Significance of RB Loss in Unlocking Phenotypic Plasticity in Advanced Cancers

Phenotypic plasticity can be defined as the ability of individual genotypes to produce different phenotypes when exposed to different environmental conditions (1). Often associated with developmental processes during embryonic differentiation, phenotypic plasticity is also a hallmark of cancer that can drive tumor initiation, metastasis, and therapy resistance (2, 3). Fluctuations in normal physiologic conditions impact cellular plasticity, and disrupted differentiation has been described as a mechanism for carcinogenesis in several cancers. Cells may undergo blocked differentiation, or dedifferentiation back to progenitor cells before achieving a terminal differentiated state, or in some cases, transdifferentiation can occur manifest as metaplasia or conversion to another lineage phenotype (2). Mechanisms of epithelial-to-mesenchymal lineage transitions have been implicated in carcinogenesis and cancer metastasis. In the context of therapeutic resistance, cancer plasticity may lead to loss of dependency on the original oncogenic driver posing a challenge for disease management (4). Changes in lineage program may manifest clinically as transformation from one histology to another, as in the case of a subset of lung and prostate adenocarcinomas that transform to small-cell carcinoma as a mechanism of resistance to EGFR or androgen receptor (AR)-targeted therapies, respectively (3, 4).

Phenotypic plasticity in cancer involves diverse and complex mechanisms that often differ across tumor types, thereby posing challenges for developing therapies to target cancer plasticity. Broadly, cancer lineage plasticity is orchestrated by reactivation of lineage-related master transcription factors accompanied by epigenetic reprogramming (3, 5). Recent data point to underlying genomic factors that facilitate cancer plasticity, including loss of the retinoblastoma gene RB1 (encoding the protein RB), which occurs across a broad spectrum of cancers (6). While the function of RB as a tumor suppressor and cell-cycle regulator is well described (7), its role as a mediator of cancer plasticity and therapy resistance has only recently emerged. Here we review the central role of the RB-E2F signaling axis and the clinical significance and biological consequences of RB loss in cancer, with a special focus on how RB deficiency impacts cancer plasticity (Fig. 1).

RB plays a central role in cell cycle as a checkpoint guardian that prevents progression from the G₁- to S-phase of the cell division cycle (7, 8). RB regulates cell cycle through its binding to the E2F family of transcription factors (9). Specifically, RB binds to the transactivation domain of E2F and represses gene transcription of factors required for G₁ to S transition including cyclin E and CDK2 (10, 11). Hyperphosphorylation of RB primarily mediated by CDK4/6 leads to its detachment from E2F, allowing E2F-target genes to be activated which are critical for S-, G₂-, and M-phases of the cell cycle (12, 13). Cell-cycle checkpoints are strictly regulated by cyclin–CDK complexes (14). Cyclins bind to and activate cyclin-dependent kinases (CDK); CDKs have specific targets depending on the cyclin they are bound to at a given state (15). Some of the key substrates of cyclin-CDKs involved in cell-cycle regulation are RB, p53, and p107 (15).

Genomic aberrations or functional loss of RB (e.g., through constitutive hyperphosphorylation, promoter hypermethylation, or other means) leads to dysregulation of this checkpoint and cell-cycle progression. Hyperphosphorylation of RB inactivates the RB protein during G₁- to S-phase during cell cycle. p16 encoded by CDKN2A is an endogenous inhibitor of CDK4/6 preventing RB hyperphosphorylation. Genomic loss of CDKN2A or promoter methylation facilitates unrestrained CDK4/6 activity and may lead to hyperphosphorylation-mediated functional loss of RB. RB phosphorylation results in the release of the E2F transcription factor and induction of DNA replication genes such as cyclin E, which facilitates transition to S-phase and drives cancer progression (16).

Given the critical regulatory role of RB on E2F activity, loss of RB results in increased activity of cyclins and CDKs and leads to unchecked proliferation (17–19). Defects in the RB pathway lead to overexpression of MAD2, a mitotic checkpoint protein that is a direct target of E2F (20). RB loss–mediated MAD2 overexpression leads to mitotic dysregulation seen in retinoblastoma, bladder, and neuroblastoma (20). Loss of RB facilitates the upregulation of CDK1 and CDK2 kinases that drive cell-cycle progression and cell division (21). A fraction of RB–E2F complexes reside on the promoters of apoptotic genes (22). Inactivation of RB may lead to abnormal expression of apoptotic genes potentially resulting in tumorigenesis (22). In murine intestinal models, Rb deficiency can lead to an accumulation of E2f3 protein and chromatin binding, resulting in activation of a G₁–S transcriptional network and enhanced cell proliferation (23). Increased E2F activity is implicated in several cancers including skin, head and neck, breast, and others (24, 25). In lung cancer, E2F1 and PTEN physically interact and PTEN regulates the transcriptional function of E2F1 (26).

