The Inhibition of CDK8/19 Mediator Kinases Prevents the Development of Resistance to EGFR-Targeting Drugs

Drug resistance is the main obstacle to achieving cures with both conventional and targeted anticancer drugs. The emergence of acquired drug resistance is initially mediated by non-genetic transcriptional changes, which occur at a much higher frequency than mutations and may involve population-scale transcriptomic adaptation. CDK8/19 kinases, through association with transcriptional Mediator complex, regulate transcriptional reprogramming by co-operating with different signal-responsive transcription factors. Here we tested if CDK8/19 inhibition could prevent adaptation to drugs acting on epidermal growth factor receptor (EGFR/ERBB1/HER1). The development of resistance was analyzed following long-term exposure of BT474 and SKBR3 breast cancer cells to EGFR-targeting small molecules (gefitinib, erlotinib) and of SW48 colon cancer cells to an anti-EGFR monoclonal antibody cetuximab. In all cases, treatment of small cell populations (~105 cells) with a single dose of the drug initially led to growth inhibition that was followed by the resumption of proliferation and development of drug resistance in the adapted populations. However, this adaptation was always prevented by the addition of selective CDK8/19 inhibitors, even though such inhibitors alone had only moderate or no effect on cell growth. These results indicate that combining EGFR-targeting drugs with CDK8/19 inhibitors may delay or prevent the development of tumor resistance to therapy.


Introduction
Cancer drug resistance, a key obstacle to achieving cures, was long thought to result from the selection of rare preexisting tumor cells that carry genetic changes conferring resistance. In recent decades, however, it became clear that resistance to both conventional and targeted drugs involves metastable transcriptional changes that allow tumor cells to adapt and survive drug exposure [1][2][3][4][5][6][7][8]. While conclusions from different studies using the same drugs have varied in regard to the relative contributions of pre-existing genetic and stable epigenetic variants towards drug resistance [9][10][11][12][13], a consensus is emerging that transcriptomic plasticity of tumor cells and drug-induced adaptation is key to the initial emergence of low-level resistance. This non-genetic resistance of tumor cell populations provides the background for subsequent selection of stable genetic variants that yield higher levels of resistance [5,13].
Most importantly from the therapeutic perspective, the induction of non-genetic resistance was found to be preventable pharmacologically, which has been first demonstrated by the ability of PKC/NFκB pathway inhibitors to prevent non-genetic induction of the multidrug transporter ABCB1 (MDR1/Pgp) [1,2]. In more recent examples, resistance to different classes of receptor tyrosine kinase inhibitors was reported to be prevented by CDK8/19 inhibitors, while showing moderate or no growth-inhibitory activity when used alone, strongly suppressed the development of resistance to the EGFR-targeting agents. These results suggest that CDK8/19 inhibition may delay or prevent the development of resistance to EGFR-targeting drugs.

Gefitinib and Erlotinib Selection and Sensitivity Testing
To determine concentrations to select for acquired drug resistance, BT474 and SKBR3 breast cancer cells were seeded in 96-well plates (2500 and 1500 cells/well, respectively) and treated with gefitinib (0-5 µM) and (in the case of BT474) with erlotinib (0-10 µM) for 10 days. Visual assessment and image densitometry were used to determine the approximate minimum concentration necessary to produce maximum growth inhibition. Drug resistant cells were selected by plating 150,000 cells in T25 flasks and treating continuously with either erlotinib (BT474: 7.5 µM) or gefitinib (BT474: 1.25 µM; SKBR3: 2.0 µM), passaging as necessary. Cells were concurrently treated with erlotinib or gefitinib in combination with senexin B (1 µM) or 15w (250 nM). Cell cultures were photographed in situ (Zeiss Axiovert 200; Carl Zeiss Microscopy, White Plains, NY, USA) before treatment, at 3 days, and then once weekly for 8-10 weeks after treatment initiation. Densitometric measurements were made using ImageJ software. Statistics were calculated by 2-way ANOVA with post-hoc multiple comparisons in Prism (GraphPad Software, San Diego, CA, USA). Representative flasks were also fixed, stained with crystal violet, and photographed at 2, 4, and 8-10 weeks after initiation of treatment.

