IC50 doses for cell growth were calculated through a curve fitting analysis from last day of growth curves using the PRIZM software (GraphPad Software Inc). Unpaired Student's t test was performed to compare cell growth using the Instat software program (Graphpad Software Inc, San Diego, CA, USA). P values were two-sided, and differences were statistically significant at P < 0.05.
We used the RET/TRKA type 2 kinase inhibitor Pz-1 as a binding moiety to generate 22 different PROTAC compounds. The first orally active PROTAC tested in clinical trials was ARV110, containing a CRBN ligand and targeted to the androgen receptor. Therefore, we based our PROTACs on different CRBN ligands (pomalidomide, lenalidomide, or avadomide) which were conjugated through a chemical linker to the solvent-exposure methyl moiety of Pz-1. Only one PROTAC, compound 19, contained a chemical ligand for HL (Fig. 1, Table 1, and Supplementary information). Various chemical linkers have been used to conjugate the the E3 ligase ligands to the target-binding moiety. They can be broadly classified as "flexible", such as PEG, aliphatic, and triazole-based linear linkers, or "rigid" linkers with ring systems. The first ones are more commonly used, although they are more susceptible to oxidation; the seconds suffer from more challenging synthesis methods. Thus, here, we used PEG-, aliphatic- as well as triazole-based linkers (Table 1 and Supplementary information). In detail, pomalidomide-based PROTACs were conjugated to either flexible (PEG, alkyl, or triazole) linkers, while for the lenalidomide- or avadomide-based ones we used PEG linkers (Fig. 1, Table 1, and Supplementary information). The precursor compound (ZW-6-020, compound 1; NMR spectra in Supplementary information, Figs. S1 and S2), used as a negative control, was the key intermediate for the generation of the PROTACs by conjugation to the linker via a N-methylformamide (Table 1, and Supplementary information). In addition, starting from compound 9, we generated another negative control, compound 13, by attaching an ethyl moiety onto the glutarimide of pomalidomide to prevent pomalidomide binding to CRBN. Compound 13 is expected to retain the binding to RET/TRKA but not to be able to induce the degradation of target proteins (Supplementary information, Figs. S3-S6).
We used KM12, expressing the TMP3-TRKA chimeric protein, and TPC-1, expressing the CCDC6-RET chimeric protein, cells to assess Pz-1-based PROTACs ability of inducing degradation of TRKA- and RET-derived oncoproteins. Pz-1 and vehicle (DMSO) were used as negative controls.
Eight of the PROTACs (compounds 8, 9, 10, 12, 15, 20, 21 and 22) strongly reduced both CCDC6-RET and TPM3-TRKA protein level already at 10 nM dose, while compound 11 induced TMP3-TRKA degradation but was almost ineffective on CCDC6-RET (Figs. 2 and 3). The remaining compounds (2, 3, 5, 6, 7, 14, 16, 17, 18, and 19) displayed an intermediate activity being able to reduce at least one of the two protein levels at 50 or 250 nM concentration. These data indicated that triazole linker and VHL ligand are unsuitable for degrading TRKA and RET kinases. In the case of pomalidomide-based PROTACs containing PEG or alkyl linkers, compounds with linkers of 7 to 13 carbon and/or oxygen atoms were the most effective against RET; TRKA could adopt also longer linkers. As predicted, the control compounds 1 and 13, described above and lacking CRBN binding ability, were unable to induce degradation of both TPM3-TRKA and CCDC6-RET up to 250 nM concentration (Figs. 2 and 3).
We treated KM12 and TPC-1 cells with the ten most potent PROTACs (i.e. compounds 8-12, 15, 18, and 20-22). As shown in Supplementary information (Fig. S7), compounds 9, 11, 20 and 21 were the strongest ones active on both TRKA and RET and displaying a similar IC in both KM12 and TPC-1 cells (7.5, 5.5 and 6.3 nM in KM12 cells and 23.3, 14.9 and 15.2 nM in TPC-1 cells, respectively). Among them, compound 9 was selected for further studies.
We investigated whether compound 9 was able to hinder proliferation of human cancer cells displaying RET- or TRKA-derived oncoproteins. To this aim, besides KM12 and TPC-1 cells, we exploited TT cells, derived from a MTC harboring RET C634W oncogenic mutant, and Lc-2/ad cells, derived from a LUAD harboring the CCDC6-RET rearrangement. We also used another MTC cell line, MZ-CRC-1, carrying the RET M918T oncogenic mutant.
