Targeted Inhibition of the E3 Ligase SCF^Skp2/Cks1 Has Antitumor Activity in RB1-Deficient Human and Mouse Small-Cell Lung Cancer

Small-cell lung cancer (SCLC) is characterized by aggressive growth, frequent metastases, and a 5-year survival of less than 5% (1, 2). While it can be sensitive to first-line therapy of cisplatin and etoposide, most patients rapidly relapse with chemotherapy-resistant tumors. Dozens of drugs have been tested clinically in SCLC, including more than 40 agents that have failed in phase III trials. The identification of driver mutations and their corresponding targeted drugs have led to significant improvements in the treatment of non–small cell lung cancer (NSCLC) and other solid tumors; however, similar advances have not been made in the treatment of SCLC. Except for recently reported modest increases in survival seen with the PD-L1 inhibitors nivolumab and atezolizumab, neither treatment options nor life expectancy have improved over the past three decades (1–4).

A unique feature of SCLC, not seen in any other solid tumor types, is the near uniform (>95%) biallelic inactivation of tumor suppressor genes RB1 and TP53 to drive tumorigenesis (1, 2). This defining feature of the disease has not led to a targeted therapy, however, because genetically inactivated RB1 and TP53 cannot be reactivated, nor is it feasible to clinically reintroduce the wild-type genes into all tumor cells in vivo. The RB1 tumor suppressor gene governs multiple cellular functions, including proliferation, cell-cycle progression, and apoptosis, via a complex collection of molecular actions (5, 6). In lung cancers, loss of RB1 was associated with higher-grade tumors and an acceleration of metastatic competency (7). pRb regulates gene expression by binding to and suppressing the E2F1 transcription factor. The E2F family of transcription factors is important for cellular homeostasis, and they are the major transcriptional regulators of cell-cycle–dependent gene expression, particularly those genes required for the G₁–G₀ to S-phase transition (8). Hence the loss of pRb leads to high E2F transcriptional activity, compromises the ability of cells to exit the cell cycle, and makes them highly susceptible to sustained proliferation in the presence of activated oncogenes (6).

While the ability of pRB to bind to E2Fs has been the focus of much research, protein interaction databases indicate that there are more than 300 proteins that might interact with pRB (6). For example, pRB exerts significant cell-cycle control that is transcription-independent, due to its well-characterized regulation of protein stability by direct effects on the ubiquitin-ligase proteasomal degradation pathway. This degradation pathway includes the SCF family of E3 ubiquitin ligases, which play essential roles in the ubiquitination, degradation, and regulation of cellular protein turnover (9). SCF complexes consist of a scaffold protein (Cul1), an adaptor protein (Skp1), an accessory protein (Cks1 aka Roc1), and an F-box component; the latter determines the recognition specificity of the protein substrate(s) for ubiquitination.

Our group previously identified one of the SCF E3 ligases, SCF^Skp2/Cks1, as an early repression target of pRB, and found that knockout of the Skp2 substrate recruiting subunit of SCF^Skp2/Cks1 complex effectively blocked tumorigenesis of pituitary melanotroph tumors and thyroid C cells in Rb1^+/− mice (10–13). Primary targets of SCF^Skp2/Cks1 (called Skp2 in this article) include the cyclin-dependent kinase inhibitor p27 (CDKN1b), a key cell-cycle regulator, which delays progression from G₁ phase into S-phase. The levels of p27 protein decline prior to cell entry into S-phase due to its Skp2-mediated ubiquitination and degradation. By binding to Skp2, pRb prevents the binding, ubiquitination, and degradation of p27 by the SCF^Skp2/Cks1 complex. In cells that have lost functional pRB, therefore, there is no restraint on Skp2-mediated p27 degradation, which leads to the loss of key G₁–S cell-cycle restriction point and promotes cell proliferation. In fact, in some studies, cell-cycle control has been shown to be better correlated with p27 levels than with changes in proteins encoded by E2F-regulated genes (3). pRB also represses Skp2 mRNA expression, and thus RB1 loss can lead to increased Skp2 levels that could further promote p27 degradation.

In this report, we show that loss or inhibition of Skp2 can restore some of the tumor suppressor actions of pRB in SCLC. The antitumor activities of small molecules that directly inhibit Skp2 activity or prevent formation of an active SCF^Skp2/Cks1 complex highlight potential actionable targets in cancer cells that have lost RB1. We have validated the antitumor activity of the Skp2 inhibitors in vitro and in vivo, using both human and mouse SCLC models. While previous studies have evaluated the therapeutic potential of Skp2 inhibitors, those studies focused on other tumor types or on the subset of tumors in which Skp2 was overexpressed (9, 14). In contrast, Skp2 inhibitors will theoretically have therapeutic effects on all SCLC, as RB1 loss and presumed increased Skp2 activity occur in virtually all SCLC tumors (1, 2).