Cetuximab Selection and Sensitivity Testing
SW48 cells were seeded in 6-well plates (40,000 cells/well) and treated with cetuximab (0-1 µM) and senexin B (0.313-5 µM) for 7 days. The number of live cells was measured by flow cytometry using a LSR II system (Becton Dickinson, Franklin Lakes, NJ, USA) with a 96-well loader, after staining with 10 µg/mL propidium iodide. For long-term selection, parental SW48 cells were plated in T75 flasks (500,000 cells/flask) and then treated with vehicle (DMSO), or with cetuximab (100 nM) and senexin B (1 µM) alone and in combination. Cells (in duplicates) were collected, live cell numbers were measured by flow cytometry and cells were replated at the original density every 6-10 days for a total of 81 days. Changes in cell number over time were calculated relative to vehicle-treated cells using a multiplying factor (the ratio of the number of cells collected to the number plated at each passage).

Effects of CDK8/19 Inhibition on Gefitinib and Erlotinib Resistance
We used breast cancer cell lines to analyze the effects of CDK8/19 inhibitor senexin B on the development of resistance to small-molecule EGFR inhibitors gefitinib and erlotinib. Single-step selections of limited duration (2-3 months) were carried out starting from 1.5 × 10 5 cells. This starting cell number and duration of exposure are too low to select drug-resistant mutants (see Discussion). Cells were plated in the presence of gefitinib or erlotinib at concentrations that produced near-maximum inhibition of cell growth after 10 days (see Materials and Methods). The results of representative single-step selections (from a total of three independent gefitinib and three erlotinib selections) conducted in BT474 breast cancer cells (ER-and HER2-positive) are shown in Figure 1. Figure 1A shows crystal violet staining of representative flasks at three time points of selection. Figure  1B shows phase contrast microscopic images of cells at multiple time points and Figure  1C presents densitometric measurements of microphotographs (four images per flask). After two weeks of continuous treatment, cells were nearly undetectable by crystal violet staining ( Figure 1A) and very few cells were detectable microscopically ( Figure 1B,C). After an additional two weeks of treatment (four weeks total), cells resumed proliferation and were actively growing at the 8-week time point, indicating that the cells had adapted to EGFR inhibitors.
To confirm the observed effects in another cell line, we have used SKBR3 breast cancer cells (ER-negative, HER2-positive) for gefitinib selection, using the same study design as with BT474 cells. Figure 2 shows the results of a representative gefitinib selection (out of 4 independent selections). Gefitinib resistance took longer to develop in SKBR3 cells than in BT474, but by 10 weeks cells appeared fully adapted to the drug (  We have asked if the prevention of gefitinib and erlotinib resistance by CDK8/19 inhibitors could be due either to synergy between EGFR-targeting drugs and CDK8/19 inhibitors or to the reversal of acquired resistance to gefitinib or erlotinib. Synergy analysis was carried out by the Chou-Talalay method [44], which compares the effects of different concentrations of drugs (gefitinib or erlotinib and senexin B) used individually or at fixedratio combinations. In this method, the drug interactions are characterized by the Combination Index (CI), where a synergistic interaction is defined by CI < 1. To determine if CDK8/19 inhibitor reversed the resistance acquired under our conditions, the same analysis was carried out on the gefitinib-or erlotinib-adapted cell populations, and the degrees of resistance to individual drugs and their combinations were determined by comparing IC50 values between the unselected and drug-adapted populations. We have asked if the prevention of gefitinib and erlotinib resistance by CDK8/19 inhibitors could be due either to synergy between EGFR-targeting drugs and CDK8/19 inhibitors or to the reversal of acquired resistance to gefitinib or erlotinib. Synergy analysis was carried out by the Chou-Talalay method [44], which compares the effects of different concentrations of drugs (gefitinib or erlotinib and senexin B) used individually or at fixed-ratio combinations. In this method, the drug interactions are characterized by the Combination Index (CI), where a synergistic interaction is defined by CI < 1. To determine if CDK8/19 inhibitor reversed the resistance acquired under our conditions, the same analysis was carried out on the gefitinib-or erlotinib-adapted cell populations, and the degrees of resistance to individual drugs and their combinations were determined by comparing IC50 values between the unselected and drug-adapted populations.
