This study was approved by the institutional review board/independent ethics committee of The University of Texas MD Anderson Cancer Center and conducted in accordance with Good Clinical Practice guidelines, defined by the International Conference on Harmonisation. All patients provided written informed consent to participate based on the principles of the Declaration of Helsinki.
Eligible patients were aged ≥18 years with previously untreated locally advanced ccRCC without evidence of metastatic disease. All patients underwent an initial diagnostic biopsy of their renal lesion to confirm clear cell histology. Eligible patients had clinical stage cT2-T3b, N0, M0 tumours, with retroperitoneal lymph nodes ≤1 cm in size (considered clinical N0) and were candidates for partial or radical nephrectomy. Additional key inclusion criteria were measurable disease according to Response Evaluation Criteria in Solid Tumors version 1.1 and an Eastern Cooperative Oncology Group performance status score of 0 or 1. Key exclusion criteria included inability to undergo a baseline tumour biopsy; a clinical status indicating the need for immediate (within 6 weeks) surgery, regardless of whether neoadjuvant therapy was to be administered; autoimmune disease; or any current/prior use of an immunosuppressant (>10 mg daily prednisone equivalent). The full trial protocol is available in the supplementary note.
This Phase 2, open-label, single-arm study (516-002 trial; clinicaltrials.gov identifier: NCT03680521) included two sequential preoperative treatment segments (Fig. 1a). In the first segment, sitravatinib monotherapy was administered for 2 weeks; in the second segment, combination sitravatinib plus nivolumab treatment was administered for ≥4 weeks (maximum 6 weeks, allowing surgical scheduling flexibility). After completing neoadjuvant therapy, patients underwent a pre-surgical restaging scan, followed by planned resection, either partial or radical nephrectomy. Patients were subject to a 48-h preoperative hold of all study drugs prior to surgery. No study drug was administered post-surgery.
Sitravatinib was administered orally once daily (QD) at the starting dose of 120 mg. The modified Toxicity Probability Interval (mTPI) method was used to set rules on a dose de-escalation plan to monitor and limit toxicity of the starting dose of sitravatinib in the combination regimen in the neoadjuvant setting28. Nivolumab was administered at the recommended dose, 240 mg every 2 weeks as a 60-minute intravenous infusion. Dose delays and modifications for adverse events (AEs) were permitted for sitravatinib.
Endpoints and assessments
The primary study objective was to evaluate clinical activity of the combination regimen; the primary endpoint was objective response rate (ORR) defined as the percentage of patients achieving a radiographic complete response (CR) or partial response (PR), per RECIST v1.1, prior to surgery. With currently available treatments, the percentage of patients with a point in time ORR prior to surgery was assumed to be 5% (null hypothesis) and this rate was thus considered uninteresting. The target percentage of patients with a point in time objective response prior to surgery using sitravatinib and nivolumab was assumed to be 30% (alternative hypothesis). Secondary endpoints were safety; pharmacokinetics (PK) of sitravatinib; immune effects, including changes in programmed cell death–ligand 1 (PD-L1) expression; time-to-surgery; and disease-free survival (DFS). Exploratory endpoints included biomarker analyses of the effect of sitravatinib alone and in combination with nivolumab.
Imaging was used for disease assessments, with an allowable window of 4 weeks prior to first study treatment for screening/baseline evaluation and within 1 week of planned surgery for on-study disease assessment. Per protocol, patients would be followed for survival every 6 months from the last study visit for at least 3 years or until death, disease recurrence, or loss to follow-up; disease recurrence was based on off-study imaging assessments performed per standard-of-care for post-nephrectomy patients. AEs were graded using the National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0. Data were collected using the Medidata RAVE platform version 2017.2.2.
