Harnessing preclinical models for the interrogation of ovarian cancer.

Ovarian cancer incorporates multiple malignancies with a variety of etiology, histopathology, and biological feature, including those derived from epithelial, germ cell, sex cord stromal, and metastatic lesions [ 1 ]. On a global scale, OC causes 152,000 deaths among 239,000 new cases annually, making OC the second leading cause of cancer-related death in women [ 2 ]. Most OC patients were diagnosed at an advanced stage with extensive peritoneal metastasis, yet symptoms are rather vague [ 3 , 4 ]. The dismal prognosis, limited treatment options, and high recurrence rate followed by stout resistance together render OC the most fatal gynecologic tumor, and the treatment of OC an ongoing challenge. The current standard of care for OC is primary debulking surgery followed by platinum-based chemotherapy [ 5 , 6 ]. Neoadjuvant chemotherapy might be considered for non-operable patients who are intolerant of surgery or deemed unlikely to reach a cytoreduction to no gross residual disease (R0) [ 7 ]. The advent of poly ADP-ribose polymerase (PARP) inhibitors, to a certain extent, revolutionized the treatment of patients with BRCA1/2 mutations. This is based on a theory named “synergetic lethality”, which was first described in 1946 [ 8 ]. Theoretically, dual inactivation of BRCA and PARP results in replication catastrophe, leading to the inevitable death of cancer cells. However, when it comes to treatment in a clinical setting, PARPi can lose its magic once resistance occurs, not to mention that the predominant beneficiary group carrying BRCA1/2 mutation only accounts for 15% of OC patients. Personalized medicine allowed advances in the exploitation of alternative strategies to tackle the aforementioned bottlenecks, yet the biological insufficiency of OC pre-clinical models to fully recapitulate the complexity of OC, to a certain extent, restraints the understanding and resolving of this heterogeneous disease [ 9 ]. A solid pre-clinical model should mirror the morphological and biological characteristics of the corresponding tumor of origin to the full extent. Historically, ever since the establishment of HELA cells, 2D cell line models have long been the mainstay of experimental cancer research [ 10 ]. Immortalized cancer cells cultured in artificial FBS-based medium boosted the exploration of cancer biology, somehow, they failed to bridge the gap between laboratory experiments and clinical trials. Abortions in phase III trials were mainly due to inadequate efficacy, which underlines the discrepancy between conventional cell lines and individual patients [ 11 – 13 ]. In response to this intractable situation, generations of pre-clinical models have been developed, among which patient-derived xenografts, organoids, tumor explants, and genetically engineered mouse models emerged as complementary tools in cancer research and drug screening (Fig.  1 ). In this review, we discuss the current status of OC models, and the strengths and drawbacks of each platform, casting light on the potentials and challenges of using these tools to unravel the biological fingerprints and vulnerabilities of OC. Fig. 1 Schematic representation of the most commonly used OC preclinical models. Both advantages and disadvantages are summarized for each model Schematic representation of the most commonly used OC preclinical models. Both advantages and disadvantages are summarized for each model

The past few years have witnessed accumulating knowledge concerning the tumorigenesis, progression, and evolution of ovarian cancer, which can be largely attributed to the development of a plethora of faithful preclinical OC models. It is now clear that OC is hallmarked by a high degree of inter- and intra-patient heterogeneity. By phenocopying the original tumor and/or expanding vital patient-derived tissues, tumor experimental models allow in-depth preclinical assessment of drug candidates and identification of tumor biomarkers at an individual level. The development and optimization of OC models are still actively ongoing, with the overall goal of better management and even the cure of OC. Given the inherent strengths and drawbacks of each model, it is imperative to be aware that one single model alone is definitely not competent to cover all OC research. Wise selection and rational combination of OC models are instrumental in solving the pain points in respect of OC. GEMMs are naturally suitable models for studying the cell of origin of OC and can be complemented by the newly developed OC organoid model. For the study of tumor invasion and metastasis, GEMMs and orthotopic PDX models can be utilized to mimic the biological process, otherwise, cell line and organoid models could be used to exploit the underlying mechanism. As for capturing and deciphering the heterogeneity and clonal evolution of OC, the highly individualized PDX, PDO, and PDE models stood out as edged tools. Ulteriorly, the patient-derived personalized models could be applied to drug development and repurposing, serving as “avatar models” for individual patients and further facilitating patient stratification, drug response assessment, and biomarker development in a clinical setting. To conclude, research in the post-genomic era yielded brand new insights into the biological and genetic fingerprints of OC. Henceforth, robust tumor models are required to validate the insights and distinguish those of value and therefore targetable. Conventional cell lines, PDXs, PDOs, PDEs, and GEMMs are all historically indispensable models for OC research. Especially, to realize the full translation from bench to beside, the new generation of patient-derived models will undoubtedly grow to be the mainstream in precision medicine. Future OC research should flexibly adopt suitable experimental models for various applications.

