Metastatic clear-cell renal cell carcinoma: a frequent NOTCH1 mutation predictive of response to anti-NOTCH1 CB-103 treatment

Abstract

Background

Clear-cell renal cell carcinomas (ccRCCs) are malignant tumors with high metastatic potential and resistance to treatments occurs almost constantly. Compared to primary tumors, there are still limited genomic data that has been obtained from metastatic samples.

Methods

We aimed to characterize metastatic ccRCC by way of whole-genome analyses of metastatic formalin-fixed samples, using OncoScan^® technology. We identified a frequent, unexpected pL1575P NOTCH1 mutation which we set out to characterize for translational purposes. We thus implemented patient-derived xenografts from metastatic samples of human ccRCC to explore its clinical significance.

Results

We showed that pL1575P NOTCH1 mutation was an activating mutation, leading to the expression of NOTCH1-intracellular domain-active fragments in both cancer cells and tumor endothelial cells, suggesting a trans-differentiation of cancer cells into tumor micro-vessels. We demonstrated that this mutation could be used as a predictive biomarker of response to CB-103, a specific NOTCH1-intracellular domain inhibitor. One striking result was the considerable anti-angiogenic effect, coherent with the presence of NOTCH1 mutation in tumor micro-vessels.

Conclusions

We identified a frequent, unexpected pL1575P_c4724T_C NOTCH1 mutation as a new biomarker for ccRCC metastases, predictive of response to the CB103 NOTCH1-intracellular domain inhibitor.

To the Editor,

The incidence and deaths of kidney cancer worldwide, and thus mainly of clear-cell renal cell carcinoma (ccRCC), have doubled since three decades, particularly in the elderly population of countries with high social-demographic index [1]. Recent therapeutic advances have considerably improved the prognosis of metastatic ccRCCs. However, almost all patients develop resistance to these treatments [2]. Metastases may derive from minority clones in the heterogeneous primary tumor [3], and there are still limited genomic data obtained from metastases [4, 5].

Using OncoScan^® technology [6], we aimed to characterize metastatic ccRCCs by way of whole-genome analyses of metastatic formalin-fixed samples.

Four patients were included in the first analysis (Additional file 1: Table S1). We identified copy number alterations and loss of heterozygosity (LOH) (Additional file 1: Fig. S1A). We compared this data obtained from the recent meta-analysis we performed on ccRCCs (Additional file 1: Fig. S1B) [4] and found some abnormalities not previously described like 9q11.2 amplification (Additional file 1: Table S2). We also identified potential mutations (File ExcelS1). Using a threshold of 9 for probability score, we retrieved 48 mutations (Additional file 1: Table S3), several of them not yet described in ccRCC. We focused on the pL1575P_c4724T_C NOTCH1 mutation, located on chromosome 9q34.3, because it was present in 3 of the 4 metastases, and because the NOTCH pathway is a potential therapeutic target in ccRCC [7]. In addition, the 9q arm is lost in 75% of RCCs (Additional file 1: Fig. S1B) [4], and we found an allelic imbalance 9q34.3 cytoband for Patients 1 and 2 (Additional file 1: File ExcelS1).

Using digital-droplet PCR (ddPCR) and specific probes for the pL1575P_c4724T_C NOTCH1 mutation, we confirmed that it was present in the lymphoblastic acute leukemia T (LAL-T) positive control sample and in the 3 metastatic samples from Patients 1, 2 and 4 with high allele frequencies (Fig. 1A), but not in Patient 3. When we tested 9 additional ccRCC metastatic samples, we identified pL1575P NOTCH1 mutations in all samples, with a mean mutant allele frequency of 53.5% (Fig. 1B).

Fig. 1

ddPCR allelic discrimination for the pL1575P_c4724T_C NOTCH1 mutation, in 3 of the 4 metastatic samples processed via Oncoscan^® analysis (A), and in ten additional metastatic samples (B). NOTCH1-ICD-expressing cancer cells and tumor endothelial cells in human samples, NOTCH1-ICD staining is mainly nuclear in the 3 metastatic samples from Patients 1, 2 and 4 (C). D Panel B illustrates the laser-microdissection of a cancer cell expressing NOTCH1-ICD (red circle and arrow) and of a cancer cell not expressing NOTCH1-ICD (yellow circle and arrow), with a significant difference in terms of percentage of NOTCH1 mutant allele frequency. *** P < 0.001. E The left panel shows the NOTCH1-ICD staining of a tumor endothelial cell (black arrow). Some tumor endothelial cells can be seen to co-express NOTCH1-ICD (red) and anti-CD31 (green) on immunofluorescence staining (middle panel). The right panel shows that laser-microdissected tumor endothelial cells expressing NOTCH1-ICD harbor the pL1575P NOTCH1 mutation with an allelic frequency of 48%

Most NOTCH1 mutations occur in the HD and/or PEST domains (Additional file 1: Fig. S2A). The pL1575P NOTCH1 mutation is located in the HD-N domain and is responsible for NOTCH1 constitutive activation in LAL-T [8], through the release of the active NOTCH1 intracellular domain (NOTCH1-ICD) into the cytoplasm and its nucleus translocation (Additional file 1: Fig. S2B).

