Identifying patients eligible for PARP inhibitor treatment: from NGS-based tests to 3D functional assays

Abstract

Within the past few years, poly (ADP-ribose) polymerase inhibitors (PARPi) have been added to the standard of care for cancer patients, mainly for those exhibiting specific genomic alterations in the homologous recombination (HR) pathway. Until now, patients who are eligible to receive PARPi have been identified using next-generation sequencing (NGS) of gene panels. However, NGS analyses do have some limitations, with a subset of patients with negative NGS-based results can exhibit a clinical benefit, responding positively to PARPi, despite the failure to detect dynamic and predictive biomarkers such as mutated BRCA1/2 genes. Furthermore, the sequencing of initial tumour does not allow to detect reversions or secondary mutations that can restore proficient HR and lead to PARPi resistance. Therefore, it is crucial to better identify patients who are likely to benefit from PARPi treatment. In this context, tumour models such as patient-derived xenografts or tumour-derived organoids could help to guide clinicians in their decision making as these models accurately mimic phenotypic and genetic tumour heterogeneity, and could reflect treatment response in an integrative manner. In this Perspective article, we provide an overview of the currently available NGS-based tests that enable the identification of patients who might benefit from PARPi, and outline breakthroughs and discoveries to expand this selection using 3D functional assays. Combining NGS with functional assays could facilitate the efficient identification of patients, thereby improving patient survival.

Background

Since the approval in 2014 of olaparib, the first-in-class inhibitor of poly (ADP-ribose) polymerase (PARP), PARP inhibitors (PARPi) have shown efficacy and acceptable tolerability in adult patients with several types of solid tumour, particularly those harbouring pathogenic mutations in BRCA1 and/or BRCA2 genes, leading to cytotoxic accumulation of unrepaired DNA double-strand breaks (DSBs) induced by PARPi as well as trapped PARP–DNA complexes. Beyond BRCA mutations, PARPi are also active in tumour cells exhibiting alterations in other genes involved in homologous recombination (HR) DSBs repair pathway¹ or others HRD mechanisms like BRCA1/RAD51C promoter methylation.

To date, four PARPi—olaparib, niraparib, rucaparib and talazoparib—mainly targeting PARP1 and PARP2 proteins have been approved in the USA and Europe for the treatment of ovarian,^2–7 breast,^8,9 pancreatic¹⁰ and prostate¹¹ cancers with recurrent, locally advanced or metastatic disease. The identification of patients who are eligible for treatment with PARPi is currently driven by the use of next-generation sequencing (NGS) of targeted gene panels or genomic signatures (genomic scares) that are specific to healthcare institutions or through FDA-approved in vitro companion diagnostic tests (CDx). However, CDx have some limitations by failing to identify a HRD status up to 18% of cases for ovarian cancers and need some improvement in terms of specificity or sensitivity parameters to better select patients who might (not) likely to benefit to PARPi treatment.¹²

In this Perspective, we begin by outlining the CDx that are available, sample processing, interpretation, and the clinical success and challenges that are associated with CDx testing. We then outline promising functional assays that could potentially help clinicians to assess PARPi sensitivity through dedicated new clinical trials.

Next-generation sequencing: the current gold standard for patient selection

Various in vitro NGS-based tests are currently used to help clinicians to identify patients with cancer who might be eligible for treatment with PARPi therapies (Table 1). These tests aim to detect somatic (not inherited) as well as germline (inherited) gene alterations, such as well-known BRCA mutations, in patient samples. Some of these tests are based on the sequencing of home-designed gene panels related to DNA repair pathways, although no international consensus specifying the best way to define the HRD genotype exists yet. Five CDx tests approved by FDA are currently commercially available: BRACAnalysis, MyChoice HRD (both from Myriad Genetic Laboratories, Inc., Salt Lake City, UT) FoundationOne, FoundationFocus CDxBRCA Assay and FoundationOne Liquid CDx (from Foundation Medicine, Inc., Cambridge, MA).^13–17 The discussion below focuses on these approved tests.

Table 1.

Current approvals of PARP inhibitors by the FDA.

