Assessing the functional significance of novel putative oncogenes remains a significant challenge given the limitations of current loss-of-function tools. Here, we describe a method that employs TALEN or CRISPR/Cas9-mediated knock-in of inducible degron tags (Degron-KI) that provides a versatile approach for the functional characterization of novel cancer genes and addresses many of the shortcomings of current tools. The Degron-KI system allows for highly specific, inducible, and allele-targeted inhibition of endogenous protein function, and the ability to titrate protein depletion with this system is able to better mimic pharmacologic inhibition compared with RNAi or genetic knockout approaches. The Degron-KI system was able to faithfully recapitulate the effects of pharmacologic EZH2 and PI3Kα inhibitors in cancer cell lines. The application of this system to the study of a poorly understood putative oncogene, SF3B1, provided the first causal link between SF3B1 hotspot mutations and splicing alterations. Surprisingly, we found that SF3B1-mutant cells are not dependent upon the mutated allele for in vitro growth, but instead depend upon the function of the remaining wild-type alleles. Collectively, these results demonstrate the broad utility of the Degron-KI system for the functional characterization of cancer genes. Cancer Res; 75(10); 1949–58. ©2015 AACR.
Although cancer genome sequencing studies have identified a large number of recurrently mutated genes (1, 2), functional studies are required to determine whether or not these genes present viable targets for cancer therapy. For instance, it is pivotal to establish that potential targets are required for tumor maintenance rather than representing initiating lesions that are no longer required for neoplastic growth. RNAi technology has been widely used to interrogate oncogene dependence by testing the impact of gene/protein knockdown in cell lines harboring the respective gene or pathway alterations. While RNAi has revolutionized our ability to perform genetic loss-of-function studies, it is known to be fraught with significant off-target effects (3–5). RNAi-based off-target effects are particularly problematic for studies using cell proliferation as a phenotypic readout, as any off-target activity toward essential housekeeping genes has the potential to yield a false-positive result. In fact, it is likely that the propensity of RNAi-based methods to yield false-positive results significantly contributes to the recently raised concerns around the lack of reproducibility in cancer biology publications (6, 7).
Many cancer genes, such as oncogenes, exhibit dominant behavior, where only one allele is mutated and thought to drive the tumorigenic effect. In order to distinguish the effects of the mutated from wild-type alleles, a method that allows for the selective inhibition of one but not the other allele would be highly desirable. It is often challenging or impossible, however, to design RNAi reagents that can discriminate between the mutant and wild-type alleles, owing to sequence constraints and the general tolerance of single-base-pair mismatches in many positions (8). Recent advances in somatic knockout technologies, in particular the CRISPR/Cas9 system (9), provide powerful tools for genetic loss-of-function studies. While being more efficient than previous somatic genome engineering technologies, CRISPR-based gene knockouts have significant limitations for the study of essential genes, as gene inactivation is not inducible, generally occurs with slow kinetics (several days), and cannot be reversed in order to mimic intermittent target inhibition. Thus, CRISPR can readily be applied toward the knockout of tumor suppressor genes, but provides less utility toward the study of essential genes such as oncogenes due to the inability to generate stable knockout clones for genes required for cancer cell proliferation.
To address some of the shortcomings of current genetic tools toward the functional characterization of novel oncogenes, we sought to develop a system that ideally would (i) have minimal off-target effects, (ii) enable allele-specific targeting, and (iii) allow for inducible and reversible gene depletion. In this study, we devised a method that allows for allele-specific regulation of endogenous proteins using a chemically regulated destabilization domain (DD) tag (10). We refer to this method as Degron-knock-in (Degron-KI) and demonstrate its application to interrogate cancer dependence of known and novel oncogenes.
