Ex vivo tissue modelling informs drug selection for rare cancers

The identification and therapeutic targeting of actionable gene mutations across many cancer types has resulted in improved response rates in a minority of patients. The identification of actionable mutations is usually not sufficient to ensure complete nor durable responses, and in rare cancers, where no therapeutic standard of care exists, precision medicine indications are often based on pan‐cancer data. The inclusion of functional data, however, can provide evidence of oncogene dependence and guide treatment selection based on tumour genetic data. We applied an ex vivo cancer explant modelling approach, that can be embedded in routine clinical care and allows for pathological review within 10 days of tissue collection. We now report that ex vivo tissue modelling provided accurate longitudinal response data in a patient with BRAFV600E‐mutant papillary thyroid tumour with squamous differentiation. The ex vivo model guided treatment selection for this patient and confirmed treatment resistance when the patient's disease progressed after 8 months of treatment.


| INTRODUCTION
The identification of molecular markers to predict treatment response and guide treatment selection, independently of tumour histology, has garnered significant interest in oncology.Currently there are over 30 molecular-targeted drugs approved for oncology indications, including melanoma, chronic myeloid leukaemia, non-small cell lung cancer and breast cancer.However, only 15% of cancer patients are eligible to receive these therapies and even fewer patients will achieve durable benefits from these genome-informed treatments. 1 In basket clinical trials, such as NCI-MATCH 2 and SHIVA, 3 where tumour-specific, actionable mutations direct treatment selection, the results have been disappointing.In the randomised Phase 2 SHIVA trial, there was no difference in the progression-free survival of patients treated with selected targeted therapies versus physician's choice (2.3 months versus 2.0 months, P = .41).Similarly, the NCI-MATCH genome-based treatment approach resulted in response rates of <5% for most molecular therapies. 4e results from precision-medicine trials are not unexpected as it is well-established that responses to therapies targeting driver oncogene mutations are not consistent across patients or cancer types.For instance, response of BRAF V600 -mutant cancer patients to combination BRAF and MEK inhibitors varies from 70% in melanoma and non-small cell lung cancer (NSCLC) to 12% in colorectal cancer. 5,6Further, tumour agnostic trials focussed on actionable mutations, such as the phase I trial investigating the KRAS G12C selective inhibitor Sotarasib, have demonstrated response rates ranging from 7.1% for colorectal cancer to 32.2% for NSCLC. 7,8As a result, there is growing interest in expanding precision medicine to incorporate both genomic and functional data. 9 this study, a functional precision cancer model was developed for a rare BRAF V600-mutant squamous cell carcinoma of the thyroid, where indications are often based on pan-cancer data.We applied an ex vivo cancer explant modelling approach, that limited tissue handling, maintained tissue integrity and enabled pathological review within 10 days of tissue collection.This ex vivo tissue model provided accurate patient response data to combination BRAF/MEK inhibition and longitudinal explant functional assessment matched patient responses.We highlight the value of patient-derived tissue explant functional models that can be embedded into clinical trials and clinical practice to support genomic data in guiding treatment selection.

| Patient response assessment and tumour biopsies
Tumour response was evaluated 8 weeks post systemic therapy using the Response Evaluation Criteria in Solid Tumours (RECIST) 1.1 Criteria.Tumour biopsy samples were obtained at time of surgical resection prior to systemic therapy and at time of disease progression.
Tumour was processed for haematoxylin and eosin (H&E) staining and immunohistochemical analysis, DNA was extracted for molecular analysis (targeted next generation sequencing panel) and generation of explant models for rapid drug sensitivity testing.