Establishment of SCLC mouse models by intratracheal delivery of Adeno-CMV-Cre has been described previously (15). Rb1^lox/lox;Trp53^lox/lox;Skp2^+/+ mice, Rb1^lox/lox;Trp53^lox/lox;Skp2^−/− mice and Rb1^lox/lox;Trp53^lox/lox; p27^T187A/T187A mice were used to establish SCLC models. Mice about 8 weeks of age were anesthetized with ketamine/xylazine and tumors were initiated by intratracheal delivery of 75 μL of DMEM/F12 medium containing 2.5 × 10⁷ PFU Adeno-CMV-Cre (prepared by the Albert Einstein College of Medicine Gene Therapy Core) and 10 mmol/L CaCl₂.

All animal procedures were reviewed and approved by the Einstein Institutional Animal Care and Use Committee (IACUC).

CT imaging was performed with an X-ray CT system (Latheta LCT-200, Hitachi Aloka Medical). Mice were anesthetized with isoflurane and imaged without any contrast reagent. Parameters used for the CT scans were as follows: tube voltage: 50 kV; tube current: 0.5 mA; axial field of view (FOV): 48 mm, with an inplane spatial resolution of 48 μm × 48 μm and slice thickness of 100 μm. Qualitative analysis of lung lesion areas was performed with LaTheta software (version 3.00).

Primary mouse RP-Lung cells and RP-LvMet cells were prepared from 0.3 cm × 0.3 cm pieces of lung or liver tumor tissues, which were minced and dissociated in collagenase A in 2 mL serum-free DMEM/F12 (Corning) for 30–60 minutes at 37°C with gentle shaking. Then sample was diluted to 10 mL in serum-free DMEM/F12 and spun at 200 × g for 5 minutes. The cell pellet was resuspended in 1 mL trypsin (Gibco) and placed in a 37°C, 5% CO₂ tissue culture incubator for 3 minutes. Sample was diluted into 20 mL serum-free DMEM/F12 and filtered through a 45-μm nylon cell strainer into a new tube. After centrifuging (200 × g) for 5 minutes, the pellet was resuspended in 5 mL of red cell lysis buffer and incubated for 5 minutes at room temperature. Cells were spun at 200 × g for 5 minutes, and resuspended with cell culture medium (DMEM/F12 containing 10% FBS, 1% penicillin/streptomycin, and 1% glutamine). Cell genotypes were confirmed by PCR of Rb1, Trp53, Skp2^−/− and p27^T187A/T187A. Human SCLC cell lines (H69, H146, H196, H446, and H720) and human NSCLC cell lines (H520 and H460) were obtained from ATCC and were cultured in RPMI1640 (ATCC) containing 10% FBS, 1% penicillin/streptomycin, and 1% glutamine.

A primary SCLC tumor from an RP mouse was dissociated as described above, and 2000 cells were embedded in growth factor–reduced Matrigel in 96 wells with DMEM/F12 medium containing mitogens (EGF, FGF2, and FGF10), R-Spondin (Wnt/β-catenin signaling agonist), Noggin and A83-01 (TGFβ inhibitors), and Y-27632 (Rock inhibitor). The frequency of organoid forming units in the dissociated suspension was approximately 0.1%, and this increased at least 10 fold on the first passage (p1). Salient feature of the RP SCLC organoids were rapid growth of tightly packed small-sized cells.

To determine the effects of the drugs on cell proliferation, cells were plated overnight in 96 well plates at approximately 5 × 10³ cells per well in 100 μL. They were treated with vehicle, compound C1 (Millipore Sigma), flavokawain A (FKA; LKT Labs), or pevonedistat (DCTD/NCI and Millennium Pharmaceuticals) at various concentrations. After 72 hours, cell numbers were determined with a CellTiter-Glo 2.0 Luminescent Cell Viability Assay reagent (Promega). Plates were read for luminescence using a Fluostar Optima Luminometer (B.M.G. Labtech) and the IC₅₀ values were determined.

To assess the effect of shRNA vectors on cell proliferation, 2 × 10⁵ cells were plated in 6-well plates in triplicate and treated with or without doxycycline. Cell numbers were determined by hemocytometer counting.

Doxycycline-induced knockdown of Skp2 was done with a pTripZ lentiviral shRNA obtained from Dharmacon. The shRNA sequence for Skp2 knockdown in mouse primary RP-LvMet cells and human H520 cells were 5′-GCAAGACTTCTGAACTGCTAT-3 and 5′-TCAAATTTAGTGCGACTTA-3′, respectively. An empty vector was used as a control. Lentiviral helper constructs psPAX2 and pMD2.G were gifts from Didier Trono (Addgene plasmid #12260 and #12259). Lentivirus stocks were generated and concentrated as described previously (16). For lentivirus infection, cells were put in 1 mL virus-containing media (DMED/F12) with 8 μg/mL polybrene, and were then spun at 1,000 × g for 30–60 minutes at 32°C. Cells were resuspended with fresh media and transferred into a culture plate and placed into a 37°C, 5% CO₂ tissue culture incubator. Successful lentivirus transduction was ensured by puromycin (400-128p, GeminiBio-Products) selection, followed by mRNA (qRT-PCR) and protein (Western blot) measurements. Doxycycline (2 μg/mL; Sigma-Aldrich) was used to induce Skp2 knockdown in cultured cells. Primary cells from RP mice were never maintained in culture longer than 10–14 days.

