (10). The treatment-refractory 9p24.1-CNG colon cancer case reported here illustrates the utility of this pipeline and further delineates the ICB response by diverse modalities.
Materials and Methods
The study was approved by the Institutional Review Board, the Regional Committee for Medical and Health Research Ethics of South-East Norway (reference number, REK 2013/2099), and the Norwegian Medicines Agency. It was performed in accordance with the Declaration of Helsinki. Written informed consent was required for participation. The study procedures and endpoints have been described previously (10) and are, together with details about the gene mutation profiling, provided in Supplementary Methods (10–12).
DNA/RNA extracted from biopsy-sampled metastatic tissue and from the whole blood sample as germline control at enrollment was sequenced using the Ion Oncomine Comprehensive Assay v1 (Thermo Fisher Scientific). The median sequencing depth was approximately 3,000×, enabling the calculation of the approximate mutant-allele fraction in heterogeneous tissue samples. The DNA sequence data were analyzed by Torrent Suite Software (Thermo Fisher Scientific); specifically, copy-number variants were detected in the copy-number detection module of the Ion Reporter Software (Thermo Fisher Scientific). The called gene variants were filtered prior to assessment and prioritization. The Molecular Tumor Board interpreted the findings with regard to the likelihood of tumor-signaling activity being involved, delivering conclusions that the following Clinical Tumor Board used to recommend potential systemic tumor-directed medication.
FISH analysis of the resected primary and locally recurrent tumors was performed to assess chromosome 9p copy-number, using a commercial probe pair that included a 350-kilobase 9p21.3-specific probe covering CDKN2A and CDKN2B, labeled with PlatinumBright550, and a 440-kilobase 9q21-specific probe spanning from RH40366 to SHGC-82248, labeled with PlatinumBright495 (both Kreatech probes KBI-10402).
Cell-free DNA was extracted from plasma sampled at each CT examination. The quantitative digital PCR assay, capable of discriminating between the KRAS p.G12D and wild-type sequences, was performed on the Raindrop System (RainDance Technologies).
RNA was extracted from liver metastasis biopsies sampled at study enrollment and 2 weeks of treatment for high-resolution sequence analysis that yielded a minimum of approximately 35 million reads per sample. Each transcript was quantified and compared for expression level in the two biopsy specimens. Differentially expressed genes were analyzed for significant biological functions and signaling pathways inherent in the tumor phenotype responses using the Ingenuity Pathway Analysis Software. Details on these procedures are provided in Supplementary Methods.
Following conclusion of the diagnostic procedure, the patient had a baseline CT examination and commenced treatment with pembrolizumab (2 mg/kg every 3 weeks). Evaluation CT examinations were undertaken every 9 weeks. Routine blood tests were done within the standard patient work-up.
In October 2015, a 51-year-old woman with an otherwise unremarkable medical background presented with anemia and was diagnosed with a conventional-type colon adenocarcinoma located in the hepatic flexure and extending into the right liver lobe. The tumor was positive for the KRAS p.G12D mutation and MSI markers and devoid of the MMR protein PMS2. The patient began neoadjuvant oxaliplatin-based chemotherapy [the Nordic FLOX regimen (13)], but CT evaluation after 3 cycles revealed locally progressive disease. The patient underwent a right-sided hemicolectomy and in-continuity resection of the involved liver segments. Histologic examination of the surgical specimen confirmed margin-negative ypT4bN0V0-stage disease. The patient was not offered adjuvant oncologic therapy.
Repeat CT scanning 4 months after the surgical procedure showed a locally recurrent tumor comprising the duodenum and pancreas, and the patient proceeded directly to salvage surgery. The histologic examination noted a positive margin toward the posterior abdominal wall and tumor infiltration in a portal vein. One month postoperatively the patient commenced a 5-week-long chemoradiotherapy regimen (delivered to the involved area of dissection).
A repeat CT record 3 months after completion of the adjuvant treatment revealed one metastatic lesion in each liver lobe in addition to an enlarged portal hilum lymph node. The patient was assessed for palliative irinotecan-based chemotherapy [the Nordic FLIRI regimen (14)], but disease progression was noted after 4 cycles. Immune checkpoint inhibitors are not approved by Norwegian health authorities for treatment of advanced MSI-positive colorectal cancer within the public health services, thus, the patient was ineligible in routine clinical practice. In February 2017, she was enrolled in the MetAction study.
Metastasis genome profiling
Following study enrollment, ultrasound-guided core biopsies were taken from the metastatic lesion in the left liver lobe without procedure-induced complications. Tumor cell content was 30% as assessed by tissue imprint. The DNA sequencing revealed point mutations commonly observed in colorectal cancer (APC p.R216X, KRAS p.G12D, and FBXW7 p.R465H), each with approximately 30% mutant-allele fraction, in addition to an estimated ≥4-fold relative 9p-CNG (Fig. 1A), confirmed by FISH (Fig. 1B). Acknowledging the approximations by applying the Oncomine Comprehensive Assay to estimate the tumor mutational burden, defined as the number per megabase of coding synonymous base substitutions and indel mutations, but excluding known driver mutations (15), only one of the identified mutations was eligible. This corresponded to a tumor mutational burden rate of 5, which is typical for MMR-proficient colorectal cancer (16), although her tumor was MMR-deficient.