| Ex vivo explant preparation
Tumour tissue was either sliced at a thickness of 250 μM using a vibratome (Leica VT 1200S, Leica Microsystems) or manually dissected into <1 mm 3 fragments within 24 h after surgery.Tissue fragments were then placed on Millicell cell culture membrane inserts (0.4 μm pore size, 30 mm diameter, Merk PICMURG50) in 6-well plates containing tissue culture media (1Â RPMI 1640, 4 mM L-Glutamine, 25 mM HEPES pH 7.0, 10% human serum, 100 U Penicillin/0.1 mg/ml Streptomycin and 50 μg/ml Gentamicin).Tissue slices were treated with the combination of 10 nM MEK inhibitor trametinib (GSK1120212) and 100 nM BRAF inhibitor dabrafenib (GSK2118436), and 0.1% DMSO as control, incubated for 72 h at 37 C in 5% CO 2 .These drug dosages were based on prior ex vivo analyses conducted on 3D organoids, 10 and differentiated treatment resistance and sensitivity in patient derived short-term melanoma cultures. 11,12Media was changed once midway with fresh media and newly prepared dilution of compounds.Drugs were dissolved in DMSO and diluted in tissue culture media.After 72 h of treatment, slices were collected and fixed in 10% neutral buffered formalin solution for 24 h at 4 C, then washed with PBS, dehydrated and embedded in paraffin before performing Haematoxylin and Eosin staining for histology as per standard clinical diagnostic practices.

| Immunohistochemical analyses
Immunohistochemistry was performed on 4 μm thick tissue sections using the heat-mediated antigen retrieval technique.Commercially available antibodies to TTF-1 (clone SPT24, Leica) and pan-TRK (clone A7H6R, Cell Signalling, Danvers USA) were performed using the Lecia Bond III automated staining platform.Antibodies to PAX 8 (clone MRQ-50, Roche), P40 (clone BC28, Biocare Medical), and BRAF V600E (clone VE1, Abcam) were also performed using VENTANA BenchMark ULTRA automated staining platform, following standardised protocol provided by the manufacturer.Tumour cells demonstrating nuclear staining for TTF1, P40, PAX8 and pan-TRK, and cytoplasmic staining for BRAF V600E were considered positive.

| Oncomine precision assay
H&E stained FFPE tissue samples were assessed by a qualified anatomical pathologist.Tissue specimens with >20% neoplastic tissue were extracted using the QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany, Cat.No.56404) and DNA quantification performed using Qubit dsDNA HS (High Sensitivity) Assay Kit (Thermo Fisher Scientific, MA).DNA libraries were prepared for sequencing using the Oncomine Precision Assay (Thermo Fisher Scientific, A46291) and next generation sequencing) was performed with the Ion Torrent Genexus Integrated Sequencer using the Ion Torrent GX5 Chip, and the genes tested have been summarised in Table S1.Sequencing was performed to a mean sequence coverage of >600Â for DNA targets, with DNA on-target reads >90% and uniformity of base coverage >96%.Limit of detection was 5% variant allele fraction (VAF) at a coverage of 400Â.Within the reportable range, sensitivity is >95% for single nucleotide variants, >95% for small indels <20 bp.Variants are analysed, filtered and annotated using Ion Reporter software.The sequencing coverage and quality statistics for each sample are summarized in Table S2.

| Patient details
Longitudinal tumour specimens were collected from a 60-year-old female who presented with a 1-month history of a rapidly increasing mass in the right neck.Fine needle aspiration of her thyroid demonstrated papillary thyroid carcinoma (Bethesda Category VI).The patient underwent a total thyroidectomy with right modified radical neck dissection.Histopathologic examination of the thyroid demonstrated papillary carcinoma with areas of dedifferentiation resembling squamous cell carcinoma (Figure 1A).The squamoid areas stained positive for the p40 squamous marker (Figure 1B).Both areas showed strong diffuse immunostaining with thyroid transcription factor (TTF-1; Figure 1C), PAX 8 (data not shown) and BRAF V600E (Figure 1D).Immunohistochemistry and fluorescent in situ hybridisation for NTRK and RET gene rearrangements were negative.Metastatic carcinoma was present in 10/30 lymph nodes identified in the accompanying neck dissection.The metastatic carcinoma showed a spectrum of morphologic appearances including conventional papillary carcinoma of thyroid with dedifferentiation to squamous appearance.
The largest metastatic deposit measured 65 mm with extensive extranodal extension, in keeping with clinically observed rapid enlargement.
There was no history of cutaneous or mucosal squamous cell carcinoma.