RNA was extracted using RNeasy Mini Kit (74104, QIAGEN). Oligo-dT and SuperScript II (Invitrogen) were used for the synthesis of the first-strand cDNA at 42°C for 60 minutes. SYBR green PCR mixture (4309155, ABI) and the standard program of ABI 7500 Fast real-time PCR were used. GAPDH was used as internal control. The qPCR primers sequences were:

• mouse GAPDH-Forward: 5′-AATGTGTCCGTCGTGGATCT-3′;

• mouse GAPDH-Reverse: 5′-GGTCCTCAGTGTAGCCCAAG-3′,

• mouse Skp2-Forward: 5′-AGCAGCCGCTGGGTGAAAGC-3′;

• mouse Skp2-Reverse: 5′-ATCACTGAGTTCGACAGGTCCAT-3′;

• human GAPDH-Forward: 5′-GGCCTCCAAGGAGTAAGACC-3′;

• human GAPDH-Reverse: 5′-AGGGGTCTACATGGCAACTG-3′,

• human Skp2-Forward: 5′-TCCACGGCATACTGTCTCAG-3′;

• human Skp2-Reverse: 5′-GGGCAAATTCAGAGAATCCA-3′.

The qPCR reactions were performed in a final volume of 20 μL. qPCR data were analyzed using ΔΔC_t analysis method. All were done in triplicates and performed three separate times.

Orthotopic SCLC models were established by intratracheal delivery of 1 × 10⁶ primary mouse RP-LvMet cells (infected with doxycycline-inducible shSkp2 lentivirus) into 6–8 weeks athymic nude mice (strain code: 490, Charles River). Fifteen days later, mice were randomly divided into two groups for daily gavage treatment with 200 μL of vehicle (control) or doxycycline (10 mg/mL). Primary mouse RP-LvMet cells (infected with doxycycline-inducible Skp2 knockdown lentivirus) were used to establish subcutaneous (s.c.) tumors by injection of 1 × 10⁶ cells into the flanks of athymic nude mice (strain code: 490, Charles River). When the tumor size was about 100–300 mm³, mice were divided into a control group and a group that was treated with doxycycline (5 μg/mL doxycycline in their drinking water plus daily oral gavage of 1 mg doxycycline), FKA (600 mg/kg/day by oral gavage), or pevonedistat (50 or 90 mg/kg/day s.c.). Similarly, 1 × 10⁶ H69 cells were injected subcutaneously into nude mice that were then treated daily with vehicle or 40 mg/kg/day C1 (in DMSO) by intraperitoneal injection when the tumor size was about 100–300 mm³. For patient-derived xenograft (PDX) models, 1 × 1 mm tumor pieces of tumors CTG-0199 or CTG-1252 (Champions Oncology) were inoculated sc into NCG mice (strain code: 572, Charles River). When tumor volume reached about 100 mm³, mice were randomized to vehicle or 600 mg/kg/day FKA treatment by gavage.

Daily drug treatments continued for the duration of the experiments. The mice were monitored daily, tumor diameter was measured using calipers, and tumor volume was calculated as (length × width²) × 0.526.

Cells or tumor masses (0.3 cm × 0.3 cm) were lysed in RIPA buffer. Protein concentrations were determined by Bio-Rad protein Assay Kit using SmartSpec 3000 Spectrophotometer for equal loading by protein content onto SDS-PAGE gels. Proteins were transferred to polyvinylidene difluoride (PVDF) membrane (IPVH00010, Millipore) and probed with the following antibodies: Skp2 (sc-7164), p21 (sc-397), and α-tubulin (sc-8035), from Santa Cruz Biotechnology; cleaved caspase-3 (#9661) and p53 (#2524) from Cell Signaling Technology; cullin-1 (ab75812) and p73 (ab40658) from Abcam; p27 (#610242, BD Biosciences), and pRb (554136, BD Pharmingen).

For coimmunoprecipitation assays, cells were lysed in Nonidet P-40 buffer [50 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 5 mmol/L EDTA, 0.5% Nonidet P-40, and 10% glycerol] containing phosphatase and protease inhibitors. Precleared extracts were incubated with rabbit anti-Skp2 antibody. Immunocomplexes were recovered using protein A-agarose (P9269, Sigma), separated on SDS-polyacrylamide gels, and transferred to PVDF membrane. For Western blotting, the primary antibodies used were anti-Skp2 antibody and anti-p27 antibody. The secondary antibody was horseradish peroxidase–labeled mouse anti-rabbit IgG (#3678S, Cell Signaling Technology). Proteins of interest were detected using chemiluminescence (NEL104001EA, Perkin Elmer).