The analysis of gefitinib/senexin B interactions in BT474 cells is shown in Figure 3A-C and Table 1. Figure 3A shows the results of a 7-day growth inhibition assay of BT474 cells treated with gefitinib, senexin B, or their 1:1 combination. IC50 values measured in these Cells 2021, 10, 144 7 of 13 assays are shown in Table 1 and CI values (determined at IC50 levels) are indicated in the graphs. Figure 3B,C and Table 1 show the results of the same analysis carried out with cells that were adapted to gefitinib ( Figure 3B) or erlotinib ( Figure 3C). Both gefitinib-and erlotinib-adapted BT474 cells showed increased gefitinib resistance (7.0-fold and 5.9-fold increase in IC50 relative to unselected cells, respectively). The addition of senexin B did not reverse resistance to EGFR inhibitors, as the resistant cells showed the same increase in IC50 values for gefitinib + senexin B combination relative to parental cells (5.9-fold for gefitinib-adapted and 7.0-fold for erlotinib-adapted cells). No significant synergy between gefitinib and senexin B was observed in parental cells or in cells selected with gefitinib or erlotinib (CI = 1.18, 1.13 and 0.91, respectively). These results indicate that the resistancepreventing effect of the CDK8/19 inhibitor was due neither to overcoming resistance that has already emerged nor to a synergistic interaction between the compounds. The analysis of gefitinib/senexin B interactions in BT474 cells is shown in Figure 3A-C and Table 1. Figure 3A shows the results of a 7-day growth inhibition assay of BT474 cells treated with gefitinib, senexin B, or their 1:1 combination. IC50 values measured in these assays are shown in Table 1 and CI values (determined at IC50 levels) are indicated in the graphs. Figure 3B,C and Table 1 show the results of the same analysis carried out with cells that were adapted to gefitinib ( Figure 3B) or erlotinib ( Figure 3C). Both gefitinib-and erlotinib-adapted BT474 cells showed increased gefitinib resistance (7.0-fold and 5.9-fold increase in IC50 relative to unselected cells, respectively). The addition of senexin B did not reverse resistance to EGFR inhibitors, as the resistant cells showed the same increase in IC50 values for gefitinib + senexin B combination relative to parental cells (5.9-fold for gefitinib-adapted and 7.0-fold for erlotinib-adapted cells). No significant synergy between gefitinib and senexin B was observed in parental cells or in cells selected with gefitinib or erlotinib (CI = 1.18, 1.13 and 0.91, respectively). These results indicate that the resistance-preventing effect of the CDK8/19 inhibitor was due neither to overcoming resistance that has already emerged nor to a synergistic interaction between the compounds.    The same analysis was carried out for erlotinib/senexin B interactions in BT474 cells. Figure 3D-F and Table 1 show the response of parental and drug-adapted BT474 cells to erlotinib, senexin B or their 5:1 combination. Both erlotinib-and gefitinib-adapted BT474 cells were resistant to erlotinib alone (3.4-and 2.9-fold relative to parental cells, respectively) and to erlotinib+senexin B combination (2.6-and 1.4-fold, respectively). CI values showed no synergy between erlotinib and senexin B for the parental and gefitinib-adapted cells (CI = 1.05 and 0.93, respectively) but this drug combination appeared synergistic (CI = 0.66) in erlotinib-adapted cells. Together with the finding that erlotinib-adapted cells were less resistant to erlotinib + senexin B combination than to erlotinib alone, this result The same analysis was carried out for erlotinib/senexin B interactions in BT474 cells. Figure 3D-F and Table 1 show the response of parental and drug-adapted BT474 cells to erlotinib, senexin B or their 5:1 combination. Both erlotinib-and gefitinib-adapted BT474 cells were resistant to erlotinib alone (3.4-and 2.9-fold relative to parental cells, respectively) and to erlotinib+senexin B combination (2.6-and 1.4-fold, respectively). CI values showed no synergy between erlotinib and senexin B for the parental and gefitinib-adapted cells (CI = 1.05 and 0.93, respectively) but this drug combination appeared synergistic (CI = 0.66) in erlotinib-adapted cells. Together with the finding that erlotinib-adapted cells were less resistant to erlotinib + senexin B combination than to erlotinib alone, this result suggests that reversal of erlotinib resistance could have contributed to the prevention of erlotinib resistance by CDK8/19 inhibitors.