| Modelling tumour response to combination BRAF/MEK inhibition
A fresh tissue sample derived from the radical neck dissection, prior to systemic therapy, was sectioned using a vibratome Six 250 μM tissue sections were generated and 2 Â 250 μM tissue sections were left treated, treated with vehicle control or treated with combination dabrafenib/trametinib for 72 h.To incorporate functional analysis within routine clinical care, tissue sections were formalin fixed and histologic examination was performed <10 days after resection.Sections were independently reviewed by two pathologists to assess tumour cell viability.As shown in Figure 2, control-treated explant slices demonstrated viable, mitotically active carcinoma with similar morphologic appearance to carcinoma with squamous differentiation as observed in the parotidectomy specimen (tumour cellularity 80%, tumour viability 90%, Figure 2A).In contrast, the explant slices treated with dabrafenib and trametinib demonstrated occasional ghost outlines of devitalised tumour cells.Viable tumour cells were not seen in the 2-3 tissue fragments treated with combination dabrafenib/trametinib (tumour cellularity 0%, viability 0%, Figure 2B).There was focal early granulation tissue formation and aggregation of lymphoid cells (Figure 2C).

| Patient response to combination BRAF/MEK inhibitor therapy
Within 2 months of surgical resection, the patient had further recurrence.A CT scan of the neck (Figure 3A) demonstrated a 3 cm mass in the region of the previously excised level 3 lymph node.Further surgery or radiotherapy was deemed not possible, and the patient was considered for palliative systemic therapy.Based on the explant analysis, the patient was commenced on combination of dabrafenib and trametinib following institution approval.The patient response correlated with the ex vivo response profile, showing a dramatic early response to therapy, with complete clinical response within 2 weeks from commencing therapy and progress CT at two 2 months demonstrating complete response (Figure 3A).staining for BRAF V600E on immunohistochemistry (Figure 3D).

| Post BRAF/MEK inhibitor progression response
Ex vivo explant testing on this new neck tumour mass was performed to determine BRAF and MEK inhibitor sensitivity.Vehicle controltreated explant sections demonstrated abundant viable tumour cells (Figure 3E).However, unlike the previous parotid explant tumour tissue, the right neck mass recurrence did not show any response following 96 h of treatment with combination of 100 nM dabrafenib/10 nM trametinib (80% tumour cellularity, 90% viability, Figure 3F).
Next generation sequencing was performed on baseline, recurrent and resistant biopsies with the Ion Torrent Genexus Integrated Sequencer using the Ion Torrent GX5 Chip.This confirmed the presence of a BRAF V600E (c.1799 T > A) mutation detected at a VAF of 6.3% in the baseline specimen and 29.1% in the tissue collected prior to systemic BRAF/MEK inhibitor therapy.Sequencing of the BRAF/ MEK inhibitor resistant specimen identified an HRAS Q61R (c.182A > G) mutation detected at a VAF of 3.1%, which was not detected on the baseline specimen, suggesting a mechanism of acquired resistance to BRAF/MEK inhibitors in this rare tumour.

| DISCUSSION
Approximately 5% of oncology drugs undergoing first-in-human trials result in Food and Drug Administration registration, with the most common cause being lack of efficacy followed by toxicity. 13This is especially problematic in rare cancer types where robust clinical trial data cannot be collected, further impacting drug approval.Biomarkerdriven treatment selection such as the identification of actional mutations via molecular testing has had variable results.For example, the first generation NTRK inhibitor, Larotrectinib, has an overall response rate of 79% across multiple tumour types.However, response rates varied according to tumour type, with response observed in 43% of melanoma (n = 7) and 96% of infantile fibrosarcoma (n = 28) patients.
Further, the poor outcomes from precision-medicine basket clinical trials that rely solely on genomic data necessitate a new approach to guide treatment decisions.We propose that a rapid, tissue-based assay, testing a select number of clinically available therapies, can be embedded into routine clinical practice to provide functional evidence of treatment response.
In this study, a patient presenting with a rare BRAF V600 -mutant squamous cell carcinoma of the thyroid, benefited from functional evaluation of ex vivo tumour response to combination BRAF/MEK inhibitors.Oncogenic BRAF V600E alterations are common (29%-83%) in thyroid malignancies, with over 50% response rates to combination BRAF and MEK inhibitors in both papillary carcinoma and anaplastic carcinoma of thyroid. 15,16However, squamous dedifferentiation is extremely rare with <100 cases reported in the literature and is associated with a poor prognosis. 17Established treatment protocols and data regarding the response of these rare tumours to systemic therapy are lacking.The addition of functional response assays to tumour genetic data is particularly valuable in prioritising and selecting appropriate therapy for patients with metastatic rare cancers.This report demonstrates the utility of an appropriately constructed ex vivo patient derived tumour explant in optimising therapeutic choices in a timely manner.
There are limitations to this ex vivo tissue approach and these include the small number of selected treatments that can be tested concurrently (up to six drugs depending on size and quality of tissue received) and the fact that only one metastatic lesion is usually examined which will not encompass heterogeneity nor clonal evolution.Furthermore, due to tissue requirements, this approach may only be