For IHC staining, tissues were fixed in 10% formalin, paraffin embedded, and sectioned at 5-μm thickness. Slides were deparaffinized, hydrated, and incubated in a steamer for 20 minutes in sodium citrate buffer (Vector Labs) for antigen retrieval. Sections were first treated with 3% H₂O₂ for 10 minutes to quench endogenous peroxidase, washed with PBS for several times, blocked with 10% normal goat serum for 1 hour, and then incubated in primary antibodies at 4°C overnight. The following antibodies were used: Ki67 (ab16667, Abcam,), cleaved caspase-3 (#9664S, Cell Signaling Technology), synaptophysin (ab32127, Abcam), and chromogranin A (ab45179, Abcam). The sections were treated with SignalStain Boost IHC detection reagent (Cell Signaling Technology) for 30 minutes. After three washes with PBS, the sections were incubated with SignalStain DAB substrate Kit (Cell Signaling Technology) was used to detect signals. IHC staining were counterstained with Harris Hematoxylin (Poly Scientific R&D Corp). Images were visualized with a Nikon Eclipse Ti-U microscope and captured with Olympus DP71 camera and DP Controller software (3.2.1.276), and saved with DP manager software (3.1.1.208). The images were further processed by Adobe Photoshop. Cleaved caspase-3 quantification was done using ImageJ software.

In the survival analysis, differences in Kaplan–Meier survival curves were analyzed by log-rank tests (Prism 6 Software). P < 0.05 was considered as statistically significant.

To study the role of Skp2 in SCLC, we used a mouse model (Rb1^lox/lox;Trp53^lox/lox) in which both alleles of the Rb1 and Trp53 genes are somatically deleted in the lung by intratracheal administration of Adeno-Cre (17). These mice develop aggressive and metastatic SCLC with histologic, pathologic, and neuroendocrine features that closely resemble advanced human SCLC (18). We generated cohorts of Rb1^lox/lox;Trp53^lox/lox;Skp2^+/+ (RP Skp2-WT SCLC) mice and littermate Rb1^lox/lox; Trp53^lox/lox;Skp2^−/− (RP Skp2-KO) mice. At 8 weeks of age, Adeno-CMV-Cre was administered intratracheally to the mice to delete Rb1 and Trp53, and initiate SCLC tumorigenesis (15). Male and female mice showed equal susceptibility to SCLC and our cohorts consist of equal numbers of both (17). At 4–6 months post Adeno-Cre inhalation, typical in situ lung lesions are found protruding into the mid-airway space, and these have small-cell morphology (scant cytoplasm), and stain positive for synaptophysin (Syp), a neuroendocrine SCLC marker, and Ki67 (Fig. 1A). Lungs were also examined by CT scans (Supplementary Fig. S1A and S1B). At 8–10 months, multiple nodular tumors with SCLC morphology are detected throughout the lungs and the liver (Fig. 1B and F; Supplementary Figs. S1A and S2A; Supplementary Table S1) as indicated by morphology, and synaptophysin and chromogranin A staining. The incidence of SCLC in the lung and livers of the RP mice with wild-type Skp2 was 91% and 74%, respectively (Table 1). Both DNA (Supplementary Fig. S2B) and Western blot analyses (Supplementary Fig. S2C) confirmed the lung tumors were deficient for Rb1 and Trp53. The liver metastases in the RP Skp2-WT mice stained positive for synaptophysin and chromogranin A, indicating they originated from the lung SCLC tumors (Fig. 1F). In contrast, RP Skp2-KO mice did not develop primary SCLC or SCLC liver metastasis, even at 24 months after Adeno-Cre administration (Fig. 1C; Supplementary Fig. S1B and S2D; Table 1). The lung tumors in these mice were negative for synaptophysin and chromogranin A, but stained positive for the epithelial marker cytokeratin (Fig. 1G).

Because Skp2 targets a number of protein substrates for ubiquitination, we determined which of these are key to SCLC development in the RP mice. As noted above, p27 is a well-established Skp2 substrate, and it must be phosphorylated on T187 for recognition by, and binding to, the Skp2/Cks1 pocket. To test the role of p27 in Skp2-dependent SCLC, we generated RP mice with a knock-in of a nonphosphorylatable T187A-p27 (RP p27-AA; ref. 19). Because p27^T187A cannot bind to and be ubiquitinated by Skp2, it will not be subject to degradation in RB1-mutant cells. As shown in Fig. 1D and G, Supplementary Fig. S2E and Table 1, there were no lung tumors with SCLC morphology or liver metastasis in these mice. We did find that mice in all three groups developed NSCLC, probably due to loss of Trp53 in other lung cell types (Fig. 1E and Table 1).

Both RP Skp2-KO and RP p27-AA mice survived significantly longer than RP Skp2-WT mice, with median survivals of 19 and 16 months for RP Skp2-KO and RP p27-AA mice, respectively, compared with 13 months for RP Skp2-WT mice (Fig. 2A). However, this difference was not as dramatic as that seen in our studies of Rb1-deficient pituitary melanotroph and thyroid C cell cancer (11). The presence of NSCLC in the mice complicates the interpretation of the survival data. In agreement with a previous report, the NSCLC tumors had morphologies consistent with large cell, anaplastic cancer with giant cells, as well as with adenocarcinomas, squamous carcinomas, and tumors with mixed morphologies (Table 1; ref. 18). While the incidence of NSCLC in the RP Skp2-WT mice (13.1%) was consistent with the previous report, the Skp2-KO and p27AA mice had higher incidences of these lung tumors (75% and 53%, respectively). Other malignancies, mainly with characteristics of leukemia and lymphomas, were also observed;however, these were likely unrelated to the RP genotype (Table 1). There was a higher incidence of the other tumors in the latter two groups as well (4.3% vs. 19% and 18%). It is possible that the longer survival of these mice, due to the lack of SCLC tumors, allowed the eventual development of NSCLC and other tumor types, and this may have been the cause of death in these mice. None of the NSCLC tumors metastasized to the liver in any of the groups, although the other tumor types did.