The first transgenic onco-mice generated by pronuclear injection of oncogene DNA were reported in the 1980s, initiating a novel field of genetically engineered mouse models (GEMMs) created by accurate manipulation of specific gene expression [ 154 ]. So far, the application of this sophisticated mouse model includes but is not limited to various types of cancers, helping to figure out the multistage in tumor initiation and progression, validate candidate genes and assess therapeutic efficacy [ 155 ]. Though tumor xenografts remain the most extensively applied mouse models in preclinical research, GEMMs circumvent several issues with xenografts that were discussed in the previous chapter [ 156 ]. In GEMMs, tumor cells are generated de novo in the context of a native milieu and within a whole organism. This way, the crucial tumor properties and modulators, such as immune cells and stromal elements are preserved and tumor cells can therefore co-evolve with the surrounding microenvironment [ 155 ]. Despite the success of GEMMs in other tumor types, OC GEMMs remain less than satisfactory. The putative reason is complex, yet can mainly be attributed to the paucity of prior knowledge of the origin and genetic basis of OC [ 157 ]. Above all, there are still controversies regarding the precursor of OC. On the genetic level, mutations in TP53 predominated the mutational spectrum of OC, other extensively reported alterations include RB1, EGFR, PTEN, PI3K/AKT, BRCA1/2, and KRAS et al. [ 158 ]. But how and to what extent perturbation of these genes contributes to the oncogenesis and progression of OC remains enigmatic. Efforts have been made to explore the consequence of the ablation or overexpression of genes of interest in OC GEMMs, which has been reviewed in Table  3 . Table 3 Key information and opinions in recent OC GEMMs research Targeted genes Targeting technology Histotype Original findings and opinions Reference Brca1, Tp53, Pten, Lkb1 CRISPR-Cas9 HGSOC 1). Quadruple deletion of Brca1, Tp53, Pten, and Lkb1 resulted in ovarian surface papillary tumors 4 months post-TAM. 2). Within 6 months post-TAM, widespread peritoneal metastasis formed in the Lkb1 deletion cohort, and some mice generated ascites by 7 months post-TAM. 3). Between 6 and 14 m post-TAM, the incidence of peritoneal metastasis was 96% and the incidence of ascites was 74% in the Lkb1 deletion cohort. Teng, Katie et al. [ 159 ] Trp53, Pten, Rb1, Cdh1 Amhr2 promoter driven Cre LGSOC, HGSOC 1). Triple deletion of Trp53, Pten, Rb1 initiated OC development in OSE cells. 2). Additional Cdh1 ablation promoted tumor dissemination and ascites formation. Shi, Mingxin et al [ 160 ]. Rb1, Brca1, Trp53 and/or Nf1 Ovgp1-TAM promotor driven Cre HGSOC, MMMT 1). FTE-specific inactivation of Brca1, Trp53, Rb1, and Nf1 resulted in STICs that progressed to HGSOC, with widespread metastases in some cases. 2). Brca1, Trp53 and Pten inactivation in the oviduct resulted in STICs and HGSOCs and was associated with diffuse epithelial hyperplasia and mucinous metaplasia. 3). Tumour initiation and/or progression in mice lacking conditional Pten alleles probably require the acquisition of additional defects. Zhai, Yali et al. [ 161 ] Pten, Apc Ovgp1-TAM promotor driven Cre, AdCre EMC 1). Oviductal epithelial hyperplasia and atypia formed ~ 1 month post-TAM. 2). Well-formed oviductal EMC-like tumors formed 9–12 weeks post-TAM. 3). 10 of 15 mice had extensive OC, 4 with omentum metastases; 1 with lung metastases. Wu, Rong et al. [ 162 ] Arid1a, Pten;Apc AdCre EMC 1). Arid1a inactivation enhanced epithelial differentiation in a murine model of EMC. 2). Arid1a inactivation resulted in prolonged survival in the Apc/Pten-deficient EMC model. Zhai, Yali et al. [ 163 ] Pten, Kras, Trp53 Amhr2 promoter driven Cre MOC, LGSOC, SOC 1). Trp53 R172H mutation promoted EOC but differently contribute to the disease in the presence or absence of the wild-type TP53 allele. 