The cistrome of E2F is not only limited to cell-cycle genes but can also have a broader effect on genes associated with diverse cellular functions. Both activation-associated (E2F1 and E2F2) and repression-associated (E2F7 and E2F8) E2F members have transcriptional targets beyond cell cycle, possibly contributing to oncogenic activity and malignant progression (19). In some cases, the E2F family members have redundant targets; a E2F1, E2F4, and E2F6 ChIP-ChIP analysis revealed interchangeable roles of E2F family members (27). E2F7 and E2F8, which are atypical repressor genes, can promote initiation of hepatocellular carcinoma when deleted (28). In an antiestrogen-resistant breast cancer model, an estrogen-independent role for estrogen receptor and the CDK4/RB/E2F transcriptional machinery was identified as critical for tumor progression (29). In prostate cancer models, E2F1 can directly regulate AR expression by binding to the promoter of the AR gene with pocket protein family members p107 and p130, suggesting that elevated E2F activity can promote progression of castration-resistant prostate cancer (CRPC; ref. 30). E2F has also been reported to physically interact with the AR to control expression of androgen-responsive genes (31). A recent study in prostate cancer highlights the expansion of E2F cistrome upon RB loss (32). The authors in this study found that changes in E2F cistrome were distinct upon RB hyperphosphorylation as compared with short hairpin RNA-mediated knockdown of RB, suggesting molecular differences based on the mechanism by which RB is rendered nonfunctional (32). HES6 which has been implicated in promoting CRPC progression preferentially regulates a transcriptional network enriched for E2F-binding sites (33). EZH2, a histone demethylase, acts as a transcriptional activator in CRPC, which maybe in part mediated by direct interaction with E2F (34).

The RB1 gene is the first described tumor suppressor gene originally identified in retinoblastoma, and later identified as dysfunctional across multiple cancer types (35–37). Loss of RB function in cancer predominantly occurs via heterozygous or homozygous genomic deletion of the RB1 gene on chromosome 13, followed by promoter methylation, mutation, and constitutive hyperphosphorylation of the RB protein (38). RB loss is seen across a diverse spectrum of cancers (Supplementary Table S1). In addition to retinoblastoma (>97%), RB1 is lost in 91%–100% of small cell lung cancer (SCLC; ref. 38), 60% of primary osteosarcomas, 17% of primary bladder urothelial carcinoma (39, 40), and 10% of primary prostate cancers (38). RB1 loss can also be acquired in later stages and is higher in the metastatic, treatment-resistant setting. Whole-exome and whole-transcriptome sequencing of 500 metastatic solid tumors revealed that RB1 was the fourth-most frequently deleted tumor suppressor gene across cancer types following TP53, CDKN2A, and PTEN (41).

Acquired loss of RB has been linked to therapy resistance in breast and prostate cancers. In preclinical studies, loss of RB has been associated with reduced sensitivity to antiestrogen therapy in breast cancer (42). Dysregulation of RB and its downstream E2F target genes has also been associated with shorter overall survival in patients with breast cancer treated with tamoxifen (42). Regardless of treatment modality, genomic characterization of patients with hormone receptor—positive (HR⁺), HER2-negative metastatic breast cancer revealed that RB1 genomic alterations are associated with increased risk of death (43).

Similarly, disruption of RB1 gene is enriched in metastatic CRPC compared with primary prostate cancer, with deletion of RB1 observed in up to a third of cases (44–46). Comprehensive genomic analysis of 429 patients with metastatic CRPC found that RB1 alteration was the only independent predictive factor for overall survival although other alterations (e.g., TP53, AR) were associated with shorter time on treatment with AR-pathway inhibitors (47). PSA is an androgen-driven serum biomarker used for the detection of prostate cancer as well as disease monitoring. Clinical and genomic features of 15 patients with metastatic CRPC within a low PSA subgroup were investigated in comparison with PSA-elevated cases (48); loss of RB1 and/or TP53 was more prevalent in low PSA populations (80% vs. 41%). This subgroup was more likely to have visceral metastasis (69% vs. 36%) and shorter overall survival (48). Hamid and colleagues investigated the prognostic significance of loss of tumor suppressor genes (TP53, PTEN, RB1) in localized, metastatic hormone-sensitive, and metastatic CRPC (45). Loss of RB1 concurrent with TP53 or PTEN was more often observed in patients with metastatic disease than those with localized prostate cancer. Of note, cumulative gene alterations were associated with poorer progression-free and overall survival in both early and advanced diseases (45). Taken together, RB1 loss is characterized as clinically aggressive variant in prostate cancer and is a poor prognostic factor regardless of disease stage.

SCLC is characterized by RB1 deficiency (>90%) while non–small cell lung cancer (NSCLC) harbors RB1 alterations in less than 10% of cases (49, 50). A recent study investigated 208 SCLC tumors using targeted next-generation sequencing and IHC (50). Only 14 patients (6%) had wild-type RB and most cases showed cyclin D1 high and mixed histologic component with NSCLC. Interestingly, Rb-deficient SCLC was associated with better survival than those with Rb-proficient SCLC in patients with extensive disease stage (P = 0.04), although this patient population also had higher tumor expression of neuroendocrine markers and proliferative (Ki67) index (50).