While the lack of SCLC in RP Skp2-KO mice supports Skp2′s essential role in SCLC tumorigenesis, these experiments do not address whether downregulation of Skp2 in established tumors would have a beneficial effect, and could therefore serve as the basis for targeted drug therapy. To answer this question, we avoided the long latency and confounding NSCLC and other tumor development of the RP transgenic mouse model, and used instead subcutaneous implantation of primary RP-LvMet cells (isolated from a SCLC metastatic liver tumor from an RP mouse) to establish SCLC. During the brief period (< 2 weeks) that these cells were cultured, we infected them with an inducible Skp2-shRNA lentivirus construct. In tissue culture, doxycycline caused knockdown of Skp2, an increase in p27, a dramatic decrease in cell proliferation, loss of viability, and the induction of cell apoptosis (Fig. 2B and C). These cells were inoculated subcutaneously in nude mice, and after allowing tumors to grow for 10 days to approximately 100–300 mm³, we treated the mice with doxycycline to knockdown Skp2 expression. The doxycycline-mediated knockdown of Skp2 had an antitumor effect in the subcutaneous xenograft mouse model, as illustrated by a significant decrease in tumor growth rates (Fig. 2D). Skp2 knockdown in these tumors was not complete (Fig. 2E), which may account for the partial antitumor effect in vivo.

Mice were also inoculated orthotopically into the lung with the inducible Skp2-shRNA RP-LvMet cells. The orthotopic tumors (Supplementary Fig. S3A and S3B) showed SCLC pathology indistinguishable from an autochthonous SCLC (Supplementary Fig. S3C and S3D). After allowing 15 days postinoculation for the tumors to become well established, the mice were treated with doxycycline until they were sacrificed upon morbid appearance. Strikingly, while multiple large lesions were seen in the livers of 9 of 9 control mice in the absence of doxycycline, only 2 of the 9 mice had liver metastases seen after Skp2 knockdown, and these liver metastases were smaller than those seen in the control mice (Fig. 2F). Whether this is a true anti-metastatic effect is uncertain, however, as it may reflect the lower primary tumor burden in the lungs of the Skp2 KD mice. These findings are consistent with a recent report in which the reactivation of the RB pathway in RB1-deficient advanced lung adenocarcinomas caused their reprogramming toward a less aggressive state and made them unlikely to metastasize (7).

Our results show that SCLC is very sensitive to Skp2 knockdown. To determine whether this effect occurs in lung cancers that have not lost RB1, we infected H520 NSCLC cells, which have wild-type RB1, with the same inducible Skp2-shRNA construct used in the SCLC cells. In contrast to the SCLC cells (Fig. 2B), Skp2 knockdown by doxycycline did not affect the proliferation of the H520 cells (Fig. 2G) despite causing a significant knockdown of Skp2 (Fig. 2H). Infection of the H520 cells with a constitutive shSkp2 similarly did not affect the proliferation of the cells.

Having successfully employed genetic methods to define the role of Skp2 in SCLC, we next used pharmacologic approaches to validate Skp2 as a potential target for drug therapy. The crystal structure of the SCF–Skp2–p27 complex has defined the interaction of Skp2 and Cks1 to form a pocket to which p27^T187 binds (14, 20–24). On the basis of this information and using in silico modeling and virtual library screening, studies investigating inhibitors of Skp2 identified a small molecule (called C1) that binds to the pocket, thereby preventing binding of p27 to the Skp2/Cks1 pocket (Fig. 3A; ref. 14). As shown in Fig. 3B, C1 inhibits proliferation of the human SCLC cell line H69 and two primary cell lines we derived from a lung tumor and a liver metastasis in the RP mice (RP-Lung and RP-LvMet), but was 6- to 50-fold less active against H460 and H520 NSCLC cells. The average IC₅₀ for C1 in a larger panel of SCLC cell lines was 2.78 μmol/L, which was 2.6- to 12-fold lower than that for the NSCLC cells (Table 2). C1 increased p27 protein in human and murine SCLC cells but not in the two NSCLC cell lines (H460 and H520; Fig. 4A). Similarly, C1 induced apoptosis (cleaved caspase-3) in the SCLC cells but not in the NSCLC cells (Fig. 4A). While C1 has previously been tested against melanoma, breast, and prostate cancer cells, our study is the first to report the activity of C1 against SCLC cells (14).

We next investigated a second approach to reduce Skp2 activity, based on the requirement of the SCF complex to be neddylated for its assembly and stability (Fig. 3A; ref. 6). Neddylation is mediated by the neddylation activating enzyme (NAE), which first helps release cullin from its inhibitory protein CAND1, and then transfers NEDD8 (a small ubiquitin-like protein) to cullin, which induces a structural rearrangement of cullin and the SCF complex, allowing the transfer of ubiquitin from the E2 enzyme to the Skp2 protein substrate (24). We used two NAE inhibitors: flavokawain A (FKA), a chalcone derived from the kava plant that has been previously studied for its antitumor effects in a number of cancer types, and pevonedistat (MLN4924), a first-in-class inhibitor of NAE that has entered clinical trials (25–27).