Baseline disease assessments were performed using computed tomography (CT), X-ray (radiography), or magnetic resonance imaging (MRI). Subsequent, on-study disease assessment included imaging of all known and suspected sites of disease identified in the preoperative setting (ie, CT or X-ray of the chest, CT, or MRI of the abdomen, and, if clinically indicated, whole body bone scan and CT with contrast or MRI of the brain and evaluation of any superficial lesions). The allowable window for imaging was 4 weeks prior to first study treatment for screening/baseline evaluation and within 1 week of planned surgery for on-study disease assessment. Imaging results were evaluated by the investigator to assess disease response per RECIST v1.1. Blood samples for PK evaluation were collected at specified timepoints prior to and following study treatment dosing. Safety assessments were conducted at the initiation of study treatment and at each clinic visit. Per protocol, patients would be followed for survival every 6 months from the last study visit for up to 3 years or more until death, disease recurrence, or loss to follow-up; disease recurrence will be based on imaging assessments performed off study per standard of care for patients post-nephrectomy.
Dose-limiting toxicities (DLTs) were assessed to guide adjustment of the starting dose of sitravatinib after 3 weeks for the first six treated patients or earlier if ≥2 patients were suspected of experiencing DLTs. The dose of sitravatinib was to be decreased if >2 of the first six patients experienced DLTs. DLTs were defined as non-haematological grade 4 AEs; non-haematological grade 3 AEs except (a) manageable nausea, vomiting, and diarrhoea persisting <2 h, (b) uncomplicated electrolyte abnormalities resolved within 72 h, (c) fatigue persisting <8 days, and d) amylase or lipase elevation not associated with pancreatitis; any toxicity that delayed surgery by >2 weeks.
Biomarker expression analyses were conducted on core needle biopsies taken at baseline (timepoint 1), during sitravatinib monotherapy at day 14 (timepoint 2), and on resected specimens at the time of surgery during sitravatinib plus nivolumab (timepoint 3). Blood samples for correlative studies were collected at screening, day 1, day 15, day 29, and day 43/surgery. Biomarker analyses included immunohistochemistry (IHC) for PD-L1 expression (using clone 28-8; cat# ab205921 [Abcam, Cambridge, UK] as used in the Agilent PharmDx system), multiplex immunofluorescence (IF) profiling, gene expression profiling using NanoString nCounter and HTG EdgeSeq technologies, and tissue and blood flow cytometry.
Patients were permitted to discontinue from study treatment or from the study at any time at their own request, or by the discretion of the Investigator or Sponsor for safety, behavioural reasons, or for significant protocol violations. Further discontinuation reasons included objective disease progression, global deterioration of health, adverse events (as detailed below), loss to follow-up, refusal for further treatment, study termination by Sponsor, and death.
Permanent discontinuation of sitravatinib was implemented in the following circumstances:
Grade 3 or 4 febrile neutropenia
Grade 4 thrombocytopenia of any duration
Grade 4 hypertension
Grade ≥3 palmar-plantar erythrodysaesthesia
Grade ≥3 haemorrhage
Grade ≥2 thrombotic events (including thrombosis, pulmonary embolism, myocardial infarction, cerebrovascular accident, and thromboembolic event)
Development of nephrotic syndrome
Grade ≥3 increased transaminase
Use of any radiation or earlier-than-planned surgery to manage cancer lesions
Increase in aspartate aminotransferase and/or alanine aminotransferase ≥3 × the upper limit of normal (ULN) and bilirubin ≥2 × ULN but without concurrent increases in alkaline phosphatase, that is not attributable to liver metastases or biliary obstruction.
Permanent discontinuation from the study was considered in the following circumstances:
If treatment with sitravatinib was withheld for ≥14 consecutive days
If after receiving sitravatinib, but prior to any nivolumab dosing, patients developed toxicities that prevented the first nivolumab administration
If significant hypertension recurred (this could also be addressed through medical management and/or dose reduction)
In the event of treatment-related, grade ≥2 decreased ejection fraction
For patients requiring acute hospitalisation for treatment of congestive heart failure.