In urothelial cancer, RB1 alterations have been reported in up to 17% of primary bladder cancers. While RB1 deletions are found in only 4% of cases, RB1 mutations are more frequent (∼13%) which lead to deregulation of RB pathway and are mutually exclusive with CDKN2A alterations (39, 40). While RB1 alterations are uncommon in esophageal cancer and colorectal cancer (<4%), alterations in the RB pathway such as promoter hypermethylation of CDKN2A is observed in more than 50% of these tumors (38). Similarly, in hepatocellular carcinoma, RB1 alterations are infrequent (<10%); however, deletion or promoter methylation of CDKN2A is found in up to 70%, suggesting RB pathway deregulation (51). Genomic loss of CDKN2A is also frequent in other cancers such as glioma (up to 85%), head and neck (up to 66%), and bladder cancer (up to 45%; ref. 38). CCND1 (cyclin D1), a component in the RB axis involved in cell-cycle progression is amplified in endometrial (26%), pancreatic (25%), and breast cancers (up to 30%; ref. 38). Cyclin D1 is overexpressed in 90% of mantle cell lymphomas which is associated with a pathogenic chromosomal translocation (52). CDK4 is overexpressed in up to 54% of endometrial cancer and mutated in approximately 16% of breast cancers (38).

Loss of RB function can be determined clinically using DNA sequencing or IHC. A recent study of 29 metastatic small cell neuroendocrine prostate cancers found RB protein loss in 90%, of which 85% had biallelic genomic deletion of RB1 (53), suggesting that detection of RB loss by a validated IHC assay may be a useful diagnostic tool to complement genomic platforms (53). While the combination of DNA sequencing and IHC assays can detect homozygous/heterozygous loss and also reduced protein expression due to promoter hypermethylation or other mechanisms, functional loss of RB due to constitutive hyperphosphorylation is not reliably detected by these assays. A recent study highlighted the potential clinical significance of RB phosphorylation status as a marker of treatment response after 2 weeks of treatment with the CDK4/6 inhibitor palbociclib in presurgery trial of early-stage breast cancer (NCT02008734; ref. 54). Palbociclib significantly reduced RB phosphorylation as compared with baseline which correlated with expression of the proliferation marker Ki67, and the authors concluded that RB phosphorylation status could potentially identify patients with primary resistance to CDK 4/6 inhibitor therapy.

Transcriptomic signatures for predicting RB functional loss may provide insights to capture additional mechanisms of RB loss and downstream consequences (55). Combining microarray datasets with 900 breast cancer samples, a study developed an RB-loss gene signature with 159 genes primarily consisting of E2F-regulated genes involved in mitosis, cytokinesis, and DNA replication (56). More recently, an independent study using Cancer Cell Line Encyclopedia developed an RB-loss transcriptional signature that had relatively better accuracy at predicting RB1 loss (55). This gene signature consisting of 186 genes was not restricted to cell-cycle genes but also encompassed DNA damage response, p53 signaling pathways, pluripotency, and adipocyte differentiation which aligns with the other noncanonical functions of RB in cancer progression (discussed later; ref. 55). A versatile gene signature reflecting diverse noncanonical functions of RB may be able to capture both genomic and nongenomic mechanisms of RB loss (Fig. 2; ref. 57), yet RB loss gene signatures are not yet used clinically.

Noncanonical mechanisms of RB function include its role in genome stability, silencing of genomic repeats, DNA repair, metabolism, inflammation, autophagy, differentiation, among others (ref. 58; Fig. 2). These diverse functions may be mediated in part through its effects on chromatin architecture. These noncanonical functions of RB along with its central role in cell division have been suggested to play a key role in cell fate determination and cancer phenotypic plasticity.

Phenotypic plasticity is associated with cancer initiation, progression, and drug resistance. Perturbations in normal physiologic conditions such as injury, inflammation, changes in microenvironment, drug treatment, oxidative stress, and senescence influence cellular plasticity (59). Changes in lineage state due to these conditions may sometimes result in cancer initiation (3, 59). For instance, chemical injury can cause committed hepatocytes to dedifferentiate into bipotential progenitor cells dependent on Notch signaling, which is associated with cancer lineage plasticity driving the initiation of intrahepatic cholangiocarcinoma (60, 61). Plasticity-induced tumor initiation is also reported in breast cancer, melanoma, colon cancer, and sarcoma (62).