2). Ovarian tumors homozygous for Trp53 R172H mutation were undifferentiated and highly metastatic, exhibited minimal TP53 transactivation activity, and expressed genes with potential regulatory functions in EOC development. Ren, Yi A et al. [ 164 ] Apc Pgr promotor driven Cre EMC 1). In 87.2% of Pgr Cre/+ ; Apc ex15lox/lox mice, lesions were found in the epithelium of the distal oviduct and fimbriae. 2). In 16.3% of mice, endometrioid cysts were detected. 3). In 27.9% of mice, endometrioid ovarian tumors developed. van der Horst, Paul H et al. [ 165 ] Trp53, Brca1, Brca2, Pten Pax8-TET promotor driven Cre HGSOC 1). Deletion of Brca1 or Brca2, Tp53, and Pten in FTE resulted in STIC lesions, HGSOC, and the progression to advanced stage disease with metastases. 2). GEMM tumor showed human HGSOC biomarkers and genomically correlated with TCGA data. Perets, Ruth et al. [ 166 ] Trp53;Rb;Brca1;Brca2 AdCre HGSOC 1). Inactivation of RB induced surface epithelial proliferation with progression to stage I carcinoma. 2). Additional biallelic inactivation and/or missense p53 mutation in the presence or absence of Brca1/2 caused progression to stage IV disease. Szabova, Ludmila et al. [ 167 ] Dicer1, Pten Amhr2 promoter driven Cre HGSOC 1). Dicer-Pten double-knockout resulted in aggressive primary fallopian tube tumors with ascites. 2). Fallopian tube removal at early age prevented tumor formation, confirming the FTE as tumor origin. Kim, Jaeyeon et al. [ 168 ] Pten, Pik3ca AdCre SOC; GCT 1). Pik3ca mutation requires a second hit to initiate tumorigenesis in the ovary. 2). Pik3ca H1047R or Pten deletion in the ovary induced serous papillary hyperplasia and cooperated to induce SOC or GCT. Kinross, Kathryn M et al. [ 169 ] Pten, Kras Amhr2 promoter driven Cre LGSOC 1). Mutant mice developed LGSOC at an early age and with 100% penetrance. 2). KRAS is a key driver of OSE transformation. Mullany, L K et al. [ 170 ] Trp53, Brca1, c-Myc Retrovirals-depended Cre SOC 1). Myc could induce malignant transformation in Brca1 and p53 deficient cells but was not sufficient for the transformation of cells deficient for either Brca1 or p53. Xing, Deyin et al. [ 171 ] Pten, K-ras AdCre EMC 1). GEMMs showed endometriosis-like lesions within the OSE but no invasive ovarian tumors up to 10 months post-infection. 2). All GEMMs developed invasive EMC as early as 7 weeks post-infection Dinulescu, Daniela M et al. [ 172 ] Trp53, Rb1 AdCre EOC 1). Dual inactivation of p53 and Rb1 is sufficient for reproducible induction of ovarian epithelial carcinogenesis in mice homozygous for conditional gene alleles. 2). Ovarian neoplasms spread intraperitoneally with ascites, and metastasize to the contralateral ovary, the lung, and the liver. Flesken-Nikitin, Andrea et al. [ 173 ] Trp53, c-Myc, K-ras, Akt Retroviral gene delivery NI 1). Addition of any two of the oncogenes c-myc, K-ras, and Akt were sufficient to induce maliganant transformation in ovarian cells deficient for p53, 2). The induced ovarian tumors in mice resembled human ovarian carcinomas in their rapid progression and intraperitoneal metastatic spread. Orsulic, Sandra et al. [ 174 ] HGSOC High-grade serous ovarian carcinoma, LGSOC Low-grade serous ovarian carcinoma, MMMT Malignant Müllerian mixed tumor, EMC Endometrioid carcinoma, MOC Mucinous carcinoma, SOC Serous ovarian cancer, GCT Granulosa cell tumors, EOC Epithelial ovarian cancer, TAM Tamoxifen, TET Tetracycline, STICs Serous tubal intraepithelial carcinomas, FTE fallopian tube epithelium, OSE Ovarian surface epithelium, NI Not informed Key information and opinions in recent OC GEMMs research 1). Quadruple deletion of Brca1, Tp53, Pten, and Lkb1 resulted in ovarian surface papillary tumors 4 months post-TAM. 2). Within 6 months post-TAM, widespread peritoneal metastasis formed in the Lkb1 deletion cohort, and some mice generated ascites by 7 months post-TAM. 3). Between 6 and 14 m post-TAM, the incidence of peritoneal metastasis was 96% and the incidence of ascites was 74% in the Lkb1 deletion cohort. 