PARPi	Indications	Usage	Setting	Prior therapy	NGS-based testing
Niraparib	Epithelial ovarian, fallopian tube, or primary peritoneal cancer	Recurrent cancer	Maintenance treatment	CR/PR to first-line platinum-based chemotherapy	None
		Advanced cancer	First-line maintenance treatment	CR/PR to first-line platinum-based chemotherapy	None
		Advanced cancer	Treatment	≥3 prior lines of chemotherapy with HRD positive status	Myriad myChoice® CDx
Olaparib	Epithelial ovarian, fallopian tube, or primary peritoneal cancer	Recurrent cancer	Maintenance treatment	CR/PR to platinum-based chemotherapy	None
		Advanced cancer	First-line maintenance treatment	CR/PR to first-line platinum-based chemotherapy with BRCA mutations	BRACAnalysis CDx Myriad myChoice® CDx FoundationOne CDx
		Advanced cancer	First-line maintenance treatment	CR/PR to first-line platinum-based chemotherapy with HRD positive status and/or genomic instability, in combination with bevacizumab	BRACAnalysis CDx Myriad myChoice® CDx FoundationOne CDx
		Advanced cancer	Treatment	≥3 prior lines of chemotherapy with gBRCA mutations	BRACAnalysis CDx Myriad myChoice® CDx FoundationOne CDx
	HER2-negative breast cancer	Metastatic cancer	Treatment	Neoadjuvant, adjuvant, or metastatic setting with gBRCA mutations	BRACAnalysis CDx
	Pancreatic adenocarcinoma	Metastatic cancer	First-line maintenance treatment	No disease progression after 16 weeks of a first-line platinum-based chemotherapy with gBRCA mutations	BRACAnalysis CDx
	Prostate cancer	Metastatic castration-resistant cancer	Treatment	Progression following prior treatment (enzalutamide or abiraterone) with HRD positive status	BRACAnalysis CDx FoundationOne CDx FoundationOne® Liquid CDx
Rucaparib	Epithelial ovarian, fallopian tube, or primary peritoneal cancer	Recurrent cancer	Maintenance treatment	CR/PR to platinum-based chemotherapy	None
		None	Treatment	≥2 prior lines of chemotherapy with BRCA mutations	BRACAnalysis CDx FoundationOne CDx FoundationFocus CDxBRCA Assay FoundationOne® Liquid CDx
	Prostate cancer	Metastatic castration-resistant cancer	Treatment	After prior treatment (androgen and taxane-based therapies) with BRCA mutations	FoundationOne® Liquid CDx
Talazoparib	HER2-negative breast cancer	Locally advanced or metastatic cancer	Treatment	Neoadjuvant, adjuvant, or metastatic setting with gBRCA mutations	BRACAnalysis CDx

CR/PR complete/partial response, gBRCA germline BRCA (pathogenic or likely) mutations, HER-2 human epidermal growth factor receptor-2, HRD homologous recombination deficiency, NGS next-generation sequencing.

Sample processing

CDx used for the assessment of germline and/or somatic alterations in genes involved in HR DNA repair pathway are based on similar workflows (Fig. 1), although some differences exist in the nature of the biological samples used (whole blood for germline mutations; DNA from frozen or formalin-fixed paraffin-embedded tumours for somatic mutations), in preanalytical process (verification of cancer cells proportion in the tumour samples for somatic mutations) as well as in sequencing procedures (a higher coverage is required for the detection of variants with low allelic fraction for somatic alterations). Specific bioinformatic pipelines are used in each context. At least 200 ng of DNA is extracted, fragmented by sonication or enzymatic methods to 200 base pairs (bp) and ligated to a specific barcoded adapter allowing samples multiplexing before sequencing. After adapter ligation, an enrichment step using either PCR or capture allows the sequencing of regions of interest (e.g. coding exon and non-coding intron sequences) and large rearrangements (e.g. deletions and duplications). Finally, bioinformatic pipelines including different steps (demultiplexing, alignment, variant calling, annotation and filtering) are used. Then, an objective classification of variants is applied for medical interpretation. For instance, according the ACGM classification, variants can be classified as: pathogenic (class 5), likely pathogenic (class 4), variant of uncertain significance (class 3), likely benign (class 2) and benign (class 1).¹⁸

Fig. 1. Next-generation sequencing workflow based on hybrid capture.