Materials and Methods
Cell line, plasmid, and compound
Cell lines and culture media are as follows: HCT116 cells (McCoy's 5A Medium GIBCO#16600 with 10% FBS); Mel-Juso, ESS1, and NCIH1048 cells (RPMI-1640 with 10% FBS); Mel202 (RPMI-1640 with 15% FBS, from Schepens Eye Institute); Karpas422 (RPMI-1640 with 10% FBS and 1× Glutamax from LifeTech). At low seeding density, Mel202 requires 30% conditioned medium to support cell proliferation. All lines were authenticated by SNP fingerprinting, and generally used within 20 passages. TALENs targeting p53 and SF3B1 were purchased from LifeTech. Backbone donor vector for homologous recombination was generated by gene synthesis from LifeTech. Left and right homologous recombination arms were obtained from genomic PCR and cloned into the restriction sites in the backbone donor vector (Nde1 and Xho1 for the left arm; EcoR1 and Not1 for the right arm, Xho1 site was removed in the end). Detailed information on the targeting sequences of TALENs and CRISPR as well as the donor vector design and PCR primers are provided in Supplementary Experimental Procedures. The chemical compounds used in this study are as follows: Doxorubicin (Sigma) and Nutlin-3 (Sigma); BYL719 (Novartis); El1 (Novartis); Shield-1 (Clontech). Shield-1 powder was dissolved in 100% EtOH at 1mmol/L and stored at −20°C. Shield-1 was added to the fresh tissue culture media right before usage.
Primary antibodies: p53 (DO-1, sc-126; Santa Cruz Biotechnology; 1C12, #2524 CST for immunofluorescence), p21 (c-19, sc-397; Santa Cruz Biotechnology), EZH2 (5246s, CST, both for Western blot and immunofluorescence), H3 (9715s; CST), H3K27Me3 (9756s; CST), PIK3CA (4249s; CST), pAKT (4058L; CST), AKT (9272s; CST), GAPDH (IMG-5019A-1, IMGENEX), SF3B1 (p221-3; MBL International, both for Western blot and immunofluorescence). Secondary antibodies: anti-mouse IgG IRDye800 (800-656-7625; Rockland), goat anti-rabbit IgG-HRP (sc-2030; Santa Cruz Biotechnology), goat anti-mouse IgG-HRP (sc-2031; Santa Cruz Biotechnology), goat anti-Mouse IgG Alexa Fluor (A11005; LifeTech), goat anti-Rabbit IgG Alexa Fluor (A11012; LifeTech). Positive clones were further verified by junction PCR and locus PCR (Supplementary Fig. S2).
In vitro cell proliferation (colony formation, three-dimensional growth assay)
Colony formation: Cells were seeded at a very low density in 6-well plates for 10 to 14 days, followed by crystal violet staining (0.5 g crystal violet, 27 mL 37% formaldehyde, 100 mL 10× PBS, 10 mL methanol, 863 dH20 to 1L). Seeded cell numbers: HCT116 (1,500 cells), NCIH1048 (10,000 cells), Mel-Juso (3,000 cells), ESS-1 (3,000 cells), Mel202 (6,000 cells). Conditioned medium (30%) was added in the Mel202 foci formation assay.
Mel202 three-dimensional (3D) growth assay: 300 cells were seeded in 80 μL of culture medium (containing 30% conditional medium) in a 96-well ultra-low attachment round bottom plate (Costar; 7007). Pictures were taken after 9 days. In both cases, fresh Shld compound was added every 5 days.
Reads were aligned to the human genome using a reference transcriptome and a modified version of Tophat, as previously described (11). Expression levels for UCSC genes and transcripts were determined using Cufflinks (12), summed for genes with multiple genomic loci, then rescaled so that the upper quartile gene expression level in each experiment was 10. Sequencing reads and gene expression values were deposited in the NCBI Gene Expression Omnibus database (GEO ID: Series GSE66719).
Details of RNAseq analysis and other Materials and Methods can be found in Supplementary Experimental Procedures.