Non-haematological toxicities of grade ≥3 and considered to be sitravatinib-related were managed with permanent discontinuation of sitravatinib. With the occurrence of grade 3 toxicities of nausea, vomiting, diarrhoea, and laboratory abnormalities that were adequately managed by routine supportive care (such as anti-emetics, anti-diarrhoeals, or electrolyte supplementation) and persisted for ≤72 hours, grade 3 fatigue lasting ≤8 days, or grade 3 amylase or lipase elevation, then sitravatinib treatment may be interrupted until resolution of toxicity to grade ≤1 or to baseline value and subsequently resumed at the same dose. Treatment with sitravatinib was discontinued in the presence of ≥2 g of proteinuria/24 h but could be restarted when protein levels decreased to <2 g/24 h.
Required dose modifications (i.e., interruption, dose reduction, or discontinuation) for nivolumab were performed per the current OPDIVO® US Prescribing Information (USPI OPDIVO [nivolumab])29, in addition to potential dose modifications for sitravatinib. Furthermore, patients permanently discontinued nivolumab in the presence of any grade 3 or 4 immune-related AEs; whereas sitravatinib could be resumed at the same or lower dose at the discretion of the Investigator until the event stabilised to grade ≤1.
The HTG EdgeSeq gene expression platform is a high throughput next-generation sequencing (NGS)-based assay that utilises low sample input and a unique nuclease protection chemistry to simultaneously assess gene expression levels in multiple genes. We used the 1392 gene HTG EdgeSeq Precision ImmunoOncology panel (PIP) to assess tumour immune response from a total of 37 patient samples.
Three batches of samples were run using the PIP; batch 1 samples comprised 16 formalin-fixed paraffin embedded (FFPE) samples with two 5 μm core needle biopsy sections, while batches 2 and 3 comprised 23 samples with three 5 μm core needle biopsy sections. Of the 37 samples, only two samples had a total surface area <6 mm2 (minimum required input into assay) and thus failed the sample input criterion.
All samples meeting the minimum input area were processed according to the manufacturer’s protocol. Briefly, the tissue was removed from the slides with a scalpel, lysed using proteinase K and HTG’s denaturation oil, and underwent target protection and clean up on the HTG EdgeSeq processor. Following this process, sequencing adaptors and barcodes were added and the library amplified. Amplified libraries were quantified with the Kapa library quantification kit, pooled, and sequenced on Illumina’s Miseq platform. The data were analysed using HTG parser and HTG EdgeSeq Reveal software, to generate raw and normalised counts. All samples passed QC2 metrics, and three samples failed QC1 (having <1.5 million reads post-sequencing).
HTG EdgeSeq read count data for 50 samples was collected and analysed for differential expression and gene set enrichment using a custom bioinformatics pipeline. Raw sample read counts were normalised to account for library size differences across samples and log2CPM expression gene values were calculated for each sample. Housekeeping (HK) gene expression was used to define a scale factor for each sample for further normalisation. This was achieved by dividing the mean expression of the housekeeping genes across samples by the mean of the house keeping genes within each sample.
Principal component analysis (PCA) analysis was performed on the normalised data to identify any possible outliers. Differential expression analysis was conducted in R (v 3.6.1) using the Limma package (v 3.40.9). Three comparisons were performed using three time points (baseline/time point 1; Day 14/time point 2; surgery/time point 3): time point 2 versus time point 1, time point 3 versus time point 1, and time point 3 versus time point 2. Custom visualisations highlighting differentially expressed genes for each comparison were generated using the ggplot2 and heatmap.2 libraries. Pathway enrichment analysis was performed using the PreRanked method from the Gene Set Enrichment Analysis (GSEA v 2.2.4) on the log2 fold change results from the differential expression comparisons. MSigDB (v7) was used to obtain the list of GeneSets and pathways for analysis.
Gene set enrichment analysis for the 3 comparison was performed using MSigDB (v 7.0) gene set collection30,31 and the GSEA software (v2.2.4) using the pre-ranked method and the permutations by gene_set parameter. GEO accession number GSE212525.