Specimens are collected and prepared following standard pathology practices (a). Eligible specimens with sufficient nucleated cells (% of tumour cells, surface) were queued for DNA extraction and quantification of double-stranded DNA. A step of fragmentation is need before ligation with indexed adapters to construct a library for sequencing. Sequence data are analysed using software dedicated to germline or somatic variants (b). All sequence data are mapped to the human genome in order to identify genomic regions of interest through a variant calling step. Finally, all relevant genomic alterations are annotated to provide a clinical identification of eligible patients to PARPi therapies. CTC circulating tumour cell, ctDNA circulating tumour DNA, FFPE formalin fixed paraffin embedded, PBMC peripheral blood mononuclear cell.

Interpretation

All five approved CDx can detect BRCA1 and BRCA2 variants that are, or are likely to be, pathogenic (Fig. 2 and Supplementary Table 1), but it is also possible to identify other genomic signatures associated with genomic instability using these CDx. The HRD positive status could be defined if a (suspected) deleterious mutation in BRCA genes was identified and/or a positive genomic instability score (GIS) ≥ 42 (for Myriad myChoice CDx). GIS was defined by assessing the unweighted sum of three individual, validated markers of genomic instability, such as loss of heterozygosity (LOH), telomeric allelic imbalance (TAI) and large-scale state transitions (LST). For example, epithelial ovarian cancers exhibit genomic instability, and approximately half of them show HRD status, thus conferring impaired DNA repair and PARPi sensitivity.^19–21 Finally, pathogenic (class 5) or likely pathogenic (class 4) variants will be reported to the patient’s physician in less than 2 weeks.

Fig. 2. Detection of gene impairments using companion diagnostic (CDx) tests.

Somatic (not inherited) as well as germline (inherited) BRCA (or other) alterations could be reported. Positive homologous recombination deficiency status can be defined with a positive genomic instability score (for myCHoice CDx) or high loss of heterozygosity (for FoundationOne CDx) associated (but not necessarily) with a (suspected) deleterious mutation in BRCA genes. BRCA BReast CAncer gene, CNA copy number alterations, HRD homologous recombination deficiency, LOH loss of heterozygosity, LST large-scale state transitions, MSI microsatellite instability, TAI telomeric allelic imbalance, TMB tumour mutational burden.

Patients will be considered eligible for PARPi therapy if they are found to harbour a HRD positive status as determined by CDx. Variants of unknown significance could be reported but could not be used for treatment.

Clinical assessment of CDx testing

Compared with placebo treatment or chemotherapy, PARPi conferred significant clinical benefit in a range of cancers, particularly in patients both with BRCA mutations and HRD positive status,^4,5,7,22 as defined by NGS approaches. To support premarket approval applications, the positive clinical performance of CDx testing was demonstrated based on efficacy data from clinical trials, as outlined below.

In a randomised controlled trial (SOLO-1; NCT01844986) that compared the efficacy of olaparib (n = 260) with placebo (n = 131) in patients with newly diagnosed advanced ovarian cancer, eligible patients were required to have a documented (suspected) deleterious pathogenic BRCA mutation as determined by institutional testing or assessed by CDx testing as part of the enrolment procedure in the study prior to randomisation.³ A relevant improvement in progression-free survival (PFS) was observed, with a 70% reduction in the risk of disease progression or death according to the BRCA mutation subset determined by institutional testing (hazard ratio: 0.30; 95% CI: 0.23–0.41; P < 0.0001) or by Myriad myChoice CDx testing (hazard ratio: 0.290; 95% CI: 0.21–0.41; P < 0.0001).

Finally, after a median duration of follow-up of 41 months, a median PFS was not reached in the olaparib group regardless of institutional or commercial CDx testing.

Challenges associated with CDx testing

To date, it remains difficult to identify strong and robust predictors of HR defects beyond alterations in BRCA and/or other mutational signatures or genomic instability signatures. Moreover, a critical need exists for consensus guidelines from an international task force that is dedicated to NGS testing, with a special focus on gene alterations that are likely to support the decision to proceed with PARPi treatment.