The same analysis for SKBR3 cells is shown in Figure 4 and Table 1. The effects of gefitinib and senexin B (Figure 4A,B) and erlotinib and senexin B ( Figure 4C,D) are shown for the parental SKBR3 cells (Figure 4A,C) and their gefitinib-adapted population ( Figure 4B,D). While senexin B showed significant growth inhibition in ER-positive BT474 cells (Figure 3), likely due to its activity as an inhibitor of estrogen signaling [33], the growth inhibitory effects of senexin B were very weak in ER-negative SKBR3 cells and never approached 50 percent. Hence, CI values at IC 50 for the combinations of EGFR and CDK8/19 inhibitors could not be determined in SKBR3 cells. Gefitinib-adapted SKBR3 cells showed a moderate increase in their IC 50 values for gefitinib (2.5-fold) and erlotinib (1.4fold), whereas the increase in resistance to the combinations of these drugs with senexin B was 1.6-fold and 1.3-fold, respectively.

Effects of CDK8/19 Inhibition on Cetuximab Resistance
Cetuximab selection was carried out in SW48 colon carcinoma cells (wild-type KRAS), using 100 nM cetuximab. As shown in Figure 5A, this concentration decreases the number of live (propidium iodide (PI)-negative) SW48 cells to 30.9% of untreated cells after 7-day treatment, whereas the dead (PI-positive) cell fraction was below 3%, indicating that the effect of cetuximab is primarily cytostatic. The addition of senexin B at concentrations below 2.5 µM had little effect on 7-day survival of cetuximab ( Figure 5A). For long-term selection, 5 × 10 5 SW48 cells were treated (i) with vehicle (DMSO) control, (ii) 1 µM senexin B, (iii) 100 nM cetuximab, or (iv) 100 nM cetuximab + 1 µM senexin B. Cells were replated and the number of live cells was determined (in duplicates) by flow cytometry every 6-10 days, for a total of 81 days. Changes in cell number (calculated at each time point taking dilutions into account; see Materials and Methods) relative to the control cells are plotted in Figure 5B.
inhibitors could not be determined in SKBR3 cells. Gefitinib-adapted SKBR3 cells showed a moderate increase in their IC50 values for gefitinib (2.5-fold) and erlotinib (1.4-fold), whereas the increase in resistance to the combinations of these drugs with senexin B was 1.6-fold and 1.3-fold, respectively.

Effects of CDK8/19 Inhibition on Cetuximab Resistance
Cetuximab selection was carried out in SW48 colon carcinoma cells (wild-type KRAS), using 100 nM cetuximab. As shown in Figure 5A, this concentration decreases the number of live (propidium iodide (PI)-negative) SW48 cells to 30.9% of untreated cells after 7-day treatment, whereas the dead (PI-positive) cell fraction was below 3%, indicating that the effect of cetuximab is primarily cytostatic. The addition of senexin B at concentrations below 2.5 µM had little effect on 7-day survival of cetuximab ( Figure 5A). For long-term selection, 5 × 10 5 SW48 cells were treated (i) with vehicle (DMSO) control, (ii) 1 µM senexin B, (iii) 100 nM cetuximab, or (iv) 100 nM cetuximab + 1 µM senexin B. Cells were replated and the number of live cells was determined (in duplicates) by flow cytometry every 6-10 days, for a total of 81 days. Changes in cell number (calculated at each time point taking dilutions into account; see Materials and Methods) relative to the control cells are plotted in Figure 5B. Senexin B alone had no significant effect on cell growth over time. 100 nM cetuximab was growth-inhibitory for the first 19 days of treatment but after that point cetuximab no longer inhibited cell growth and the cells continued to grow until the end of the study (81 days), at a rate that was even higher than the control (marked by the slope of the curve), indicating the development of a cetuximab-resistant phenotype. The addition of senexin B initially decreased response to cetuximab, suggesting an apparently antagonistic interaction until the time when cells treated with cetuximab alone developed resistance. In contrast, the combination of cetuximab with the CDK8/19 inhibitor continued to inhibit cell growth until the end of the study (day 81), with no apparent development of resistance. The difference in calculated cell numbers between cetuximab alone and cetuximab + senexin B treatments exceeded four orders of magnitude starting between days 62 and 69. Therefore, despite the apparent antagonism in the short-term, the addition of CDK8/19 inhibitor greatly improves the outcome of cetuximab treatment in the long-term, by preventing the development of cetuximab resistance.