We found that both FKA and pevonedistat inhibited the proliferation of the H69, RP-Lung, and RP-LvMet SCLC cells (Table 2). In addition to inhibiting proliferation, pevonedistat caused cell death in preformed SCLC organoids, indicating its effects were cytotoxic and not merely cytostatic (Supplementary Fig. S4). In contrast to C1, however, these agents showed less selectivity for SCLC, compared with the two NSCLC cell lines. As expected, both FKA and pevonedistat decreased the neddylation of Cul 1 (Fig. 4B) and increased p27 and p21 (Fig. 4C) in the SCLC cells. Interestingly, pevonedistat induced a large increase in Skp2 protein levels, in contrast to the previously reported effect for FKA (Fig. 4C; ref. 27). It is presumed that this Skp2 is not part of a SCF^Skp2 complex, as the assembly of the complex requires neddylated Cul 1, and therefore lacks E3 ligase activity, which then leads to the large increase in p27. We confirmed this with coimmunoprecipitation assays showing that pevonedistat treatment interrupted the interaction between Skp2 and p27 (Fig. 4D). The differences in Skp2 levels between pevonedistat and FKA treatment may reflect the greater specificity for NAE of the former compound, as FKA's activity is not restricted to the effects on NAE (27).

Previous studies of panels of drugs on a large number of human SCLC cell lines found that the sensitivity of the cells to the drugs was highly correlated with their response to etoposide, that is, cells that were resistant to etoposide were generally resistant to all other agents (28). We found this not to be the case for C1 or pevonedistat, however, as there was no significant correlation with their IC₅₀s compared with their IC₅₀ for etoposide (Fig. 4E). The possibility that the combination of C1 and etoposide could have synergistic antitumor activity was also investigated; however, no evidence for such combinatorial effects were observed. Although combining C1 with pevonedistat in the SCLC cell lines did cause a significantly greater growth inhibition than either agent alone, there was also no evidence that this interaction was synergistic (Fig. 4F).

Having demonstrated an inhibitory effect of the genetic knockdown of Skp2 on SCLC in vitro and in vivo, and the pharmacologic effect of Skp2 inhibition on SCLC in vitro, we next evaluated the antitumor effects of the Skp2 inhibitors in vivo. The three compounds, C1, FKA, and pevonedistat, had significant antitumor activity against SCLC subcutaneous xenografts in vivo. These included the H69 human cell line (Fig. 5A), primary liver metastasis cells from the RP mouse (RP-LvMet; Fig. 5B and E), and two human SCLC PDXs (Fig. 5C and D). Both PDXs had mutant RB1 and TP53, had characteristic small-cell morphology, and expressed SCLC neuroendocrine markers, as did the tumors derived from H69 and RP-LvMet (Fig. 5F; Supplementary Fig. S5). Inhibition of tumor growth by the drugs was at least 50% in all cases, with increases in apoptosis and evidence of a cessation of growth with prolonged drug treatment (Fig. 5A–J). RP-LvMet tumors from pevonedistat-treated mice had significant increases in Skp2, p27, and cleaved caspase-3 levels (Fig. 5G and H). It is noteworthy that FKA was active in the CTG-0199 PDX (Fig. 5C), which was derived from a patient with SCLC who had previously received carboplatin and paclitaxel chemotherapy, indicating that the Skp2 inhibitors may be effective as a second-line therapy in chemo-resistant tumors. There was no change in body weight in mice treated with C1 or pevonedistat (Fig. 5K).

Despite having a high number of mutations and genomic alterations, second only among solid tumors to melanomas, SCLC has few obvious therapeutic targets for drug development, and no targeted drugs have been shown to have clinical activity (29). While recent FDA approval of PDL1 inhibitors represent the first new drugs approved for SCLC in over 20 years, their activity was modest, and the clinical activity of checkpoint inhibitors seems to be less pronounced in SCLC than in other tumors with similarly high mutation burdens (1, 4). Recent genomic and proteomic profiling of SCLCs have identified potential new targets, including genes involved in DNA repair, histone methylation, and the Notch pathway (1, 2). However, each of these may be suitable targets in only small subpopulations of patients with SCLC. In contrast, mutations and complex genomic translocations led to biallelic inactivation of RB1 in 106 0f 108 human SCLC tumors that were examined, and the remaining two tumors had massive genomic rearrangements (chromothripis) accompanied by the loss of pRB expression (30). The importance of the universal inactivation of RB1 in SCLC was confirmed in genetically engineered mouse models, in which the somatic biallelic deletion of Rb1 and Trp53 was both required and sufficient for the development of aggressive and metastatic neuroendocrine lung cancers (18).