Clinical data from the placebo-controlled, randomised NOVA trial assessing niraparib therapy in ovarian cancers have shown that 134 out of 553 patients were enrolled in non-gBRCAmut/HRD-negative group (HRD-negative status <42 scored by myChoice CDx) without identified genomic instability.²² In this group, niraparib exhibited a clinical benefit with a median PFS of 6.9 months vs 3.8 months (hazard ratio: 0.58; 95% CI: 0.36–0.92; P = 0.02). These results from NOVA were close to those obtained in other studies and although the improvement in PFS was less than that for gBRCA cohort (median, 21.0 vs 5.5 months; hazard ratio: 0.27; 95% CI: 0.17–0.41; P < 0.001) or HRD positive subgroup of the non-gBRCA cohort (median, 12.9 vs 3.8 months; hazard ratio: 0.38; 95% CI: 0.24–0.59; P < 0.001), it suggests that more patients that lack clearly identified biomarkers could benefit from PARPi.^23,24

Preliminary data suggest that promoter methylation (BRCA1, RAD51C) or a new mutational signature of HRD, called “Sig3” (mutational signature established from whole genome sequencing, independent of the effect of these mutations on the functionality of the encoded proteins), could also be candidate biomarkers for PARPi sensitivity in ovarian cancer and other tumour types.^23,25,26

In the PROfound or TRITON-2 trials, the safety and efficacy of PARPi (olaparib and rucaparib, respectively) were evaluated in men with metastatic castration-resistant prostate cancer. Interestingly, patients were prospectively enrolled based on genomic alterations in DNA damage response (DDR) pathway beyond BRCA mutations, such as ATM or CDK12, considered as potential biomarkers for PARPi sensitivity. In contrast with patients carrying BRCA mutations, a limited responses (clinical, biological or radiological) were observed in non-BRCA DDR gene alterations subgroups, suggesting that prognostic impact of all DDR mutations are not equal and requiring supplementary investigations to identify new genomic alterations related to PARPi benefit.^27,28 An important limitation of current HRD analyses is constituted by the fact that they are frequently performed at baseline, before any treatment, especially in the case of cancers difficult to biopsy (e.g., ovarian cancers). In this context, the occurrence of reversions or secondary mutations that might restore HR and thereby lead to PARPi resistance is not assessed, and this can lead to inappropriate administration of PARPi. The NGS could efficiently detect reversion or secondary mutations, if the tumour samples would be picked up at the beginning (at the time) of the treatment by PARPi. Another possibility will be to work on liquid biopsies (circulating tumour DNA, circulating tumour cells) performed before PARPi administration. However, NGS itself exhibits some limitations: its sensitivity is limited as compared with digital PCR (particularly for the detection of variants with low allelic fraction [below 1 or 2%]); it does not consider epigenetic alterations (BRCA1 or RAD51C promoter methylation, miRNA expression modifications) that could be involved in HR alteration and many variant remained of uncertain (or unknown) significance cannot be selected for PARPi treatment decision. The significance of all variants remaining under investigation, it is sometimes unable to identify mutations with functional consequences on HR pathway.

Taken together, these results indicate that NGS-based tests can thus produce both false-positive and false-negative results with regard to the HR status and prediction of sensitivity to PARPi, which strongly indicates the need for complementary functional assays.

Functional assays to assess PARPi sensitivity

Using ex vivo, in vivo or in vitro approaches, patient-derived three-dimensional (3D) models are now widely used in the field of oncology in functional assays (Fig. 3) to assess PARPi sensitivity and to identify new biomarkers in order to predict patient clinical outcome.^29–31

Fig. 3. A proposed integrative approach to assess PARPi sensitivity.

Combined to next-generation sequencing, PARPi sensitivity could be assessed directly through dose–response tests applied on patient-derived tumour organoids (PDTO). Furthermore, both PDTO and tumours (or spheroids isolated from paracentesis) could be exposed to X-ray in order to assess homologous recombination capacity and to identify indirectly those who are deficient and could benefit from PARPi therapies. HR homologous recombination, NGS next-generation sequencing.