Cancer plasticity is also associated with metastatic disease progression and therapy resistance. An early example of phenotypic plasticity contributing to disease progression is in melanoma (63). By amplifying a melanocyte lineage regulator, MITF, melanoma cells acquire plasticity along with invasive and metastatic properties (63). Distinct melanoma cellular states related to MITF and NFκB activity also influence resistance to MAPK pathway inhibitors such as trametinib and dabrafenib (64, 65). Treatment resistance conferred by cancer phenotypic plasticity has also been reported in prostate cancer, lung cancer, breast cancer, pancreatic cancer, colon cancer, glioblastoma, bladder cancer, and basal cell carcinoma (4, 66–71). In many of these cases, cancer plasticity has been linked to epigenetic dysregulation or changes in chromatin permissiveness, ultimately affecting their transcriptome and resultant phenotype. While there is no robust causal evidence linking any specific genomic aberration(s) with cancer plasticity, RB loss of function appears to be a facilitator of lineage reprogramming in prostate and lung cancers.

Androgens are critical for growth and differentiation of the normal prostate and maintenance of homeostasis, and prostate adenocarcinoma is driven by the AR (72, 73). Resistance to AR-directed therapies is mostly due to reactivated AR signaling (74, 75). In up to 15%–20% of CRPC, tumor progression is associated with loss of AR signaling and acquisition of alternative neuronal/neuroendocrine lineage programs. Clinically, this can manifest as histologic transformation to a small cell neuroendocrine prostate carcinoma (NEPC). NEPC is clonally derived from prostate adenocarcinoma and has a similar genomic profile (46), yet are enriched for alterations involving RB1. Deletion or mutation of RB1 is seen in approximately 20% of castration-resistant adenocarcinoma and up to 70% of NEPC, and are often concurrent with TP53 in NEPC (46). In a genetically engineered Pten-deleted prostate cancer mouse model, deletion of Rb1 facilitates lineage plasticity and the development of NEPC (76). Gene expression profiling revealed loss of Ar expression along with upregulation of Ezh2; Ezh2 inhibition led to a reversal in phenotype with reexpression of Ar and resensitization to the AR-antagonist enzalutamide. Sox2, an important driver of pluripotency is also upregulated following Rb1 loss (76). A more recent study added a Chga-reporter to this mouse model and defined the role of FoxA2 in prostate cancer lineage plasticity (77). Another study reported human prostate cancer models with knockdown of both RB and p53 leads to upregulation of SOX2 and the emergence of NEPC and antiandrogen drug resistance (78). Although RB1 loss is an important medicator of phenotypic plasticity in prostate cancer, NEPC features have also been observed in a Pten/Trp53 loss-of-function model, with NEPC cells arising from a luminal cell precursor as delineated through lineage tracing (79). Mechanistically, this leads to an induction of Sox11 promoting neural differentiation, rather than Sox2 as witnessed in the context of RB1 loss. Of note, there was a significant reduction in Rb expression seen in Pten/Trp53 loss mice and further downregulation as the cells underwent overt NEPC differentiation (79). In a patient-derived xenograft (PDX) model that develops transdifferentiation from a high-grade primary prostate adenocarcinoma to NEPC after castration, single-copy loss of RB1 was seen in the primary AR-active adenocarcinoma, and RB function was further abrogated as the tumors transformed to an AR-negative NEPC as measured by RB-loss gene signature (80).

Functional loss of RB mediated via promoter hypermethylation or constitutive hyperphosphorylation can also occur in NEPC (81). Hyperphosphorylation-mediated functional loss of RB was described in a cell line–derived xenograft model that transdifferentiated to NEPC upon developing resistance to enzalutamide (82, 83). In this specific study, the authors highlighted dramatic changes in chromatin accessibility and EZH2 co-occupation of the AR cistrome that transcriptionally modulated stem cell and neuronal gene networks (82, 83). A recent study performed Assay for Transposase-Accessible Chromatin using sequencing on 40 prostate cancer models and highlighted the ability of chromatin profiles to classify subtypes of treatment-resistant disease (84). As expected, the authors saw an enrichment for RB1 deep deletion in NEPC models as compared with other subtypes (84). Interestingly, two of these models had an intact RB1 but a deletion of CDKN2A which correlated with hyperphosphorylation of RB and functional loss (85).

Using models with loss of RB function, other master transcriptional regulators have been described that drive lineage plasticity. A neural transcription factor BRN2 is overexpressed in NEPC and has been experimentally validated as a driver of NEPC in a cell line–derived prostate cancer model that developed resistance to enzalutamide treatment (83). As mentioned above, these cells also display a hyperphosphorylation-mediated RB functional loss (82, 83). Using NEPC PDXs with homozygous loss of RB1, another study demonstrated a potential mechanism for SRRM4 and REST as transcriptional regulators in the emergence of NEPC (86, 87). Overall, these studies support loss of RB (functional or genomic) as an important facilitator of prostate cancer lineage plasticity and therapy resistance. These studies also point to epigenetic reprogramming and changes to chromatin permissiveness, which may be in part due to the role of RB in normal development (reviewed later).