1). Triple deletion of Trp53, Pten, Rb1 initiated OC development in OSE cells. 2). Additional Cdh1 ablation promoted tumor dissemination and ascites formation. 1). FTE-specific inactivation of Brca1, Trp53, Rb1, and Nf1 resulted in STICs that progressed to HGSOC, with widespread metastases in some cases. 2). Brca1, Trp53 and Pten inactivation in the oviduct resulted in STICs and HGSOCs and was associated with diffuse epithelial hyperplasia and mucinous metaplasia. 3). Tumour initiation and/or progression in mice lacking conditional Pten alleles probably require the acquisition of additional defects. 1). Oviductal epithelial hyperplasia and atypia formed ~ 1 month post-TAM. 2). Well-formed oviductal EMC-like tumors formed 9–12 weeks post-TAM. 3). 10 of 15 mice had extensive OC, 4 with omentum metastases; 1 with lung metastases. 1). Arid1a inactivation enhanced epithelial differentiation in a murine model of EMC. 2). Arid1a inactivation resulted in prolonged survival in the Apc/Pten-deficient EMC model. 1). Trp53 R172H mutation promoted EOC but differently contribute to the disease in the presence or absence of the wild-type TP53 allele. 2). Ovarian tumors homozygous for Trp53 R172H mutation were undifferentiated and highly metastatic, exhibited minimal TP53 transactivation activity, and expressed genes with potential regulatory functions in EOC development. 1). In 87.2% of Pgr Cre/+ ; Apc ex15lox/lox mice, lesions were found in the epithelium of the distal oviduct and fimbriae. 2). In 16.3% of mice, endometrioid cysts were detected. 3). In 27.9% of mice, endometrioid ovarian tumors developed. 1). Deletion of Brca1 or Brca2, Tp53, and Pten in FTE resulted in STIC lesions, HGSOC, and the progression to advanced stage disease with metastases. 2). GEMM tumor showed human HGSOC biomarkers and genomically correlated with TCGA data. 1). Inactivation of RB induced surface epithelial proliferation with progression to stage I carcinoma. 2). Additional biallelic inactivation and/or missense p53 mutation in the presence or absence of Brca1/2 caused progression to stage IV disease. 1). Dicer-Pten double-knockout resulted in aggressive primary fallopian tube tumors with ascites. 2). Fallopian tube removal at early age prevented tumor formation, confirming the FTE as tumor origin. 1). Pik3ca mutation requires a second hit to initiate tumorigenesis in the ovary. 2). Pik3ca H1047R or Pten deletion in the ovary induced serous papillary hyperplasia and cooperated to induce SOC or GCT. 1). Mutant mice developed LGSOC at an early age and with 100% penetrance. 2). KRAS is a key driver of OSE transformation. 1). GEMMs showed endometriosis-like lesions within the OSE but no invasive ovarian tumors up to 10 months post-infection. 2). All GEMMs developed invasive EMC as early as 7 weeks post-infection 1). Dual inactivation of p53 and Rb1 is sufficient for reproducible induction of ovarian epithelial carcinogenesis in mice homozygous for conditional gene alleles. 2). Ovarian neoplasms spread intraperitoneally with ascites, and metastasize to the contralateral ovary, the lung, and the liver. 