Multiplex immunofluorescence (IF) analyses
For multiplex IF analysis, the Opal chemistry, and multispectral microscopy Vectra/Polaris scanner system (Akoya Biosciences, Waltham, MA) were used; analysis was performed using the inForm software. A total of 21 cases were stained for Multiplex IF Panel 1, 2, 4, and 5 using similar methods to those previously described32,33. Briefly, 4 μm-thick FFPE samples from consecutive sections were stained using 20 biomarkers divided into multiplex IF panels against: Panel 4, CK, CD3, LAG3, TIM3, ICOS, VISTA, and OX40; and Panel 5, CK, CD68, Arg-1, CD11b, CD33, CD14 and CD66b. The multiplex IF panels were applied in 52 samples (different time points) per panel, three samples per panel were considered not eligible for image analysis. All the markers were stained in sequence according to each multiplex IF panel using their respective fluorophore contained in the Opal 7 kit (catalogue #NEL797001KT; Akoya Biosciences) and the individual tyramide signal amplification fluorophores Opal Polaris 480. The slides were scanned using the Vectra/Polaris 3·0·3 (Akoya Biosciences) at low magnification, 10x (1.0 µm/pixel) through the full emission spectrum and using positive tonsil controls to calibrate the spectral image scanner protocol32. A pathologist selected a median of five regions of interest (ROIs) for scanning in high magnification using the Phenochart Software image viewer 1.0.12 (660 × 500 µm size at resolution 20×) in order to capture various elements of tissue heterogeneity. Each ROI was analysed by a pathologist using InForm 2.4.8 image analysis software (Akoya Biosciences). Marker co-localisation was used to identify specific cell phenotypes in each multiplex IF panel. Densities of each cell phenotype per panel were quantified, and the final data were expressed as number of cells/mm2. All data were consolidated using the R studio 3.5.3 (Phenopter 0.2.2 packet, Akoya Biosciences).
Additional multiplexed IF staining on a subset of available tissue samples for tumour and immune cell markers was also performed in FFPE tumour samples at screening, Day 14, and surgery using NeoGenomics MultiOmyxTM technology, including the NeoLYTX v2.0 software. This technology evaluates the expression of a panel of 19 biomarkers, including arginase 1, CD3, CD4, CD8, CD11b, CD14, CD15, CD16, CD33, CD56, CD68, CD163, CTLA4, FOXP3, HLA-DR, Ki67, PD1, PDL1, and tumour segmentation markers PanCK or CA9. Staining was performed using a single 4 μM FFPE slide. Within each staining round, two cyanine dye-labelled (Cy3, Cy5) antibodies were paired and recognized two markers. The staining signal was then imaged and followed by novel dye inactivation, enabling repeated rounds of staining. Proprietary deep learning-based algorithms were applied to identify and classify individual cells associated with each marker. Both tumour segmentation IF markers and pathologist-defined tumour areas based on haematoxylin and eosin slides were used to define areas of analysis for cell classification within tumours. These results were combined to generate co-expression summaries and compute spatial distribution statistics for phenotypes of interest.
IHC assay (PDL-1, Clone 28-8)
Staining of tumour tissue for PD-L1 was conducted in FFPE sections from tissue obtained at the three pre-determined time points (baseline/time point 1; Day 14/time point 2; surgery/time point 3) using automated immunostaining. A total of 46 samples from 21 patients were stained using clone 28-8 (cat# ab205921; Abcam).
The immunohistochemistry protocol is briefly described here: tissue sections (4 μm) were stained in a Leica Bond Max automated stainer (Leica Biosystems, Vista, CA); the tissue sections were deparaffinised and rehydrated following the Leica Bond protocol. Antigen retrieval was performed for 20 min with Bond solution #2 (Leica Biosystems, equivalent EDTA, pH 9·0). The primary antibody (PDL-1, clone 28-8 [Abcam], dilution 1:100) was incubated for 15 min at room temperature and detected using the Bond polymer refine detection kit (Leica biosystems) with DAB as chromogen, the slides were counterstained with haematoxylin, dehydrated, and cover slipped.