Mechanisms of resistance to EGFR-targeted therapies in lung adenocarcinoma are similar to other targeted therapies, with activating mutations involving EGFR, bypass mechanisms, and loss of target dependence (4). Like prostate cancers that lose AR dependence, loss of EGFR-signaling dependence has been associated with histologic transformation to SCLC (seen in 5%–14% of cases; refs. 88–90). This lineage plasticity is associated with retention of early genomic aberrations (including EGFR mutation), loss of EGFR expression, and acquisition of a transcriptional program that resembles primary SCLC. Patients that develop small-cell transformation are managed like de novo primary SCLC (91). A study evaluating RB1 status of 11 treatment-emergent transformed SCLC found that RB1 was lost in 100% of cases and was concurrent with TP53 mutation or heterozygous loss in all cases (88). Notably, RB1 loss and TP53 mutation are observed almost universally in de novo SCLC (91%–100%; ref. 92), and loss of Rb1 and Trp53 is sufficient for SCLC development in mouse models (93). A recent study found that in patients with EGFR-mutated lung adenocarcinoma, the presence of TP53 and RB1 alterations (43/863, 5%) was associated with greater risk for SCLC transformation (7/39, 18%) suggesting that may be early precursors of SCLC or facilitators of lineage plasticity (89). Furthermore, the triple-mutant cases (EGFR, RB1, TP53) had a higher incidence of whole-genome doubling and inferior clinical outcomes (89). An independent study evaluated whole-genome sequencing of serial biopsy timepoints of 4 patients with EGFR-mutant lung adenocarcinoma that transformed to SCLC, which also supported loss of RB1 and TP53 as potential early harbingers of SCLC transformation (90). IHC analysis of a larger cohort of 75 patients treated with EGFR inhibitor therapy found that lung adenocarcinoma harboring inactivated RB and p53 had a 43× greater risk of small-cell transformation (90). However, because not all patients harboring these alterations underwent SCLC transformation, other regulators may be required to achieve lineage plasticity. In EGFR- and TP53-mutant cell lines, knockdown of RB1 was insufficient to induce transformation to SCLC or drug resistance (88). Of note, an independent study using a Kras-driven murine lung adenocarcinoma model with Trp53 deletion demonstrated that Rb1 loss leads to increased expression of Nkx2-1 and FoxA2 facilitating lineage infidelity and cancer metastasis; however, these tumors did not express neuroendocrine markers (94). Histologic transformation to SCLC has also been reported in ALK-rearranged adenocarcinoma during acquired resistance to the ALK inhibitor alectinib (95), which has also been linked to loss of RB (96).

CDK4/6 inhibitors are routinely used for the treatment of HR⁺ breast cancer and polyclonal RB1 mutations have been associated with acquired resistance (97). However, in the PALOMA-3 study (521 patients), RB1 mutations were only found in a minority (4.7%) of patients treated with the CDK4/6 inhibitor palbociclib plus fulvestrant, suggesting that RB1 genomic alterations are not the primary mechanism of resistance to CDK4/6 inhibition (98). RB1 loss may also mediate breast cancer initiation. Conditional deletion of Trp53 and Rb1 in murine mammary epithelium accelerated development of mammary neoplasms (88%); however, deletion of Rb1 alone was not sufficient to induce neoplasms (99). Of note, Rb1 inactivation led to secondary genetic alterations and increased genomic instability associated with Trp53 mutations (99). Functional loss of RB is observed in up to 30% of triple-negative breast cancer (TNBC) and concurrent TP53 alterations are found in up to 40% of basal-like breast cancer (100). Rb1/Trp53-deleted cells in murine primary mammary epithelial cells including luminal progenitors induced basal-like tumors via epithelial-to-mesenchymal transition that shared similar features of TNBC (100). Basal-like breast carcinomas also resemble morphologic features of RB−/p16+ human papillomavirus (HPV)-related squamous cell carcinoma (101).

Disruption of the Rb1 gene in mouse embryos is lethal, and embryos die within 16 days of gestation with notable defects in erythroid and neuronal development (102, 103). While cells with Rb1 loss can initiate differentiation as they express early tissue-specific markers, late differentiation markers such as TrkA and TrkB (neuronal markers) are reduced suggesting a lack of complete differentiation (104).