1). Addition of any two of the oncogenes c-myc, K-ras, and Akt were sufficient to induce maliganant transformation in ovarian cells deficient for p53, 2). The induced ovarian tumors in mice resembled human ovarian carcinomas in their rapid progression and intraperitoneal metastatic spread. HGSOC High-grade serous ovarian carcinoma, LGSOC Low-grade serous ovarian carcinoma, MMMT Malignant Müllerian mixed tumor, EMC Endometrioid carcinoma, MOC Mucinous carcinoma, SOC Serous ovarian cancer, GCT Granulosa cell tumors, EOC Epithelial ovarian cancer, TAM Tamoxifen, TET Tetracycline, STICs Serous tubal intraepithelial carcinomas, FTE fallopian tube epithelium, OSE Ovarian surface epithelium, NI Not informed Another factor that confounds the path of developing OC GEMM might be the lack of validated ovary-specific promoters to facilitate tissue or cell-specific functioning of the gene-editing system. In 2003, Connolly et al. reported the first transgenic mouse with poorly differentiated ovarian carcinoma which was developed by induced expression of the transforming region of SV40 driven by the Müllerian inhibitory substance type II receptor gene promoter (MISRII) [ 175 ]. SV40 Tag could functionally inactivate the tumor suppressor gene RB and P53, leading to malignant transformation of epithelial cells [ 176 ]. Using this method, approximately 50% of cases successfully developed into bilateral OC with peritoneal metastasis and ascites, which shared high similarity with clinical OC patients and thus were of high clinical relevance. In addition, the malignant ascites of the model were further utilized to establish a cell line model, which exhibited the key properties of epithelial OC. Notably, as a major defect in this transgenic model, the infertility of the female mice precluded the stabilization and expansion of this transgenic line [ 175 ]. In order to solve this issue, Connolly’s group generated an affected male founder, TgMISIIR-Tag-DR26, the female offspring of which would develop bilateral ovarian tumors with varying latency and similar histological characteristics of HGSOC. By backcrossing the TgMISIIR-TAg transgenic line, the authors obtained murine ovarian carcinoma (MOVCAR) cell lines from the malignant ascites of tumor-bearing C57BL/6 TgMISIIR-TAg transgenic mice [ 177 ]. Next, the advent of the Cre-loxP system helped to explore more possibilities in OC GEMMs development [ 157 ]. Regarding the mechanism of this mammalian gene-editing technology, Cre recombinase discovered from bacteriophage P1 recognizes a 34 base pair specific sequences called loxP site and meditates exact recombination between two loxP sites that flank the target gene. Apart from gene excision, the preset location and orientation of loxP sites can also mediate gene translocation and inversion [ 178 ]. Later, strategies were explored including AdCre injection into the ovarian bursa and Amhr2, Pax8, and Ovgp1-mediated Cre expression in Müllerian-derived epithelia and rendered various OC phenotypes under the manipulation of different suspicious OC driver genes [ 179 ]. The most extensive application of GEMMs lies in the study of cancer initiation and progression. A quadruple combination of perturbations including Pten, Trp53, Rb1, and/or Cdh1 was adopted by Shi M et al. using Amhr2cre/+ mice, developing invasive OC with extensive peritoneal metastasis by targeting ovarian surface epithelium [ 160 ]. The cell-of-origin of HGSOC remains controversial during the past decades. GEMMs are suitable models to decipher this critical question [ 180 – 182 ]. Flesken-Nikitin A et al. identified the hilum region of the mouse ovary as a previously undefined stem cell niche of the OSE and were susceptible to malignant transformation into epithelial OC. Hilum cells showed preferential transformation after conditional deletion of Trp53 and Rb1 using the Ad-Cre/LoxP system [ 183 ]. By introducing genetic abnormalities of combined RB family inactivation and Tp53 mutation in Pax8 + FTE and Lgr5 + OSE or OSE-derived organoids, Zhang S et al. confirmed that HGSOC may originate from both FTE and OSE and the biological behavior of tumor might vary between different tumor of origins [ 184 ]. Collectively, OC GEMMs still face major challenges both in development and application. This kind of model may have a relative disadvantage in mirroring the heterogeneity of OC, but as critically emphasized, GEMM bears an irreplaceable value in the study of carcinogenesis and the cross-talk between immune and stromal components and cancer cells.