Analysis of the expression of PD-L1, was performed by a pathologist using a standard microscope approach. PD-L1 was evaluated in viable malignant cells and reported as percentage of malignant cells with any positive membrane expression, <1% was determined as PD-L1 negative; ≥1% was determined as PD-L1 positive. Samples with fewer than 100 malignant cells were considered inadequate for PD-L1 analysis. The optimal cut-off for identifying PD-L1 positivity in ccRCC has not been identified and is not used in clinical decision-making23. Linear regression analysis on PD-L1 change from baseline to time of surgery and change in sum of target lesions showed no association (p = 0.23).
Flow cytometry of freshly disaggregated tumour tissue
Fresh tissue from 80 samples (n = 24 patients) underwent flow cytometry analysis. In some cases, normal kidney samples were also collected at the time of surgery and stained. Fresh tissue was mechanically disaggregated using a BD Medimachine System (BD Biosciences) and was subsequently filtered to generate a single cell suspension prior to staining. The sample was processed and stained within 24 h of collection. Surface staining was performed in FACS Wash Buffer (1× DPBS with 1% BSA) for 30 min on ice using fluorochrome-conjugated monoclonal antibodies from BD Biosciences, BioLegend, and Life Technologies. Cells were then fixed in 1% paraformaldehyde solution for 20 min at room temperature. For panels containing transcription factors, cells were fixed and permeabilised using the BD Transcription factor kit according to the manufacturer’s instructions. Samples were acquired using a BD Fortessa X20 and analysed using FlowJo Software v 10.7.1 (Tree Star). Dead cells were stained using AQUA live/dead dye (Invitrogen) and excluded from the analysis. Single colour controls were used to generate and adjust compensation matrices, and fluorescence minus one (FMO) controls were used to set positive gates for markers where the negative and positive populations do not clearly separate. Subgating is only performed when more than 100 events are present in the parental population as a QC control. Supplementary Table S3 shows the flow cytometry panel design and the associated gating strategy is in Supplementary Fig. S7.
Flow cytometry of blood
Flow cytometry analysis was conducted retrospectively on cryopreserved peripheral blood mononuclear cells (PBMCs). Prior to use, PBMCs were stored in liquid nitrogen in 1 mL aliquots. PBMCs from 95 samples were analysed. All samples were stained and acquired at the same time to avoid any technical variation. Prior to staining, PBMCs were thawed, washed, and resuspended in FACS wash buffer (1× DPBS with 1% BSA). Flow cytometry staining and sample QC was performed as described in the previous sections.
It was planned that the study would enroll 25 patients, anticipating that 18 of these would be evaluable for clinical activity. A one-sided (alpha 0.025) Exact test for single proportion was used to test the hypothesis of whether the percentage of patients with ccRCC achieving a point-in-time objective response prior to surgery was ≤5% against alternative hypothesis that the objective response was >5%; the corresponding 95% confidence intervals (CIs) were calculated using the exact Clopper-Pearson method. An efficacy-evaluable population of 18 patients would provide 80% power with a two-sided Type 1 error of 0.05 (equivalent to one-sided alpha 0.025) to demonstrate a significant difference between the target response rate (30%) versus background (5%). Event time was censored on the date of surgery, or date of last follow-up assessment documenting absence of recurrence or death, whichever occurred later for patients who were alive and disease-free. No interim analysis was planned.
The efficacy analysis included patients who received ≥1 dose of each study drug and underwent on-study disease assessment prior to surgery. The safety analysis included all patients who received ≥1 dose of either sitravatinib or nivolumab. The PK evaluable population comprised all patients who received sitravatinib and had non-missing concentration-time data. Correlative studies were performed using samples from the enrolled, safety, or efficacy populations, according to the specific analysis being conducted; the number of samples used in each analysis varied according to specimen availability.
The primary endpoint of proportion of patients achieving an objective response (CR or PR) was summarised. Information regarding pathologic CR was summarised descriptively. DFS was described using the Kaplan-Meier method. Median follow-up for DFS was calculated using the reverse Kaplan–Meier method. Median extent of follow-up was calculated based on descriptive statistics from first dose to last date known alive.
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.