Indirect assessment

Beyond assessing the response to PARPi by directly exposing cells or tumour tissue to these agents, some assays developed over the past 10 years can potentially predict PARPi sensitivity. Among them, the repair capacity (RECAP) test was first performed to assess the formation of RAD51 foci in proliferating cells (in S/G2-phase) following irradiation of fresh breast cancer samples^32,33 as well as fresh primary ovarian cancer cells.³⁴ As RAD51 is required for the repair of DNA DSBs by HR, a lower percentage (<10%) of proliferating cells with RAD51 foci was demonstrated to be associated with PARPi-sensitive models and a higher percentage associated with PARPi-resistant models. Based on this scoring, an HR status could be assigned for each sample (irradiated or at basal state), such as HR-deficient status, reflecting impaired HR pathway and potential PARPi sensitivity, or HR-proficient status, reflecting functional HR pathway and potential PARPi resistance. Usually, a cut-off at 10–20% of proliferative cells with more than five RAD51 foci in their nuclei enables the HR status in patient-derived xenograft (PDX) or tumour samples to be correlated with sensitivity to PARPi.^32,35 However, the RECAP test has not yet been used to distinguish clinical responders from non-responders to PARPi: a prospective clinical validation obtained by a dedicated clinical trial is undoubtedly required to allow such an application in routine practice. Another approach to extend this selection of potential PARPi responders could be to combine the detection of RAD51 activity with other factors involved in DNA repair pathways. For example, Kubelac et al.²¹ undertook a dynamic immunohistochemistry analysis of proteins involved in HR (RAD51, ATM, FANCD2), non-homologous end-joining (53BP1) and base excision repair (PARP-1, PAR) mechanisms using samples from advanced ovarian cancers before and after neoadjuvant chemotherapy to assess DNA repair competency and prognosis following treatment. Although the expression of PARP-1 (a target of PARPi) remained stable, at diagnosis and does not significantly changes after neoadjuvant chemotherapy tumours that harboured a combined profile of RAD51 proficiency and 53BP1 deficiency at diagnosis (suggesting an HR proficient phenotype and putative PARPi resistance) demonstrated significantly worse survival in multivariate analyses.²¹ This study suggests that setting-up integrated analyses that include the study of the expression of an extended panel of DNA repair-related proteins is of interest.

Another study documented a multidimensional fluorescence microscopy analysis of PARPi toxicity that simultaneously measured DNA-damage-associated gH2AX, HR competency (as revealed by RAD51 foci), DNA replication, cycle cell blockade, level of PARP inhibition, and mitotic disorders in low and high throughput at the single cell level. Such an analysis provides a complementary tool to usual RECAP test analyses that offers a better appreciation of the HR competency and sensitivity to PARPi.³⁶ However, like RECAP, this innovative approach also requires further clinical validation studies.

Direct assessment

3D tumour models using freshly resected samples, such as solid tumours (localised or metastatic), spheroids (from fluid paracentesis) or CTCs (from blood) can be used to evaluate the therapeutic effects of PARPi. Patient samples can be engrafted into immunocompromised mice to assess the anti-tumour activity of PARPi either alone or in combination. Such PDX models have been shown to recapitulate the genomic features of a patient’s tumour as well as its sensitivity to the combination of olaparib and the alkylating agent temozolomide in a co-clinical trial investigating small-cell lung cancer (SCLC).³⁷ Moreover, this novel strategy supported the discovery of a new molecular signature that is predictive of clinical response to these therapies and paves the way for rethinking clinical trial design, particularly in aggressive and some rare neoplasms with difficult therapeutic care. In addition to tumour resections, blood liquid biopsy samples can provide isolated CTCs from solid tumours; these CTCs are subcutaneously tumorigenic in mice and also mirror the clinical response of patients to chemotherapy.³⁸ Indeed, Lallo and colleagues showed that a CTC-derived xenograft of SCLC exhibited a heterogeneous response to olaparib, either alone or in combination with AZD1775, an inhibitor of the WEE1 kinase³⁹ particularly in a SCLC-derived PDX model with a mutation in PALB2, a gene that encodes partner and localiser of BRCA2 (PALB2) but that is not usually associated with HR deficiency in SCLC. Interestingly, this combination offers a new opportunity to stabilise SCLC, which carries a poor prognosis, and to overcome chemoresistance in selected samples taken at relapse, compared with those taken at diagnosis. Furthermore, PDX models provide a tool with which to explore and stratify new subgroups of patients according to their eligibility to receive PARPi by identifying a subset of highly sensitive and long-term responsive BRCA-proficient tumours in breast or ovarian cancers. PDX models have also demonstrated the possibility of restoring cell death in PARPi-resistant models by adding CHK1 or ATR inhibitors which lead to sensitise cells to PARPi therapies.^40,41 Nevertheless, engraftment time, logistical and cost challenges, as well as ethical considerations can limit co-clinical trials using PDX.