The DD tag is a derivative of human FKBP12 that is stabilized in the presence of the small-molecule ligand Shield-1 (Shld), but becomes unstable and subjected to proteasomal degradation once Shld compound is withdrawn (10). We reasoned that introduction of the DD tag by homologous recombination at the N-terminus of endogenous genes should impart allele-specific protein stability regulation through addition or withdrawal of Shld (Fig. 1A). To test the feasibility of this approach, we examined if this system can be used to inducibly regulate the function of TP53, a critical tumor suppressor gene. The DD-tag fusion was knocked into the TP53 gene locus of the diploid HCT116 cell line using TALENs, and several clones were isolated by puromycin selection. The DD-tag fusion increases protein size by 12 kDA (Supplementary Fig. S1A), thus allowing for easy distinction of heterozygous clones, which express both DD-tagged and untagged protein, compared with homozygous clones, where only the DD-tagged version but no untagged endogenous protein is detectable. “Homozygous” DD-tagged clones can either represent clones that have the DD tag inserted in all alleles, or alternatively may harbor a DD knock-in of one allele with the remaining alleles being inactivated (leading to knock-out) through error-prone non-homologous end joining, thus resulting in expression of only the DD-tagged protein. While these two possibilities cannot easily be distinguished by Western blot, they become readily evident by targeted sequencing of the genomic insertion site (Supplementary Fig. S2). Notably, out of 24 puro-resistant clones analyzed, 8 (33%) exhibited expression of only the DD-tagged p53 protein, indicating homozygous targeting (Supplementary Fig. S1A). The homozygous DD-p53 clone displayed efficient depletion of TP53 upon Shld withdrawal, whereas p53 protein levels were not altered by Shld removal in the parental HCT116 cells (Fig. 1B and C; Supplementary Figs. S3A and S4A). We next examined if the extent of p53 depletion by the Degron-KI system was sufficient to mirror a TP53 knockout phenotype by investigating its impact on DNA damage response and MDM2 inhibition (13), which require an intact p53 pathway. Similar to parental HCT116 cells, when cultured in the presence of Shld to stabilize the DD-p53 protein, the homozygous DD-p53 clone robustly activated the DNA damage response pathway when treated with the DNA-damaging agent doxorubicin (Fig. 1D and E). In addition, the sensitivity of DD-p53 clone (in the presence of Shld) to the MDM2 inhibitor Nutlin-3 was similar to that of the parental cells (Fig. 1F), indicating that the DD-p53 protein retains wild-type function. Strikingly, Shld withdrawal in DD-p53 cells strongly suppressed the induction of p21 upon doxorubicin treatment (Fig. 1D and E) and rendered these cells insensitive to the MDM2 inhibitor Nutlin-3 (Fig. 1F). By contrast, Shld withdrawal had no effect in the parental cells (Fig. 1D, E, and F). Together, these findings demonstrate that the Degron-KI system is able to recapitulate a TP53 knockout phenotype.
We next sought to investigate whether the Degron-KI system can be used for the functional validation of oncogene dependence. Recurrent hotspot mutations in EZH2 have been described in diffuse large B-cell lymphoma (DLBCL; ref. 14), and EZH2-mutated cell lines are sensitive to pharmacologic EZH2 inhibitors (15–19). We used CRISPR/Cas9 to introduce a DD tag in the EZH2 locus in Karpas422 cells (Supplementary Fig. S2C and S2H), which harbor the EZH2 Y641N mutation and are sensitive to EZH2 inhibitors (16, 18, 19). The EZH2-wild-type cell line HCT116 was used as a negative control. CRISPR-based DD-tag insertion at the endogenous EZH2 locus in HCT116 cells was even more efficient than DD-p53 tagging with TALENs. In fact, when wild-type Cas9 was used, untagged EZH2 protein was almost undetectable in the polyclonal population following puromycin selection, indicating that the majority of cells had homozygous expression of the DD tag (Supplementary Fig. S5A and S5C). By contrast, use of a Cas9 nickase mutant (20–22) resulted in mostly heterozygous DD-EZH2 targeting (Supplementary Fig. S5B), thus providing an approach to effectively generate both heterozygous and homozygous Degron-KI clones in HCT116 cells. Withdrawal of Shld caused destabilization of the DD-tagged EZH2 alleles in both HCT116 and Karpas422 Degron-KI clones (Fig. 2A and D; Supplementary Figs. S3B, 4B, and 6A), and resulted in robust inhibition of H3K27 tri-methylation, the major histone substrate of EZH2 (Fig. 2A and D; ref. 23). Notably, Shld withdrawal markedly impaired the proliferation of DD-EZH2–tagged Karpas422 cells, whereas no effect on cell growth was observed in the HCT116 control cell line (Fig. 2C and E), thus recapitulating the selective dependence observed with pharmacologic EZH2 inhibitors (Supplementary Fig. S7A and S7B; ref. 19). Moreover, the inhibition of both H3K27 tri-methylation and proliferation occurred in a dose-dependent manner (Fig. 2B, F, and G), demonstrating that the Degron-KI system can be used to mimic concentration-dependent inhibition by pharmacologic agents.