Following tumor board consensus supporting treatment with a PD-1 inhibitor and patient's commencement of pembrolizumab treatment, blood tests already showed a decline in tumor markers at 2–3 weeks (Fig. 2A). The moderately elevated liver enzymes decreased over the entire treatment course; liver function markers as well as differential blood counts, electrolytes, and other renal function markers have all been within the reference limits throughout the therapy. The first CT evaluation after 3 treatment cycles revealed a partial response, according to the RECIST v1.1, which remained consolidated at the 1-year treatment evaluation (Fig. 2B). The patient has been continuing the treatment for 20 months with ongoing partial response (by November 2018).
As shown by Supplementary Fig. S1, the patient experienced a rise in plasma free-T4 and a corresponding fall in thyroid-stimulating hormone only 2–3 weeks after commencement of therapy. Grade 1 hyperthyroidism (by Common Terminology Criteria for Adverse Events v4.0) was evident when the patient started the third treatment cycle, whereas grade 2 hypothyroidism was recorded at commencement of cycle 4. Since then, the patient has been on continuous medication with levothyroxine that has been dose-adjusted on the basis of the continuous monitoring of plasma thyroid-stimulating hormone and free-T4. The patient has not reported other adverse events.
Circulating tumor DNA response
Plasma analysis showed a decline in the KRAS mutant-allele fraction from 23% at baseline to 2.4% at 2 weeks of study treatment (Fig. 2C). A mutant-allele fraction of only 0.1% was detected in the concurrent liver metastasis biopsy. Plasma specimens sampled at each of the CT evaluations were negative for KRAS p.G12D (Fig. 2C).
Metastasis phenotype response
The RNA transcript profile of the left liver lobe metastasis (the individual RNA sequence counts are provided in Supplementary Table S1) changed over the first 2 weeks of PD-1 blockade (Supplementary Table S2). Although expression of CD274 (PD-L1) and JAK2 diminished, PDCD1LG2 (PD-L2) expression increased in the on-treatment biopsy (Supplementary Table S3). Among the most discriminating RNA profile-defined functions altered between the baseline and 2-week samples were activation of responses ascribed to immune-cell trafficking, lipid metabolism, and inflammation, whereas biological features related to epithelial neoplasia were inhibited (Fig. 3A; Supplementary Table S4). Moreover, profile-defined signaling pathways that were induced by pembrolizumab included activation of the liver-X/retinoid-X hormone receptor transcription factors (LXR/RXR) and acute-phase reaction with the complement system (Fig. 3B; Supplementary Table S5). An 18-gene T-cell–inflamed expression profile that predicted response to pembrolizumab across multiple solid tumors has been reported (17). In our dataset, a majority of these IFNγ-responsive genes were activated in the metastasis at 2 weeks of pembrolizumab treatment, whereas IDO1 expression was repressed (Fig. 3C).
We report a radiographically partial and durable response of advanced colon cancer on PD-1 blockade based on 9p24.1-CNG detected in a metastatic liver lesion. A decline in circulating routine tumor markers and the mutant KRAS DNA was already observed after the first therapy cycle, mirroring the conversion of the metastasis phenotype from epithelial features to characteristics of an inflamed TME. The response measured with each of diverse modalities was consistent with the MetAction diagnostic procedure concluding that the CNG of chromosome 9p24.1 represented the driver alteration. This cause of ICB susceptibility is rare in solid tumors and particularly in colorectal cancer (9). One other case of advanced 9p24.1-CNG colon adenocarcinoma treated with a PD-1 inhibitor has been reported (18).
The therapeutic response came with a common toxicity, a transient hyperthyroidism preceding hypothyroidism that was managed by hormone substitution. Patients with advanced NSCLC have better response and outcome measures to ICB if they develop thyroid dysfunction, which is an independent predictive factor (19, 20). This suggests a mechanistic connection between the thyroid hormone metabolism and ICB efficacy. In mice, the T4 metabolite triiodothyronine stimulates tumor-antigen–presenting dendritic cells to impel IFNγ-mediated T-cell cytotoxicity (21).
Some studies have reported on the analysis of circulating tumor DNA (ctDNA) to assess ICB efficacy in advanced cancer. A prospective pilot study of 10 patients with detectable ctDNA at commencement of therapy, including one MSI-positive colorectal cancer case, showed that complete ctDNA loss at the first treatment assessment (8 weeks) is associated with durable radiographic tumor response as well as long survival outcomes (22). An association between ctDNA and radiographic outcomes is also reported for 24 NSCLC cases (23). Our case showed 30% KRAS mutant-allele fraction in the baseline metastasis and only 0.1% at 2 weeks of treatment. The corresponding figures for ctDNA, 23% versus 2.4%, may reflect different methodologic sensitivity or tumor load at other metastatic sites. Nevertheless, plasma KRAS p.G12D was undetectable from the first radiographic response evaluation at 9 weeks onwards.