Loss of RB impacts tissue-specification during the development of various tissue types (Fig. 3). Deletion of Rb1 in murine retinal progenitor cells leads to increased cell death of specific progenitor populations and ectopic proliferation mediated by E2f (105). Deletion of E2f1 or inhibition of Cdk2 activity is sufficient to prevent tumor initiation and retinoblastoma development in Rb1-deficient retina progenitors (106, 107). Deletion of Rb1 in lung epithelium results in increased number of neuroendocrine cells within the airway resulting in hypercellular neuroendocrine lesions suggesting that Rb loss preferentially induces neuroendocrine cell fate during lung development (108). Disruption of the Rb pathway in mouse intestinal cells (small intestine and colon) leads to hyperplasia and dysplasia of the intestinal epithelium, which may be due to dysregulation of terminal differentiation of these cells (109). RB also plays a critical role for the differentiation of the epidermis, as Rb1 loss results in hyperplasia and hyperkeratosis of the epidermis of mice and is accompanied by aberrant expression of differentiation markers (110). Rb is essential for the maintenance of the postmitotic state of terminally differentiated keratinocytes (110). In osteogenesis, loss of RB blocks late osteoblast differentiation (111). RB binds to and transactivates an osteoblast transcription factor RUNX2, and the complex is recruited to osteoblast-specific promoters (111). In adipogenesis, RB plays a dual role (112). RB blocks reentry into the cell cycle of preadipocytes, but subsequently gets hyperphosphorylated (inactive) facilitating cell-cycle exit required for terminal differentiation of adipocytes (112). Further studies support this dual role of RB in adipocyte differentiation; one study shows RB recruits histone deacetylase (HDAC3) to silence PPARγ target genes blocking adipocyte differentiation (113), and another study demonstrated positive regulation of terminal adipocyte differentiation via physical interaction of RB with CEBPs (114). Given that osteoblasts and adipocytes arise from mesenchymal tissue, a study investigated the role of RB in determination of these two fates (115). Rb1 loss in murine mesenchymal progenitor cells favored adipogenesis over osteogenesis and ultimately led to reduced levels of calcified bone and high levels of brown fat (115). These studies suggest that RB plays a critical role in development and more specifically, the contextual loss of RB greatly influences cell fate determination and differentiation.

While understanding the role of RB during development and the downstream impact of its loss of function has provided great insights into tumor initiation, it is equally important to understand the biological consequences of RB loss in adult differentiated tissue because RB alterations commonly occur in late-stage cancers correlating with treatment resistance. In mouse embryonic fibroblasts (MEF), loss of Rb1 or components of RB pathway leads to increased expression of stem cell markers such as Oct-4, Nanog, Sox2, and Klf4, and subsequently leads to formation of tumors and further differentiated cells in mice (116). Induction of stemness upon Rb1 loss in MEFs is consistent with previously published studies that reported loss of G₁ control and immortalization (117, 118). A recent study reported direct binding of Rb1 to Sox2 and Oct4 loci suppressing their expression and inhibiting the reprogramming of MEFs to induced pluripotent stem cells which was independent of Rb cell-cycle regulatory functions (119). In human fibroblasts, activation of caspase 3 and 8 leads to regulation of RB and subsequently activation of Oct-4 which is critical for the derivation of induced pluripotent stem cells (120). In murine postmitotic myocytes, transient suppression of Rb1 and Arf results in progenitor cells (myoblasts) that retain the capability to differentiate and form myofibers upon transplantation in vivo (121).

RB interacts with and recruits epigenetic modifiers to regulate transcription under various contexts (Fig. 4). RB can recruit histone deacetylases (HDAC) to repress E2F targets (122, 123). DNA methyltransferases can also be recruited to the RB–HDAC–E2F complex to repress target genes (124). RB can occupy repetitive sequences and recruit EZH2 to deposit repressive histone marks to silence these regions (125). When RB is lost, repeat sequences may be derepressed, which has been associated with lymphoma susceptibility in vivo (125). Repeat elements such as satellite sequences are found in heterochromatin regions at centromeres. In retinal pigment epithelial cells, RB1 loss led to defects in centromeric condensation and centromere dysfunction during mitosis and aneuploidy (126). Several independent studies have reported the role of RB in the formation of constitutive heterochromatin regions in the pericentric as well as telomeric chromatin (127–130).

While constitutive heterochromatin is critical for faithful chromatin segregation, facultative heterochromatin wherein repressive marks are deposited to silence set of genes upon specific developmental cues, is critical for cell-fate determination (131, 132). Interestingly, RB has been reported to mediate facultative heterochromatin formation upon cellular senescence (133) and can recruit heterochromatin protein 1 (HP1) to specific promoters to silence target genes (134). Moreover, emerging evidence suggest that RB can interact with chromatin insulators such as CTCFs and cohesin complexes, suggesting that they may play a role a broader role in nuclear organization and gene expression (135, 136). The noncanonical role of RB in chromatin reorganization and epigenetic dysregulation may relate to its canonical role in cell-cycle progression as chromatin undergoes elaborate disassembly during cell division and requires accurate reassembly for faithfully recapitulating lineage.