To further establish the broad applicability of the Degron-KI system, we next applied it to PIK3CA, a well-established oncogene with several small-molecule inhibitors in clinical trials (24–26). DD tags were knocked into the PIK3CA locus of NCI-H1048, a lung cancer cell line harboring the oncogenic PIK3CA H1047R mutation. In response to Shld withdrawal, homozygous DD-tagged PIK3CA clones (Supplementary Fig. S2D and S2J) demonstrated robust suppression of p-AKT (a marker of PI3K pathway activation) and inhibition of cell growth (Fig. 3A and B; Supplementary Fig. S4C). These phenotypes closely mirrored suppression of the PI3Kα pathway by the selective PI3Kα inhibitor NVP-BYL719 (25). As was the case for the DD-EZH2 clones, both pathway modulation and growth inhibition in DD-tagged PIK3CA clones could be titrated by adjusting the amount of Shld (Fig. 3D and E), thus recapitulating the dose-responsiveness of a pharmacologic PI3Kα inhibitor (Fig. 3D and E). Furthermore, PI3Kα protein levels and downstream signaling (p-AKT inhibition) were markedly suppressed at 8 hours after Shld withdrawal, with some additional reduction by 24 hours (Fig. 3C). Collectively, the ability of the Degron-KI system to mirror the activity of pharmacologic inhibitors such as EZH2 and PI3Kα inhibitors demonstrates the utility of the Degron-KI system for the functional validation of oncogene dependence.
We next set out to apply the Degron-KI system toward the study of a novel and less well-understood oncogene. The spliceosome component SF3B1 was recently found to be recurrently mutated in myelodysplastic syndromes, chronic lymphocytic leukemia, and uveal melanoma (27–31). SF3B1 mutations are generally heterozygous and cluster at codons 625, 666, and 700, a mutation pattern that is characteristic for oncogenes and suggests possible gain of function. Yet it remains unclear what functional role SF3B1 mutations may play during oncogenesis and whether deregulated SF3B1 activity is required for cancer maintenance. SF3B1 small-molecule inhibitors are currently under development and have entered clinical trials (32). The fact that SF3B1 is part of the spliceosome, however, raises the concern that pan-SF3B1 inhibitors may exhibit significant toxicities in normal tissues and thus present a very narrow therapeutic window. In order to address whether or not SF3B1 is broadly required for cellular growth, even in the absence of oncogenic alterations, we generated both heterozygous and homozygous DD-SF3B1 clones in Mel-Juso, a melanoma cell line expressing wild-type SF3B1 (Supplementary Fig. S2E). Depletion of one copy of SF3B1 in heterozygous DD-SF3B1 clones had no impact on growth (Fig. 4A and B, clone #3). By contrast, Shld withdrawal in homozygous DD-SF3B1 clones led to robust SF3B1 protein depletion and markedly impaired growth (Fig. 4A and B, clones #1 and #2; Fig. 4E; Supplementary Figs. S4D and 6B). As before, this effect was Shld-dosage dependent (Fig. 4C and D). Together, these findings indicate that while SF3B1 is not haploinsufficient, complete inhibition of SF3B1 is not tolerated in SF3B1 wild-type cells, therefore substantiating the toxicity concerns for a pan-SF3B1 inhibitor.