Beyond murine hosts, investigation of PARPi in xenografts using zebrafish larvae could offer a new option to reduce engraftment time (from 3–12+ months in mice to 1-3+ months in zebrafish larvae), to explore tumour heterogeneity and to test PARPi as a monotherapy or in combinatorial regimens over periods of time that are consistent with the clinical decision-making time frame.^42,43 However, logistical and pharmacological considerations such as death post-transplant, low number of transplanted cells failing to recapitulate tumour heterogeneity, short-term drug assays could limit their use in clinical practice.

Concurrently, patient-derived organoids have emerged as a breakthrough technology. These 3D multicellular structures grown from pluripotent or adult stem cells genetically and phenotypically reflect the characteristics of a particular organ or tissue.⁴⁴ It is also possible to establish patient-derived tumour organoids (PDTOs) from tumour samples, paracentesis fluid or CTCs, and these organoids can be used for many translational applications such as disease modelling, drug development and precision cancer treatment.^45,46 By assessing the capacity to predict clinical drug response and the functional consequences of tumour heterogeneity in 36 models from 23 ovarian cancer patients, a new platform showed that PDTOs recapitulate the clinical patient response to the standard of care (carboplatin in combination with paclitaxel).⁴⁷ The area under the curve (AUC) resulting from the exposure of PDTOs to carboplatin plus paclitaxel (where a smaller AUC indicates a greater response to carboplatin and paclitaxel combination treatment) showed a statistically significant correlation (P < 0.01) with the clinical response, as measured by chemotherapy response score, normalisation of the serum biomarker CA-125 and radiological responses. Moreover, all PDTOs were also found to be resistant to olaparib, rucaparib and niraparib, consistent with NGS results showing an absence HRD positive status as well as with a proposed cut-off used to identify PDTOs with IC50 value above it (considered as low-responders subgroup) from PDTOs with IC50 value below cut-off value (as high-responders subgroup). This cut-off is based on in vivo plasma PARPi concentrations obtained from previously published pharmacokinetic data. In another study eight out of 33 ovarian cancer models were identified to have an HRD-associated mutational signature potentially predictive of PARPi sensitivity but, notably, only two of these models (6%) were considered to be olaparib sensitive based on functional assay.⁴⁸ In breast cancer organoid models, a high total number of BRCA1/2-mutation-associated signatures was associated with the higher olaparib and niraparib sensitivity than a low total number of mutations.⁴⁹ Beyond gynaecological malignancies, pancreatic-cancer-derived organoid models lacking genetic alterations commonly found in the HRD panel exhibited differential sensitivity to PARPi: eight out of 24 were sensitive to olaparib and 11 of 24 were sensitive to talazoparib.⁵⁰ Similar data derived from other organoids^51–54 provide a strong rationale to explore gene alterations of unknown significance using functional assay-based basket trials and to assess PARPi (either alone or in combination) as an optional treatment for a subset of endometrial, gastric, lung or metastatic colorectal cancers.^51–54 Altogether, these observations support the potential benefit of extending the use of PARPi beyond BRCA-mutated patients and strongly recommend the use of functional assays.