The activity of pRb is controlled by cyclin-dependent kinases, and high levels of active pRb are found in G₁ phase and quiescent cells (31). While the role of pRb in regulating cell-cycle arrest at the G₁–S restriction point has been well-established, questions have arisen as to whether this regulation is solely due to pRb's effect on E2F transcription factors and the transcription of cell-cycle genes (6). Pinpointing the specific function of pRb, whose loss is critical for driving tumorigenesis in RB1-deficient cells, is an essential first step for the identification of a downstream pathway for drug targeting. Our previous reports that the loss of Skp2 was sufficient to prevent tumorigenesis in Rb1 knockout mice supported a critical role for a transcription-independent action of pRb, that is, the regulation of protein stability by the ubiquitin–proteasome system (12, 19).

pRb directly binds to the N terminus of Skp2 in the SCF complex, which causes the repression of Skp2-E3 ligase activity. pRb binding blocks the binding of p27 to Skp2-SCF, prevents the ubiquitination and degradation of p27, and maintains cell-cycle arrest. Hence the loss of RB1 results in unregulated activation of Skp2, unregulated loss of p27, and reduced cell-cycle control. The importance of this action of pRb was shown using a pRb mutant that could not bind to E2F but retained full activity in inhibiting the Skp2–p27 interaction and inducing cell-cycle arrest (3). In the absence of RB1, Skp2 can act as an oncogene, a conclusion supported by findings that Skp2 gene is amplified in 44%, and overexpressed in 75% of human RB1-mutant SCLC (32, 33). Although loss of pRb leads to E2F1 activation, we previously found that cyclin A, a target of E2F1, restrained the activity of E2F1. When Skp2 is codeleted with Rb1, accumulated p27 sequesters cyclin A away from the E2F1 promotor, which not only potently activates E2F1 target genes but also converts an oncogenic E2F1 into a proapoptotic E2F1 via activating p73 expression (10, 34, 35). Increased accumulation of the Skp2 degradation target p27 transforms the oncogenic pRb repression target E2F1 to an apoptotic E2F1, revealing synthetic lethality between Rb1 and Skp2, which remained unabated when Trp53 was additionally deleted (10, 13).

Understanding the nature of the interaction between Skp2, pRb, and p27 is valuable not only for defining the basis for the tumor suppression actions of RB1, but also to aid in the identification of potential target sites for drug design. The assembly of four components forms a functional SCF^Skp2: Skp2, its adaptor protein (Skp1), its accessory protein (Cks1), and its substrate (p27). While SCF^Skp2 lacks the traditional catalytic sites of other enzymes that allows for direct inhibitor design, it does have biochemically distinct potential drug–binding pockets that could serve as targets for small-molecule design (9, 36). The design of drugs targeting these pockets requires detailed structural analysis of the relevant protein–protein interactions, and this knowledge is incomplete as there are currently no X-ray cocrystal structures of small-molecule inhibitors bound to SCF^Skp2.

Data that are available suggest our understanding of the complex, based on the crystal structure of the Skp1/Skp2/Cks1 subunits combined with a peptide representing the binding region of p27, may not fully reflect the in vivo structure (36). Nevertheless, a suitable pocket was identified at the p27–Skp2–Cks1 interface, and a molecule (C1) that binds to this pocket and inhibits Skp2 ligase activity has been identified using virtual library screening (14, 36). While other potential binding pockets and inhibitory small molecules have been reported, C1 is one of only two molecules that docked to its expected target site using ICM-Dock, a docking algorithm shown to have the best accuracy when compared with other algorithms (36). Our finding that C1 significantly increases p27 levels in SCLC cells is consistent with physical data showing that C1 binds specifically to the Skp2–Cks1 protein complex, that this binding sterically clashes with the interaction of p27 with Skp2-Cks1, and that it causes a concentration-dependent inhibition of the ligase activity of SCF^Skp2 in a cell-free system (14). Our finding of a 2.6- to 12-fold greater sensitivity to C1 of SCLC cells, compared with NSCLC cells, suggests that C1 has a degree of selectivity and could serve as a prototype for additional medicinal chemistry optimization. Our evidence for in vivo antitumor activity in a mouse SCLC model further support this conclusion. Further studies are needed to determine if the noted selectivity is due to the loss of Rb1.

While directly targeting the p27–Skp2–Cks1 interface has the advantage of high specificity for the intended target, as discussed above, to date only compounds of modest potency have been identified. We therefore looked for pathways responsible for the formation and/or stability of the Skp2 complex for which there were known potent inhibitors. Blocking the neddylation of the complex on cullin 1 was one such approach that we reasoned could result in a decrease in active Skp2 levels. FKA is a NAE inhibitor that works in RB1-deficient prostate cancer cell lines by binding to the ATP-binding site of the APPBP–UBA3–NEDD8–ATP quaternary complex and inhibiting its formation (37). Because FKA is a relatively nonspecific and low potency inhibitor of NAE and Cul1 neddylation, its clinical utility is uncertain.