A potential avenue to circumvent the normal tissue toxicity of pan-SF3B1 inhibition would be the development of SF3B1 inhibitors that selectively target the mutant protein. However, given the challenges and resources required toward the development of mutant-selective inhibitors, it will be critical to confirm that SF3B1 mutations are in fact required for tumor maintenance. We therefore sought to employ the Degron-KI system to assess the impact of mutant-specific SF3B1 inhibition in two SF3B1-mutant cancer cell line models, the endometrial cell line ESS-1, and the uveal melanoma cell line Mel202. When analyzing DD-tagged ESS-1 clones, we noticed that this cell line is polyploid, expressing at least two wild-type and one mutant allele (K666N hotspot mutation). Allele-specific targeting was confirmed through allelic phasing of the DD tag with the K666N SF3B1 mutation by RT-PCR (Supplementary Fig. S8A–S8D), yielding several clones that selectively deplete either all copies of SF3B1, two copies, or only one copy of SF3B1, including clones that selective target only the mutant or wild-type alleles (Supplementary Table S1). Similar to the findings in the SF3B1 wild-type cell line, Shld withdrawal in clones that deplete all copies of SF3B1 strongly inhibited the growth of ESS-1 cells (Fig. 5A, clones #1 and #2), indicating that SF3B1 is also required for the growth of ESS-1 cells. Surprisingly, however, the selective depletion of mutant SF3B1 protein, which was confirmed by Western blotting (Fig. 5B; Supplementary Fig. S4E), did not impact the proliferation of ESS-1 cells (Fig. 5A, clones #3 and #4). By contrast, the selective depletion of the wild-type protein profoundly inhibited cell growth, despite continued expression of mutant SF3B1 protein (Fig. 5A, clones #7, #8, and #9). To exclude the possibility that the dependence on wild-type SF3B1 in this cell line is due to a gene dosage effect, as cells harbor only one mutant and likely two wild-type copies of SF3B1, we investigated the impact of Shld withdrawal on clones that had a DD tag inserted in both the mutant and one wild-type copy of SF3B1, where cells continued to express presumably only one remaining copy of the SF3B1 wild-type protein. Shld withdrawal in these clones also had no discernable growth impact (Fig. 5A, clones #5 and #6), despite having a similar total level of remaining SF3B1 protein compared with clones #7, #8, and #9 (Fig. 5B; Supplementary Fig. S4E). Collectively, these findings suggest that the in vitro growth of ESS-1 cells critically depends on the remaining wild-type copies of SF3B1 but not mutant SF3B1.
Given these unexpected findings in ESS-1 cells, we wanted to investigate the effects of allele-specific SF3B1 inhibition in an additional cancer cell line model. The uveal melanoma line Mel202 harbors the SF3B1 R625G hotspot mutation, which is recurrently mutated in primary uveal melanoma tumors (27, 28). This mutation has previously been described to correlate with alternative splicing of the genes UQCC and CRNDE in uveal melanoma tumors (33), but a causal relationship between SF3B1 mutations and these alternative splicing patterns has not yet been established. Mel202 cells were found to be expressing one mutant and multiple wild-type alleles of SF3B1. Using the Degron-KI system, we again were able to obtain Mel202 clones that had DD-tag insertions in single or multiple alleles of SF3B1. During the clone characterization, we identified clones that exhibited a constitutive knockout (KO) of the mutant allele due to a small in-del mutation (Supplementary Table S2, clone #3). This mutant knockout clone showed a reversal of the altered splicing pattern in UQCC and CRNDE to that of the SF3B1 wild-type cells (Fig. 5C; Supplementary Fig. S9), thus providing a robust readout for the functional inactivation of mutant SF3B1. Indeed, in clones that selectively DD-tag the R625G mutant but not wild-type SF3B1 alleles, Shld withdrawal reversed the altered splicing pattern of UQCC and CRNDE (Fig. 5C), indicating that the Degron-KI system allowed for robust inactivation of mutant SF3B1. By contrast, in clones that selectively DD-tag the wild-type SF3B1 alleles but not the R625G allele, Shld withdrawal did not affect the splicing pattern of UQCC and CRNDE (Fig. 5C). These findings demonstrate that SF3B1 mutations are causally linked to the alternative splicing of UQCC and CRNDE. Next, we wanted to examine the effects of selective SF3B1-mutant depletion on gene expression and alternative splicing at a global, genome-wide level. Analysis of the RNAseq data using CuffDiff (12) showed that selective depletion of the DD-R625G–mutant protein in MEL202 cells had little impact on the global gene expression pattern (Supplementary Fig. S10). By contrast, using MATS (34), we found that selective depletion of the SF3B1 mutant affected the alternative splicing of 138 genes, including several genes that were found to be alternatively spliced in primary uveal melanoma tumors harboring SF3B1 mutations, e.g., ABCC5, SERBP1, SNRPN, DLST, DYNLL1, ENOSF1, SLC3A2, STIM1, TMEM14C (ref. 33; Supplementary Tables S3–S4; Supplementary Fig. S11). Moreover, we found that the majority of alternative splicing events in response to mutant SF3B1 depletion consisted of alternative 3′ splice sites, with little impact on 5′ splice sites (Supplementary Fig. S12). These findings are consistent with SF3B1′s role as a core component of the U2 snRNP complex that promotes the initial formation of the spliceosome at 3′ exon/intron boundaries (35). Together, these findings indicate that SF3B1 mutations alter spliceosome activity that lead to “aberrant” use of alternate 3′ splice sites.