Discussion
In the present study, we have analyzed the effects of CDK8/19 inhibition on the development of adaptive resistance to EGFR-targeting small molecules (gefitinib and erlotinib) or an antibody (cetuximab). We have carried out single-step selection for resistance to EGFR-targeting drugs by plating a relatively small cell number (1.5 × 10 5 cells for BT474 and SKBR3 and 5 × 10 5 cells for SW48 cells) and using relatively short periods of drug exposure (up to 10 weeks). According to recent analysis [47] mutation rates per base per cell division in cancer are on the order of 10 −8 -10 −7 and a minimum of 2 × 10 7 cells (10 T175 flasks) would be required to have even a single mutant; it also takes 6-12 months of selection to achieve full drug resistance. Nevertheless, in our study, resistance in all the cell/drug combinations developed in every flask in less than 10 weeks, suggesting that such resistance represented drug adaptation of cell populations and was not mediated by genetic mechanisms. In all the tested cell-drug combinations, the development of this adaptive resistance was prevented by a highly selective CDK8/19 inhibitor senexin B. The role of CDK8/19 in the development of resistance was verified through the use of another highly selective CDK8/19 inhibitor, 15w, which bears no structural similarity to senexin B. To simplify future analysis of the role of CDK8/19 Mediator kinase in drug resistance, we are currently generating a series of tumor cell lines with CRISPR/CAS9 knockout of CDK8 and CDK19, followed by re-expression of the wild-type Mediator kinases and their kinase-dead mutants.
Although CDK8/19 inhibitors consistently prevented the development of adaptive resistance to EGFR-targeting drugs in all four tested cell line/EGFR inhibitor combinations, this effect in most cases could not be attributed to drug synergy or to reversal of resistance. In particular, gefitinib-adapted BT474 cells showed the same increase in resistance to gefitinib or erlotinib as to combinations of these drugs with senexin B, although erlotinibadapted cells were less resistant to the combination of erlotinib with senexin B than to erlotinib alone. No synergy was apparent in gefitinib or erlotinib combinations with senexin B (in contrast to a strong synergy found between erlotinib and THZ1, an inhibitor of CDK7/CDK12/CDK13 [48]). In the course of cetuximab selection, the addition of senexin B even decreased the response to cetuximab in the short-term but nevertheless prevented the development of cetuximab resistance. Hence, the observed prevention of resistance most likely reflects the inhibition of CDK8/19-mediated transcriptional reprogramming [18,34] that prevents the emergence of adaptive resistance to EGFR-targeting drugs.
Numerous mechanisms of resistance to EGFR inhibitors have been described [42,44,49,50]. These include activating mutations in EGFR itself, mutations of BRAF, KRAS and PIK3CA, as well as amplification or non-genetic upregulation of HER2 and MET, and downregulation or mutation of PTEN. The resistance was also associated with the activation of various compensatory pathways, such as Wnt/β-catenin and NFkB, and compensatory growth factor receptor signaling, including overexpression of HER family ligands, as well as HGF and FGF. Epithelial-mesenchymal transition (EMT) has also been identified as a major mechanism of resistance to EGFR-targeting drugs [49]. As discussed above, the small cell number used in our selections and the observed rapid development of resistance argue against the involvement of any genetic mechanisms. In addition, cross-resistance of gefitinib-and erlotinib-adapted cells argues against the involvement of inhibitor-specific mutations in the EGFR gene. Since drug-resistant cell populations were isolated here via rapid single-step adaptation, they almost certainly represent mixtures of cells with different transcriptomic mechanisms of resistance. With the advent of single-cell (sc) RNA-Seq, it became clear that adaptive drug resistance involves rapid transcriptional changes, which are observed concurrently in the same cell population [5][6][7]50]. In particular, a heterogeneous pattern of resistance-associated transcriptional changes was observed in scRNA-Seq analysis of the effects of short-term (5-day) cetuximab treatment in three head and neck squamous cell carcinoma cell lines [51]. Transcriptomic heterogeneity within the same cell lines was found to increase as an immediate response to cetuximab, and this effect was suggested to indicate that different cell subclones in the same cell line are activating alter-native pathways to overcome EGFR inhibition [51]. In the future, we are planning to use scRNA-Seq to identify the spectrum of CDK8/19-dependent transcriptomic mechanisms that lead to non-genetic resistance to EGFR-targeting drugs.
We have previously reported that CDK8/19 inhibitors prevent the induction of resistance-associated anti-apoptotic factors upon treatment with DNA-damaging drugs or ionizing radiation [25] and suppress the emergence of estrogen-independent ER-positive breast cancer cells upon prolonged estrogen deprivation [33]. These results, combined with the present study, indicate a pleiotropic effect of CDK8/19 on the emergence of drug resistance and suggest potential utility of CDK8/19 inhibitors in preventing the development of resistance to different classes of drugs.