Conventional cell line models, ever since the establishment of the HELA cell line, had long been acknowledged to profoundly facilitate the understanding of tumor initiation, progression, metastasis, and drug discovery [ 10 ]. Well-established cell line models recapitulate the hallmarks of parental tumors in a way that can mimic the molecular aberrations and vital biological events. As regards to OC, there have been approximately 100 cell lines publicly available, most of which, however, lack proper annotations at histological, cellular, and molecular levels [ 14 ]. Based on transcriptomic data, Barnes BM and colleagues clustered 44 OC cell lines into 5 transcriptionally different groups, highly corresponding to the 5 major histological subtypes of OC [ 15 ]. It is worthwhile to mention that a bonafide list of reliable cell line models for distinct human pathology is needed to guide type-specific OC research [ 16 , 17 ]. Currently, compiling works have been devoted to the establishment of OC cell line models with defined pathology and well-characterized phenotypic and genomic profiles. The team of Kreuzinger C innovatively established reliable high-grade serous OC (HGSOC) cell lines from clinical OC samples, while providing a detailed background in clinical parameters, chemosensitivity status, and molecular alterations. And in a further step, a total of 34 established cell lines were used to decode potential therapeutic targets specific for HGSOC [ 18 , 19 ]. In addition, cell lines for ovarian clear cell carcinoma (CCC) [ 20 ], low-grade serous ovarian carcinoma (LGSOC) [ 21 ], endometrioid adenocarcinoma (EMC) [ 22 ], and other pathology were cultured and introduced by researchers worldwide. Notably, the OC cell line toolbox is further expanded by the development of drug-resistant cells and syngeneic mouse cell lines as well. A fundamental pain point in OC treatment is the stout resistance to chemotherapies, which calls for in vitro drug-resistant models to help understand the underlying mechanism. Methodologically speaking, one option is to derive resistant models from clinically relevant samples [ 23 , 24 ], another is to establish isogenic drug-resistant cell lines, usually through stepwise sequential exposure to increased concentration of drugs [ 25 ]. For example, the carboplatin-resistant A2780 cells [ 26 ], taxol-resistant OC3/TAX300 cells [ 27 ]. Interestingly, comparisons between isogenic drug-resistant cell lines and their background counterpart may serve as the lens to look into the trivia during the development of OC drug resistance [ 28 , 29 ]. As for syngeneic murine cell lines, ID8 is the most widely used and also the most extensively characterized model for epithelial OC. The ability of ID8 cells to form sporadic lesions in the peritoneal cavity of immune-competent mice closely mimics the biological events in stage III and/or IV OC in patients [ 30 ]. Henceforth, syngeneic murine cell lines are still representable and the most convenient models to use for the study in the organism with full immunity and comprehensive stroma components. OC cell line model, as a fundamental tool, remains to be the cornerstone in cancer models for its easy manipulation, low cost, and short doubling time [ 31 ]. The US National Cancer Institute (NCI) 60 human tumor cell lines representing 9 distinct tumor types (including ovarian cancer) were assembled in the late 1980s under a disease-oriented concept, aiming at identifying the anti-tumor effect of compounds in particular tumor types [ 32 , 33 ]. Moving beyond its original intention, the NCI 60 panel, combined with