On the basis of the central role of cell-cycle dysregulation in cancer, there has been significant drug development focused on targeting regulators of cell cycle (Fig. 5). Three CDK4/6 inhibitors (palbociclib, ribociclib, abemaciclib) are FDA approved in breast cancer in combination with hormonal therapy. CDK4/6 inhibitors are in clinical trials for several other cancer types including prostate cancer, pancreatic cancer, lung cancer, lymphoma, glioma, and squamous cell carcinoma [reviewed in detail elsewhere (137)]. D-type cyclins (cyclins D1, D2, and D3) bind to CDK4 and CDK6 to specifically regulate the transition from the G₁- to S-phase by phosphorylating RB (138). Given the mechanism of action of CDK4/6 inhibitors, tumors with RB deficiency are not expected to respond. As a result, several clinical trials using CDK4/6 inhibitors (e.g., NCT01536743, NCT01976169, NCT02334527) have used biomarkers to exclude patients with RB pathway disruption (e.g., IHC for RB and p16). There are several ongoing studies that also use RB1 as a biomarker for patient selection (Table 1).

There are approaches to target mitotic regulators and other cyclin-CDKs in the context of RB deficiency. In preclinical studies focused on osteosarcoma, breast, and colon cancer, cells can be sensitized to cyclinA/CDK2 or cyclinE/CDK2 inhibition after disrupting RB–E2F interaction (139). In breast cancer, it was demonstrated that ablation of RB function results in activation of CDK2 which could be exploited for therapeutic targeting (140). Several other studies in acute myeloid leukemia, breast cancer, glioblastoma, and melanoma highlight the therapeutic potential of inhibiting CDK2 in the context of RB loss (141–146). In addition, substrates of CDK2 are dictated by the cyclins they are bound to (147). A recent study identified number of novel nuclear substrates for CDK2 which included epigenetic modifiers such as KDM1A, SETDB1, and DOT1L and DNA repair proteins like RAD54L and XRCC1 (148). Techniques for identifying specific substrates of kinases are rapidly improving which may further inform novel therapeutic strategies for inhibiting CDK2 in the context of RB-deficient cancers (149). With several degraders already in phase II clinical trials, proteolysis targeting chimeric (PROTAC) technology demonstrate a powerful platform for targeted degradation of challenging candidates (150, 151). Proof-of-principle studies using chimeric oligonucleotides in PROTAC technology also showcase capability of targeting DNA-binding proteins including transcription factors like E2F which may have potential applications in RB-loss tumors in the near future (152, 153).

A recent study sought to identify synthetic lethal candidates that can be targeted in RB1-deleted SCLC. The authors performed a CRISPR/Cas9 screen in an RB1-deleted SCLC cell lines that conditionally expressed RB1 and identified aurora kinase B (AURKB) as a potential dependency and therapeutic candidate (154). Aurora kinases (AURKA, AURKB, and AURKC) are involved in chromosomal segregation and are key regulators of mitosis. Inhibition of AURKB in the context of RB1 loss exacerbated mitotic abnormalities leading to polyploid formation and a dramatic decrease in tumor growth (154). Another independent study used a drug screen to identify targets that are synthetically lethal with RB1 mutation (155). The authors tested 36 cell-cycle inhibitors in over 500 cell lines from diverse epithelial, mesenchymal, and hematologic cancer lineages. Inhibiting AURKA had a significant effect on RB1-mutated cell lines from different lineages including lung cancer, breast cancer, myeloma, retinoblastoma, and glioblastoma (155). Aurora kinase A and B are also overexpressed in NEPC, and antitumor activity has been observed with aurora kinase inhibition in RB1-deficient NEPC preclinical models and patients (81, 156, 157). In breast cancer models that acquire resistance to CDK4/6 inhibition via RB dysregulation, aurora kinase inhibition along with targeting of another spindle assembly checkpoint kinase TTK has shown promising antitumor activity (158). HPV-driven cancers are often RB deficient, and aurora kinase inhibition induces cell death in HPV-positive cancers (159). There is an ongoing biomarker-selected clinical trial for RB1 loss head and neck squamous cell carcinoma (including HPV-driven) with the aurora kinase A inhibitor alisertib in combination with pembrolizumab (NCT04555837).

In addition to aurora kinases and TTK protein kinase, other checkpoint proteins and chromosomal segregation proteins have been reported as potential targets in the context of RB1-deficient tumors. Around 30% of TNBCs exhibit RB1 loss and a recent study showed DNA-damage checkpoint protein CHK1 and chromosomal segregation protein PLK1 are overexpressed in these tumors and their inhibition leads to significant tumor growth arrest in vivo (160). Of note, a phase II clinical trial (NCT02873975) is evaluating the efficacy of CHK1 inhibitor for advanced solid tumors that harbor alterations in homologous repair pathway, alterations that lead to replicative stress often associated with RB1 loss and CCNE1 amplification.

Three broad strategies have been proposed to target cancer plasticity [reviewed in detail elsewhere (161)]. Here we will specifically discuss these three strategies in the context of RB1 loss–mediated cancer plasticity (Fig. 5).