We also tested pevonedistat, a highly potent (1,000-fold higher compared with FKA) and specific neddylation inhibitor. In human cancer cell lines, pevonedistat inhibits modification of cullin proteins (including in SCF-Skp2) by NEDD8, resulting in increased levels of CRL (cullin-ring-ligase) ubiquitination substrates, but not substrates ubiquitinated by other non-cullin RING E3 ligases (Fig. 3A). In contrast to proteasome inhibitors with broad activity (e.g., bortezomib), it suppresses only a small subset of intracellular protein turnover. Thus, through modulation of NAE, the NEDD8 pathway regulates abundance of Skp2 and its substrates. Other mechanisms of action are also possible, for example, the neddylation of E2F1 controls its target specificity such that its deneddylation switches it from stimulating proliferation to promoting apoptosis (38). Of relevance to SCLC, p53-deficient cancer cells may be more sensitive to pevonedistat than those with unmodified p53 (39).

The importance of the neddylation pathway in SCLC is supported by studies that found dysregulation of the levels of CAND1 (decreased expression), Cul1 (increased), neddylated Cul1 (increased), and Skp2 (increased) in human SCLC specimens (24–26). Increasing the levels of SCF-Skp2 in SCLC cells could amplify the effect of loss of RB1 on p27 degradation and cell-cycle regulation. Consequently, by reducing the levels of Skp2, a neddylation inhibitor could indirectly cause p27 accumulation, and when used in combination with a direct Skp2 inhibitor, lead an additive or better inhibitory effect on SCLC. However, we found no evidence that the combination had complementary effects on the inhibition of cell proliferation compared with the drugs used alone.

Dependency Map analysis to identify genetic dependencies found the Skp2 is a pan-essential gene (505/625 of CRISPR Common Essential Genes) (depmap.org/portal/gene). While this analysis also found a highly significant (P = 2.7 × 10⁻⁵) enrichment of Skp2 sensitivity in SCLC, it also raises the possibility that its inhibition could have deleterious effects on normal cells. Studies of mice with a targeted deletion of Skp2 (Skp2^−/−) mice suggest that any untoward effects would be manageable, particularly in the context of a relatively short inhibitor treatment. The Skp2^−/− mice showed no evidence of illness or gross anatomic abnormalities up to 10 months of age, and both male and female were fertile (40). The body weights of the Skp2^−/− mice were reduced compared with their littermate controls, but this seems to reflect changes occurring during embryogenesis that would not be relevant in the treatment of adults with a Skp2 inhibitor. Cellular changes noted include centrosome overduplication and accumulation of p27 and cyclin E, and the consequences of these on normal cells would have to be monitored during drug treatment. In our studies with the Skp2 inhibitor C1, we did not find any changes in body weight in treated mice (Fig. 5K).

Our previous work documented the importance of Skp2 in pituitary and prostate cancers driven by the loss of Rb1, and we have now extended these observations to SCLC using molecular, genetic, and translationally relevant pharmacologic approaches to demonstrate that Skp2 can act as a therapeutic target for the treatment of advanced SCLC. We have identified and validated a downstream actionable target of pRb in SCLC, and show that small molecules inhibiting this newly defined signaling pathway slows tumor growth in human and mouse SCLC. While our data show a consistent synthetic lethal interaction between RB1 and Skp2, whether Skp2 inhibition would be also effective in tumors with wild-type RB1 remains an open question. While our data from the H520 cell line are consistent with a selective effect based on RB1, Skp2 knockdown would need to be evaluated in additional cell lines before a definitive answer can be obtained. In addition to SCLC, RB1 is frequently mutated in other solid tumors, including advanced gastroenteropancreatic neuroendocrine carcinomas, osteosarcomas, and metastatic breast cancers, and this is usually associated with poor prognosis and resistance to most chemotherapy drugs (41–43). Interestingly, recent proteogenomic analysis of the paradoxical amplification of RB1 seen in some human colon cancers was found to be associated with pRb hyperphosphorylation and the consequent inactivation of pRb, suggesting that select colorectal cancers could also be candidates for anti-Skp2 therapy (44). Thus, the use of Skp2 inhibitors could have an impact on a broad range of advanced and intractable cancers.

No potential conflicts of interest were disclosed.

Conception and design: H. Zhao, N.J. Iqbal, E.L. Schwartz, L. Zhu

Development of methodology: J. Locker

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Zhao, N.J. Iqbal, V. Sukrithan, C. Nicholas, Y. Xue, J. Locker

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Zhao, N.J. Iqbal, C. Nicholas, J. Locker, J. Zou, E.L. Schwartz

Writing, review, and/or revision of the manuscript: H. Zhao, N.J. Iqbal, C. Nicholas, J. Locker, E.L. Schwartz, L. Zhu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Nicholas

Study supervision: E.L. Schwartz, L. Zhu

Other (assisted with production of lentivirus, for skp2 knockdown): C. Yu

We thank Angela Davies and Champions Oncology for providing the SCLC PDXs and the NCI Division of Cancer Treatment and Diagnosis and Millennium Pharmaceuticals for providing pevonedistat. H. Zhao, N. Iqbal, V. Sukrtithan, C. Nicholas, Y. Xue, C. Yu, E.L. Schwartz and L. Zhu received support from NIH RO1CA230032. This work was supported in part by an NIH Cancer Center Support grant to the Albert Einstein Cancer Center (CA013330).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.