high-throughput technology, served as a pipeline in the field of drug screening [ 34 ]. Among the 60 extensively characterized cell lines, IGR-OV1, OVCAR-3, OVCAR-4, OVCAR-5, OVCAR-8, and NCI/ADR-RES are considered to be the 6 most representative OC cell lines to undergo numerous validations of promising small molecule compound [ 35 ]. Other large-scale datasets including The Cancer Cell Line Encyclopedia (CCLE) [ 36 ], the Cancer Genome Project (CGP) [ 37 ], and Cancer Therapeutic Response Portal (CTRP) facilitate the in-depth study of drug response across cell lines. Established cell lines are and will continue to be of paramount significance in every aspect of OC research. It is now beyond doubt that cell lines facilitated the study of cellular dependency in high-throughput pharmacological interrogation and genetic screening. Bulk of cell line studies on the genome-wide level revealed the genetic alterations related to drug sensitivity and resistance. Papp et al. integrated genomic, epigenomic, and expression analyses to shed light on the molecular abnormalities in an ovarian cancer cell line panel comprised of 45 OC cell lines and revealed unique molecular dependencies of several targeted therapies [ 38 ]. Garnett MJ et al. conducted systematic pharmacogenomic profiling in a pan-cancer cohort including 17 OC cell lines to provide an extensive view of the genomics underlying drug sensitivity. Using cell line models, large-scale loss-of-function screens such as siRNA, shRNA, and CRISPR-Cas9 library yielded new insights on identifying “driver” mutations among the “passenger” ones [ 39 – 42 ]. Mengwasser KE and colleagues conducted shRNA and CRISPR screening with DNA-repair-based libraries on 2 pairs of BRCA2 isogenic cell lines, including ovarian PEO1 B2MUT cells, revealing FEN1 and APEX2 as BRCA2 synthetic lethal targets, which could possibly be utilized in future treatment of BRCA-deficient populations [ 43 ]. In addition, the examples of utilizing OC cell line models to predict drug efficacy and explore underlying molecular mechanisms in OC initiation, progression, metastasis, and response to clinical and novel therapies are far too numerous to be detailed. Experimental technique evolving cell line models are well-developed and still undergoing a continuous boom. In vivo tumorigenicity of some cell lines endowed them with the utility to form xenografts for the study of tumor initiation and metastasis as well as in vivo drug response in animal experiments [ 44 , 45 ]. Though historically served as the main force in cancer research models, conventional cell lines are continuously questioned in their fidelity and clinical relevance [ 46 – 50 ]. Serial analysis of gene expression (SAGE) database identified distinct gene clusters overexpressed in cell lines and solid tumors respectively, which might give rise to the deviations of cell lines in respect of chemotherapy resistance [ 47 ]. Comparison between OC cell lines and clinical samples revealed that cell line models failed to capture clinical MDR gene expression patterns with their high selectivity in the expression of genes associated with MDR [ 48 ]. A research comprised of 41 OC cell lines indicated that selective pressure against BRCA1/2 mutations during adaption to 2D culture might contribute to a lower incidence of BRCA1/2 deleterious mutations than the population-level incidence [ 51 ]. Another comparative study looked at the CNVs, mutations, and mRNA expression profiles between commonly used OC cell lines and HGSOC samples and revealed pronounced gaps in molecular profiles. Startlingly, the most popular OC cell lines such as SK-OV-3, A2780, OVCAR-3, CAOV3, and IGROV1 showed low correlation value with individual tumors while rarely applied cell lines such as KURAMOCHI, OVSAHO stood out as “good cell lines” [ 50 ]. However, being a