The conversion of the liver metastasis phenotype following initiation of the PD-1 inhibitor, as portrayed by the RNA sequence data, was perhaps the most intriguing observation. The interrelated phenotypes comprising CD274 (PD-L1) and JAK2 expression and sequences ascribed to intestinal epithelial neoplasia were attenuated, consistent with a rapid and substantial tumor-cell elimination. In contrast, expression of the chromosome 9p24.1 partner PDCD1LG2 (PD-L2) was augmented in the 2-week biopsy sample. It has been proposed that PD-L2 positivity of tumor cells and TME cells predicts treatment response to PD-1 blockade through T-cell reactivity (24, 25). Hence, the heightened PD-L2 expression in the on-treatment metastasis specimen might have reflected activation of tumor-targeting immunity.
The computational RNA sequence analysis further revealed enrichment of pembrolizumab-induced functions and signaling pathways with designations such as immune-cell trafficking, but also inflammation, acute-phase response, and complement system activity. The complement complex belongs to the rapidly responding innate immunity and exerts opposing effects within tumor-antigen presentation and T-cell–mediated cytotoxicity on the one side and TME activation of tumor-supportive systemic inflammation on the other (26). However, the data also supported pembrolizumab-induced activity of IFNγ-responsive genes involved in tumor-eliminating T-cell effects (17), and the patient has had no clinical or biochemical evidence of systemic inflammation during the treatment course. Expression of the IFNγ-responsive IDO1, encoding the essential indoleamine 2,3-dioxygenase-1 of various TME-cell types that participate in immune tolerance to tumor antigens (27), was repressed at 2 weeks of ICB. Noting the failure or termination of clinical trials adding an indoleamine 2,3-dioxygenase-1 inhibitor to PD-1 blockade, it may be that this particular combination is mechanistically redundant.
The RNA transcript profile also revealed a possible role of lipid metabolism and particularly LXR/RXR activity in the patient's metastatic cancer on ICB response. In mouse models, administration of an LXR agonist reduced the numbers of systemic and tumor-infiltrating myeloid-derived suppressor cells (an innate immune-cell population), which evoked T-cell cytotoxicity and restrained tumor growth; similar effects on peripheral blood myeloid-derived suppressor cells and cytotoxic T-cells were observed in patients (28). It has been known for almost three decades that LXR/RXR function as coregulators of the thyroid hormone receptor–mediated transcription (29). However, in our patient's case, it is not clear how the initial hyperthyroidism and the concurrent ICB-induced LXR/RXR activity in the metastatic cancer were mechanistically interrelated.
The patient described here, presenting to the MetAction study with metastatic colon cancer refractory to all standard therapies, has experienced durable, partial radiographic tumor regression following administration of pembrolizumab, based on the detection of 9p24.1-CNG in a liver metastasis. Our experience may support the use of PD-1 blockade in histology-agnostic cases with 9p24.1-CNG.
Request to inspect and analyze the data that underlie the results reported in this article, including the sequence datasets that are stored in the Services for Sensitive Data facility at University of Oslo (Oslo, Norway), should be directed to the corresponding author, and access will be provided in accordance with the General Data Protection Regulation of the European Union.
Disclosure of Potential Conflicts of Interest
D. Heinrich is a consultant/advisory board member for Merck Sharp & Dohme (MSD). E. Hovig has ownership interest in OncoImmunity AS and PubGene Inc. A.-L. Børresen-Dale is a consultant/advisory board member for Arctic Pharma AS, Saga Diagnostic AS, and PubGene AS and has ownership interest in Arctic Pharma AS. No potential conflicts of interest were disclosed by the other authors.
Conception and design: A.H. Ree, H.G. Russnes, D. Heinrich, E. Hovig, A.-L. Børresen-Dale, K. Flatmark, G.M. Mælandsmo
Development of methodology: H.G. Russnes, I.R. Bergheim, K. Beiske
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.H. Ree, H.G. Russnes, D. Heinrich, C. Johansen, K. Beiske, A. Negård
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.H. Ree, Vigdis Nygaard, H.G. Russnes, D. Heinrich, Vegard Nygaard, E. Hovig, K. Beiske, A.-L. Børresen-Dale, K. Flatmark
Writing, review, and/or revision of the manuscript: A.H. Ree, Vigdis Nygaard, H.G. Russnes, D. Heinrich, Vegard Nygaard, C. Johansen, I.R. Bergheim, E. Hovig, K. Beiske, A. Negård, A.-L. Børresen-Dale, K. Flatmark, G.M. Mælandsmo
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.H. Ree, Vigdis Nygaard, H.G. Russnes, Vegard Nygaard, C. Johansen, G.M. Mælandsmo
Study supervision: A.H. Ree, H.G. Russnes, K. Flatmark, G.M. Mælandsmo
The authors thank the Norwegian Cancer Genome Consortium for the opportunity to use their resources in the analysis and handling of DNA/RNA sequence data. The esthetic creation of the figure artwork by Ms. Dawn Patrick-Brown is highly appreciated. This work was supported by grants from the Research Council of Norway (218325) and South-Eastern Norway Regional Health Authority (2017109).