Despite the efficient reversion of the altered splicing pattern, however, mutant-selective depletion of SF3B1 had no significant impact on in vitro growth in both two-dimensional and 3D tissue culture systems (Fig. 5D and F, clones #1 and #2). By contrast, as was previously observed in ESS-1 cells, the wild-type selective depletion of SF3B1 significantly impaired the growth of MEL202 cells (Fig. 5D and F, clone #4). The results from these two SF3B1-mutant cancer cell line models suggest that the SF3B1-mutant allele is not required for tumor maintenance, at least under the in vitro conditions tested. Moreover, the critical dependence on the remaining wild-type copies of SF3B1 indicates that SF3B1 hotspot mutants must be defective in some essential cellular activity that requires compensation by the remaining wild-type protein.
The Degron-KI system described here presents a method for the functional characterization and validation of novel cancer genes, and provides several features that address some shortcomings of current genetic loss-of-function tools. First, in contrast to RNAi-based methods, the Degron-KI approach should have minimal off-target effects, as the DD tag should only create fusion genes for the gene of interest. Indeed, a genome-wide survey by RNAseq of several DD-SF3B1–tagged Mel202 clones revealed that DD-tag fusions were only detected at the SF3B1 gene locus (Supplementary Fig. S13). Using RNAseq, we also confirmed that the addition of Shld has minimal to no effects on global gene expression (Supplementary Fig. S10). Moreover, the ability to study each clone in the presence and absence of Shld provides an ideal isogenic setting and a robust control for any potential CRISPR-induced genomic alterations. A potential limitation of the Degron-KI approach is that it only works efficiently when applied as an N-terminal fusion (10, and data not shown) and thus requires that proteins can tolerate N-terminal tagging. While C-terminal degron fusion systems have been described (36, 37), we found that protein depletion was much less effective compared with the DD-Shld system (data not shown). In addition, for some genes the expression level of DD-tagged protein was reduced compared with the untagged protein, possibly due to slight alteration in protein stability (incomplete stabilization through Shld) or translation efficiency (due to the P2A ribosome skipping site). Importantly, however, for 5 of 6 genes that we have tested so far, we found that homozygous DD-tagged clones behaved similarly to the parental cell line, suggesting that in most cases the DD-tagged protein is sufficiently expressed and functional. Further supporting the high reproducibility of the Degron-KI method, our analysis of multiple independent DD-tagged clones revealed similar phenotypic results for each gene studied.
A second feature of the Degron-KI approach is that it enables inducible, titratable, and reversible gene depletion. A key requirement for any loss-of-function method is the ability to efficiently inhibit protein function. Although the extent of protein depletion with the Degron-KI system was robust for the proteins tested in our study, it appears to be incomplete in most cases. Therefore, the Degron-KI approach described here will most likely create hypomorphs rather than complete null states. Consistent with this notion, Shld withdrawal in the DD-tagged clone led to robust but slightly weaker inhibition of p-AKT and cell proliferation in NCIH1048 DD-PIK3CA cells compared with the PI3Kα small-molecule inhibitor BYL719. However, it is important to note that in all cases tested, the extent of protein depletion was sufficient to recapitulate genetic (p53) or pharmacologic (EZH2, PIK3CA) loss-of-function phenotypes, as judged by downstream pathway inhibition and impact on growth (in the case of oncogene dependence). Another important feature of the Degron-KI system is that, in contrast with current genetic tools, it allows for the titration of gene inactivation by varying the dose of Shld. The titration of Shld in DD-PIK3CA lines, for instance, was able to mirror the dose-dependent inhibition of pathway and growth modulation by a pharmacologic PI3Kα inhibitor. Collectively, the reversibility and ability to titrate protein depletion with the Degron-KI system provide a closer surrogate of target modulation by pharmacologic inhibitors.