representative cell line does not necessarily mean that it can successfully and efficiently form xenografts in vivo, again restraining the utility of cell line models [ 44 , 52 ]. As the genomic fingerprints, even druggable targets might be lost over extensive passages in vitro, the utilization of this conventional model is fairly discouraged [ 48 ]. Notably, given the random nature of the incidence of genomic drift, it’s possible for identical cell lines to differ even among laboratories. Clearly, OC cell lines commonly used in published studies do not stand for the full complexity of OC. To summarize, conventional OC cell line models, by their very nature, cannot afford the rapid expansion of precision medicine alone.

Another patient-derived model taken into consideration is the patient-derived explants (PDE), which refers to the short-term ex vivo culture of freshly obtained, surgically resected human tissue, either in chunks or cut into slices [ 147 ]. Although the methodology of PDE and PDE-related drug testing has been around for a long time, unfortunately, it has not translated into a wide range of applications and thus has not become mainstream in cancer research, particularly in OC. Limited reports have been made on the preclinical application of this model. Hence we will briefly introduce the PDE model and its application in OC research. Technologically, the thickness of explants might affect the viability during culture and drug response considering the efficiency of nutrition diffusion and metabolites transportation. Manual slicing with surgical equipment used to be the most common method in tissue preparation yet has now been gradually replaced by mechanized methods such as the vibratome [ 148 ], which can uniform the thickness of slices. According to the report of Parajuli N et al., a slice thickness of 160 μm is optimal for tissue handling and viability [ 149 ]. There is no universal formula for PDE culture conditions, tissue-specific supplements are thought to improve the viability of different types of PDEs [ 150 , 151 ]. Explants could be placed on support media such as gelatin sponges and pore membranes or could be cultured in a free-floating manner [ 147 ]. OC tissue could be cultured for about 5 days on gelatin sponges in RPMI1640 media with serum added [ 151 ]. Abreu S et al. reported OC free-floating PDE culture which reached the longevity of 30 days in DMEM media with serum addition [ 152 ]. Both studies lack supplementary additives to extend the lifetime of OC PDE models. Another technical issue is the endpoint analysis of PDE, which is critical for the evaluation of drug response. The most frequently used strategy is to measure cell viability with MTT assay after enzymatic digestion of the explant. Immunohistochemistry analysis of protein markers of cell proliferation and death is another commonly adopted method. Recently, multiple-immunofluorescence was adopted by researchers, offering innovative analytical approaches and better interpretation for PDE assays [ 147 ]. Though less adopted in OC research, the PDE model has its own unique advantages compared with PDX and PDO models. Firstly, PDE can be rapidly generated post-surgery without much technological and financial burden [ 153 ]. Besides, PDE retained the cell-cell interaction and cell polarity in a much more proper way. The existence of resident immune cells was validated by immunohistochemistry on OC PDE, showing evidence of the perseverance of tumor-infiltrating CD4 + and CD8 + T cells, macrophages and B cells [ 152 ]. Though more intuitively correlating drug response with patient pathology, the longevity of PDE severely restricted its practical applications. However, we hold the belief that by rationally combing this short-term culture with other models in OC research, PDE can assist in OC precision medicine.