Following 8
months of treatment, the patient developed pain and numbness in the right arm.MRI and PET/CT showed a poorly defined soft tissue mass in the right neck extending from the paraspinal tissue to anterior scalene muscles (Figure 3B).Tissue biopsy confirmed that the treatment-resistant right neck mass was morphologically similar to the baseline specimen (Figure 3C) but with only patchy positive F I G U R E 1 Histological examination of the parotid mass excised during initial surgery prior to systemic therapy.(A) Infiltrative tumour with squamous differentiation on haematoxylin and eosin-stained sections (scale shown in μm).(B) Squamous phenotype was confirmed with P40 immunohistochemistry stain (scale shown in μm).(C) Tumour cells were positive for TTF-1 supporting thyroid origin (scale shown in μm).(D) The histological appearance, immunophenotype together with the diffuse BRAF V600E stain confirmed metastatic disease from the thyroid carcinoma (scale shown in μm).

F I G U R E 2
Ex vivo modelling of BRAF/MEK inhibitor response in BRAF V600E -mutant thyroid squamous cell carcinoma excised during initial surgery prior to systemic therapy.(A) The parotid explant control-treated tissue sections included malignant squamous cells showing brisk mitotic activity (arrows), with cellularity of 80% and viability of 90% (Haematoxylin and Eosin, scale shown in μm).(B) Following 72 h treatment with dabrafenib/trametinib, the explant tissue showed no viable tumour cells (Haematoxylin and Eosin, scale shown in μm).(C) Focal early granulation tissue formation with occasional ghost outlines of degenerate tumour cells and lymphoid aggregate were observed post dabrafenib/trametinib treatment (Haematoxylin and Eosin, scale shown in μm).possible at time of routine surgical resection or with an invasive core biopsy.Nevertheless, functional testing is important addition to current genomic testing, and can be incorporate as a routine part of personalised cancer therapy work-up and therapeutic decision making to enhance accuracy.F I G U R E 3 Ex vivo modelling of biopsy obtained from right neck recurrence following 8 months of combination dabrafenib and trametinib confirms BRAF/MEK inhibitor resistance.(A) CT scans demonstrating a necrotic right neck node at baseline (left) followed by complete response post dabrafenib and trametinib therapy (right).(B) New poorly defined enhancing soft tissue lesion in the right side of the neck, extending from the paraspinal tissue to the scalene muscles on MRI.This new lesion was biopsied for the second explant testing.(C) Core biopsy of the right neck recurrence demonstrated similar histological features to the original primary and metastatic tumours (Haematoxylin and Eosin, scale shown in μm).(D) In contrast to the primary tumour, the BRAF V600E immunohistochemistry stain showed heterogeneous staining with both positive (bottom right) and negative (top left) tumour cells (Haematoxylin and Eosin, scale shown in μm).(E) The control-treated explant tissue demonstrated tumour cells with overall 80% tumour cellularity and 90% viability (Haematoxylin and Eosin, scale shown in μm).(F) Following treatment with dabrafenib, there was no significant tumour response in the ex vivo model, with comparable tumour viability to the controltreated tissue specimen (Haematoxylin and Eosin, scale shown in μm).