We would like to note that the Degron-KI method is highly efficient and broadly applicable. We were able to recover high proportions of correctly DD-tagged clones for 4 different genes tested across 7 cell line models, including a nonadherent DLBCL cell line that is difficult to transfect. Many cancer cell lines exhibit aneuploidy, which creates a challenge for somatic engineering as many loci will be present at more than two copies. Using CRISPR-based targeting, however, we were able to engineer “homozygously” DD-tagged clones in a single round of transfection even in cell lines that were triploid or tetraploid for the locus of interest. Together with a recent study that developed a similar approach but was restricted to a single gene and a single cell line (38), these findings indicate that the Degron-KI approach should be broadly applicable to many genes and cell line models.
Lastly, a key feature of the Degron-KI system is the ability for allele-specific targeting. The application of this technology to the study of the putative oncogene SF3B1 yielded several important insights. SF3B1 mutations cluster at hotspots and are suggestive of gain of function, but their mechanism of action and oncogene dependence remained unclear. Although previous studies demonstrated a correlation between SF3B1 mutation and aberrant splicing, allele-specific targeting enabled us to provide the first demonstration that this altered splicing pattern can be directly reversed by inhibition of the mutant but not the wild-type protein. Our data indicate that SF3B1 mutations alter spliceosome activity, leading to aberrant use of alternate 3′ splice sites. Moreover, allele-specific targeting allowed us to separate the impact on alternative splicing of the mutant protein from the cell-essential spliceosome function performed by the wild-type SF3B1 protein. Surprisingly, the reversal of aberrant splicing in response to SF3B1-mutant–specific inhibition did not translate to growth-inhibitory effects in the two SF3B1-mutant cancer models tested. Perhaps even more unexpected, SF3B1-mutated cells are critically dependent on the function of the remaining wild-type SF3B1 allele(s), and sole expression of mutant SF3B1 is not sufficient to maintain in vitro growth. Together, these findings indicate that the splicing alterations induced by SF3B1 mutations are not required for in vitro cancer proliferation. The dependence on the wild-type SF3B1 allele suggests that the SF3B1 mutations have characteristics of a loss-of-function allele with respect to in vitro growth, potentially explaining why SF3B1 mutations are always heterozygous, as homozygous alterations may not be compatible with cell growth. It is possible that SF3B1-mutant function is only required for in vivo growth or other aspects of tumorigenesis, or that growth dependence was lost with repeated passaging in tissue culture. Alternatively, SF3B1 mutations may present a predisposing or initiating event that is not required for continued cancer growth. Despite the potential caveats outlined above, the findings described raise caution as to whether the selective targeting of SF3B1-mutant protein will yield robust antitumor effects. Moreover, the finding that SF3B1 appears to also be essential in cells lacking oncogenic SF3B1 alterations raises potential concerns regarding the therapeutic index for pan-SF3B1 inhibitors.
As illustrated by the case of SF3B1, the Degron-KI approach described here presents a powerful tool to assess dominant and recessive cancer gene function, and significantly advances our approaches for credentialing novel cancer targets.
Disclosure of Potential Conflicts of Interest
H.M. Chan is research investigator at Novartis. No potential conflicts of interest were disclosed by the other authors.
Conception and design: Q. Zhou, H.M. Chan, F. Stegmeier
Development of methodology: Q. Zhou, H.M. Chan, J. Min
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q. Zhou, D. Ruddy, I. Kao, M. Lira, V. Gibaja
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Zhou, A. Derti, J. Min
Writing, review, and/or revision of the manuscript: Q. Zhou, A. Derti, H.M. Chan, Y. Yang, J. Min, M.R. Schlabach, F. Stegmeier
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Q. Zhou, D. Rakiec, Y. Yang
Study supervision: H.M. Chan, J. Min, F. Stegmeier
The authors thank Xiaoqin Xiang, Rosalie deBeaumont, Lesley Griner, Feng Fei, Geoffrey Bushold, Josh Korn, and Li Li for the technical advice. Raymond Pagliarini kindly provided the construct containing the DD sequence. Christine Fritsch helped to advise the culture condition of the PIK3CA-mutant cell line NCIH1048.
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.