From basic studies in developmental biology, we understand that loss of RB1 leads to epigenetic dysregulation and changes in chromatin organization which may contribute to lineage reprogramming. Evidently, in RB-deficient preclinical models of prostate cancer and other plasticity models, blocking EZH2 leads to tumor growth arrest and changes in cellular reprogramming (76, 82). EZH2 inhibition also impacts SCLC plasticity by expanding non-neuroendocrine immunogenic fraction of SCLC, suggesting combination of EZH2 inhibition with STING agonism (162). Histone demethylases such as KDM5A and KDM1A (LSD1) have been demonstrated to repress NOTCH signaling in the context of RB1 loss and maintain neuroendocrine differentiation of SCLC tumorigenesis and have been proposed as therapeutic targets (163, 164). Changes in DNA methylation also occur in the context of RB1 loss and NEPC, which is being explored as biomarkers and targets (165, 166). RB1 loss leads to chromatin remodeling and certain subunits of the SWI/SNF complexes are dysregulated in NEPC that could be potentially targeted (167). PROTAC technology holds promise for targeting epigenetic mediators of lineage plasticity (e.g., PRC2 complex; refs. 168, 169).

Therapies that reverse lineage plasticity making the tumors regain dependence on their original oncogenic driver might make advanced tumors clinically manageable if they could be resensitized to targeted therapies. This approach is also related to blocking epigenetic drivers of plasticity. For instance, blocking HDAC and 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR) in EGFR-mutant treatment-resistant lung adenocarcinoma undergoing plasticity restores sensitivity to tyrosine kinase inhibitors (170). In RB1-loss and AR-low plasticity models in prostate cancer, EZH2 inhibition leads to an increase in AR signaling which can be combined with an AR antagonist to achieve synergy (76, 82). However, reexpression of AR is not always seen in AR-negative NEPC models after EZH2 inhibition (81) potentially due to a more terminally differentiated state and highlighting the context specificity of such therapies. There are ongoing clinical trials that combine EZH2 inhibitors with AR antagonists (e.g., NCT04179864, NCT04104776) albeit these trials are not selected for RB1 loss.

Unraveling novel drivers or dependencies of the new cell identity is quite challenging yet an exciting strategy to tackle cancer plasticity. Recent studies highlighted the dependency on phospholipid glutathione peroxidase GPX4 in tumors with high mesenchymal cell state (171). Similarly in receptor tyrosine kinase, AXL has been demonstrated as a vulnerability that can be targeted in EGFR-mutant treatment-resistant lung adenocarcinoma plasticity. In the context of RB1-mediated lineage plasticity in prostate cancer, novel drivers such as BRN2 and ONECUT2 have been identified as targetable master regulators (83, 172). Other transcriptional mediators of prostate cancer lineage plasticity such as SOX2, SOX11, ASCL1, INSM1, NEUROD1, HOXB5, HOXB6, and NR1D2 have been identified and maybe exploited as therapeutic targets (78, 79, 173). Expression of lineage-specific markers that emerge after transdifferentiation, such as DLL3 in the context of small-cell transformation, is also opening new avenues for targeting (174). Large-scale CRISPR screens spearheaded by the Cancer Dependency Map project offer promising opportunities to further identify novel druggable candidates that are shared across de novo and acquired RB1-loss cancers.

Functional loss of RB not only dysregulates cell cycle but also has profound effects on genome stability and epigenetic reprogramming, ultimately impacting cancer plasticity. Phenotypic plasticity is associated with metastatic progression, aggressive disease, and acquired resistance to therapy. A better understanding of the genetic and epigenetic drivers of cancer plasticity, and emerging clinical biomarkers such as RB1 loss, provides new opportunities for therapeutic interventions. The goals of therapy may be context driven to target cell cycle or acquired lineage programs to reduce tumor burden, reverse or prevent plasticity, and/or resensitize to targeted therapies. Ultimately, the translation of novel strategies for targeting RB loss and cancer plasticity will require interdisciplinary commitment at the intersection between molecular and developmental biology and clinical/translational oncology research.

H. Beltran reports grants from Bristol Myers Squibb, Circle Pharma; grants and personal fees from Diacchi Sankyo; personal fees from Janssen, Merck, Pfizer, Foundation Medicine, Blue Earth, Amgen, Bayer, LOXO Eli Lilly, Sanofi, Curie Therapeutics, AstraZeneca, and Novartis outside the submitted work. No disclosures were reported by the other authors.

We apologize to the many researchers whose work we were unable to cite because of space limitations. We thank members of the Beltran lab for insightful discussion of topics relevant to this review. V.B. Venkadakrishnan is supported by DoD PCRP Early Career Investigator Award (W81XWH2210197), and National Cancer Center Postdoctoral Fellowship Award. Y. Yamada is supported by the Japan Society for the Promotion of Science. K. Weng and O. Idahor were supported by Dana-Farber/Harvard Cancer Center Young Empowered Scientists for ContinUed Research Engagement (YES for CURE) program funded by NIH/NCI grant R25CA221738. H. Beltran is supported by the Prostate Cancer Foundation, DoD PCRP (W81XWH-17-1-0653) and NIH/NCI (R37CA241486-01A1, P50 CA211024-01A1).