To identify NOTCH1-ICD, we performed immunostaining using an anti-NOTCH1 antibody (aa1755-1767-intracellular) which specifically recognizes an ICD epitope. We found predominant nuclear staining in the 3 metastatic samples from Patients 1, 2 and 4 (Fig. 1C), but not all cancer cells were stained. Using laser-microdissection, the pL1575P NOTCH1 mutation was mainly present in cancer cells expressing NOTCH1-ICD (57% vs.17%, P < 0.01, Fig. 1D).

Surprisingly, we found that some CD31-expressing endothelial cells, but not all, co-expressed NOTCH1-ICD (Fig. 1E). Using laser-micro-dissection, we also identified the NOTCH1 mutation with an allelic frequency of 48% (Fig. 1E), suggesting vascular mimicry [9].

Five patient-derived xenograft models obtained from ccRCC metastases were developed in our research unit (XRCC1 to XRCC5). We identified the NOTCH1 mutation in all five models at Passage 1 (P1). Except for the XRCC5 model, the allelic frequency significantly increased between P1 and P5 (Additional file 1: Fig. S3). We chose XRCC4 and XRCC5 because of the marked enrichment for NOTCH1 mutation in the XRCC4 xenograft (48% allelic mutation frequency at P5) and a much lower allele mutation frequency for the XRCC5 xenograft (19% at P5). XRCC4 and XRCC5 were obtained from patients responding to sunitinib (Additional file 1: Fig. S4), predicting response to sunitinib in the two models [10]. We treated them with two NOTCH1 inhibitors: LY411575, a γ-secretase inhibitor with an IC50 of 0.39 nM, and CB-103, a NOTCH1-ICD specific inhibitor (Additional file 1: Table S4 and Additional file 1: Fig. S2B). Using LY411575 administered daily by gavage, we did not observe any anti-tumor effect. In contrast, CB-103 mono-therapy induced a significant anti-tumor effect for both models, more marked with XRCC4 (Fig. 2A, Additional file 1: Table S5, Fig. S5A). Unexpectedly, like with sunitinib, there was also a strong induction of necrosis with CB-103 mono-therapy. The additive effect of the combined treatment was low in terms of tumor growth inhibition (Fig. 2B and Additional file 1: Fig. S5B), but tissue effects were much more marked than with sunitinib or CB-103 mono-therapy, in particular for necrotic area extent with very low cell viability. There was also a significant gradual decrease in micro-vessel density and proliferation (P < 0.01, Fig. 2C–D). CB-103 tumor growth inhibition was stronger with XRCC4 than with XRCC5 (Additional file 1: Table S5), coherent with a higher allelic mutation frequency in XRCC4. Finally, when we assessed NOTCH1-ICD expression and NOCTH1 allelic mutation frequency in tumors after treatment, we found a significant decrease for both markers. Model XRCC4 treated with CB-103 monotherapy, NOTCH1 allelic frequency decreased from 64 to 1%, and NOTCH1-ICD was no longer seen to be expressed on western blot (Fig. 2E, F).

Fig. 2

In vivo anti-tumor effect of CB-103 in a XRCC4 xenograft model. A CB-103 monotherapy, sunitinib monotherapy, or the combination of CB-103 and sunitinib significantly inhibit tunor growth after 30 days of treatment (n = 6 per treatment group). This is associated with a significant gradual increase in the percentage of necrotic areas (B), a significant decrease in microvessel density (C) and in cell proliferation (D). After 30 days of CB-103 monotherapy treatment, the pL1575P_c4724T_C NOTCH1 mutation and NOTCH1-ICD protein expression disappear on Western blot (E). (*P < 0.05, ***P < 0.001) (F) NOTCH1 mutant allele frequency on laser-micro dissected cancer cells after treatment with CB103 or not, using ddPCR

Using two different methods, we showed that the pL1575P_c4724T_C NOTCH1 activating mutation is frequent in metastatic ccRCCs, and this may be explained by a high sensitivity of Oncoscan^® technology using formalin-fixed samples [6, 11]. Like in LAL-T, we also evidenced the benefit of using CB-103 designed to block active forms of NOTCH1-ICD [12]. CB-103 is currently being evaluated in phase II clinical trials for various malignancies, with acceptable toxicity profile [13]. Since NOTCH signalling is closely linked to immune system regulation, further studies are required to determine if this NOTCH1 mutation is associated with response to immune checkpoint inhibitors, which would be of high translational relevance [14]. Our study opens the way to further development in metastatic ccRCC.