Challenges and opportunities

As described above, PARPi can be used for the treatment of a wide variety of solid cancers in specific subgroups of patients who are deficient in HR (Table 1); current studies are investigating their potential in the therapeutic management of haematological malignancies.^55–58 However, the major challenge remains to clearly identify the patients who will really benefit from PARPi treatment — both to increase the efficacy of these agents and to avoid inappropriate toxicity. This article has highlighted the fact that genomic analysis can fails to identify all the patients that will respond to treatment and, conversely, sometimes results in the inappropriate selection of patients who will fail to respond. They also emphasise the need to consider genomic data together with the results of functional tests to substantially improve the selection of patients who are most likely to respond to PARPi therapy. This approach offers a new way of rethinking researcher–clinician interactions and incorporating 3D tumour-cell-based assays in clinical practice.⁵⁹

Among tumour-based functional assays, PDTO-based assays appear promising for this purpose. However, they do present some limitations that need to be addressed in the future to allow a bench to bedside process. First, it should be noted that the success rate of establishing PDTOs (currently 10–80%, depending on cancer type and culture conditions) needs to be improved, with culture media, extracellular matrix and tumour cell seeding conditions requiring optimisation and standardisation as much as possible. Another challenge of 3D-cell-based assay is the need to assess PARPi sensitivity over a short time span, consistent with clinical decision making, especially in a front-line or first-line setting. In the field of precision medicine, miniaturisation enabled by microfluidic-based culture systems could offer new opportunities to reduce sample requirement for drug testing and to allow high-throughput screening (or at least to test a greater number of anti-tumour agents such as standard of care and innovative molecules in the context of treatment response prediction).

Another important point to consider is the impact of the microenvironment on the tumour response to treatment. For instance, PARPi could increase tumour mutational burden, activate immunosuppressive pathways and reprogramme immune tumour microenvironment, offering a novel therapeutic strategy to combine immune checkpoint inhibitors (ICI) with PARPi in ongoing clinical trials.⁶⁰ However, to date, accurate and predictive biomarker to identify eligible patients to ICI–PARPi combinations are lacking. These results also highlight the need for innovative tools to predict the response to treatment to enable better patient selection. Ongoing research is investigating the possibility of enriching the microenvironment in PDTOs with autologous leucocytes in order to more closely mimic the situation observed in tumours and to predict the response to such ICI–PARPi combinations.^61,62

Finally, advances made in the isolation of CTCs could be of importance in the context of the selection of patients eligible for PARPi and their follow up. Although longitudinal solid biopsies at diagnosis and relapse can present a significant challenge in various cancers, they remain essential to perform 3D cell-based assays and to reveal primary or secondary resistance. If CTCs could be used during the follow-up of patients either to follow tumour burden or to identify secondary mutations that are likely to lead to resistance to PARPi, they also could be useful for establishing 3D models at the time of recurrence. Accordingly, functional assays could be suitable for predictive purposes, to identify resistance mechanisms and to provide alternative therapeutic strategies.

Taken together, these considerations suggest that, in the future, tumour-based functional assays could be incorporated to genomic studies to prevent unnecessary treatment or in the consideration of new treatment options. There is now a critical need for correlation studies that are able to validate the predictive value of such assays in order to transform their exciting potential in a real clinical value in the context of precision oncology.

Author contributions

Conception and design: P.-M.M., L.P. Methodology: NA. Acquisition of data and writing: all authors. Administrative, technical, or material support: NA. Study supervision: P.-M.M., L.P.

Ethics approval and consent to participate

Not applicable.

Data availability

Not applicable.

Competing interests

P.-M.M. was supported by the Cancer Institut Thématique Multi-Organisme of the French National Alliance for Life and Health Sciences (AVIESAN) Plan Cancer 2014-2019 (doctoral grant) and from the Comprehensive Cancer Center F. Baclesse. P.-M.M., L.-B.W., B.L. and L.P. are members of the “Réseau Normand d’Innovation Thérapeutique en Oncologie (ONCOTHERA)” network. “ONCOTHERA” European project is co-funded by the Normandy County Council and the European Union within the framework of the Operational Program ERDF/ESF 2014-2020. Work of the laboratory on ovarian cancer organoids is supported by Cancéropôle Nord-Ouest (“ORGRAFT” project), Ligue contre le Cancer (Calvados’s commitee), Fondation de l’Avenir (#AP-RM-19-020), the “Fondation ARC pour la recherche sur le cancer” (#PJA20191209649), and by the Normandy County Council and the European Union within the framework of the Operational Program ERDF/ESF 2014-2020 (“ORGATHEREX” and “ORGAPRED” projects).

Funding information

None.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-021-01295-z.