We performed growth curves of the 5 cell lines by treating them with different doses of the compound. As shown in Fig. 4, the IC dose of compound 9 for KM12, TT, TPC-1 and Lc-2/ad cells ranged between 2.6 and 10.4 nM. Accordingly, the compound was able to induce a strong reduction of the concentration of RET- or TRKA-derived oncoproteins at 10 nM (Supplementary information, Fig. S8). Instead, MZ-CRC-1 proliferation was only modestly affected by compound 9 (IC 600.5 nM) (Fig. 4). Consistently, RET M918T protein levels were reduced only at very high doses (> 1000 nM) of compound 9 (Supplementary information, Fig. S8). Such lack of effect of compound 9 on RET M918T could be due either to an intrinsic resistance of this mutant to PROTAC-mediated degradation or, in alternative, to a defective E3-ligase/proteosome machinery in MZ-CRC-1 cell line. To dissect between these two possibilities, we exogenously expressed RET M918T in HeLa cells, known to be able to sustain degradation mediated by pomalidomide-based PROTACs. In HeLa cells compound 9 was able to induce RET M918T degradation already at 10 nM, like RET C634R used as a control (Supplementary information, Fig. S8). In addition, RET M918T was also transiently expressed in KM12 cells, that, based on our data are susceptible to TMP3-TRKA degradation upon treatment with compound 9 (Fig. 2 and Supplementary information, Fig. S8). As shown in Supplementary information (Fig. S8), also in this case RET M918T protein was degraded, parallel to endogenous TPM3-TRKA degradation, with a similar potency as RET C634R upon compound 9 treatment. These results indicated that Pz-1 based PROTACs can degrade also RET M918T oncoprotein and that PROTACs activity is cell context dependent. Interestingly, both in HeLa and in KM12 cells a reduction of the degrader activity on RET and TRKA proteins was observed when the PROTAC was used at very high doses (> 1000 nM). This is a typical PROTACs phenomenon, named "hook effect", and consists in their decreased activity at high concentrations, because of binding sites saturation on either the target protein (POI) or the E3 ligase which prevents the formation of the required ternary complex.
To compare the ability to induce protein degradation with the potency in inhibiting cell proliferation, we also selected 4 additional PROTACs, one with a similar activity to compound 9 (compound 20), one with no activity (compound 13) and two with an intermediate activity (compound 17 and compound 3) (Fig. 2 and 3). As shown in supplementary information (Figs. S9 and S10), compound 20 was able to inhibit cell proliferation with an IC dose (around 1-2 nM) similar to compound 9. On the contrary, the other compounds were far less efficient with IC tenfold greater than compound 9.
Finally, to test whether the effect of compounds 9 and 20 was dependent on RET/TRKA inhibition, the two compounds were tested in comparison to Pz-1 in thyroid immortalized and cancer (Nthy-ori 3-1 and B-CPAP, respectively), lung cancer (A549), and colon cancer (HTC-116) cells, all negative for RET or TRKA oncoproteins. As shown in Supplementary information (Fig. S11), these cells were resistant to compound 9 and compound 20 as to Pz-1, featuring a IC dose for cell proliferation > 250 nM. Differently from Pz-1, compound 9 and compound 20 were unable to target VEFR2 protein exogenously expressed in Hela cells (Figure S12).
To characterize the mechanism underlying decrease of protein levels, we measured TPM3-TRKA protein half-life upon treatment of KM12 cells with compound 9 and cycloheximide, a toxin able to block protein synthesis. Cycloheximide alone had no effect on the levels of TPM3-TRKA protein up to 8 h, indicating that TPM3-TRKA half-life is longer than 8 h. In cells co-treated with cycloheximide and compound 9, TPM3-TRKA protein and phosphorylation levels significantly decreased as soon as after 2 h, indicating that compound 9 strongly shortened protein half-life. Specific TRKA autophosphorylation sites (Y674/675 and Y490) were tested (Fig. 5A).
To verify whether the degradation of TPM3-TRKA was mediated by the 26S proteasome, we treated KM12 cells with the proteasome inhibitor MG132. MG132 restored TPM3-TRKA protein levels upon compound 9 treatment, demonstrating that TPM3-TRKA degradation was mediated by the proteasome (Fig. 5B). Interestingly, MG132 was unable to restore TRKA phosphorylation since its inhibition was due to the Pz-1 component of the PROTAC that function as a traditional small molecule kinase inhibitor whose mechanism of action is independent from the proteasome.
Then, to verify whether compound 9 induced TPM3-TRKA poly-ubiquitylation, we transiently expressed TPM3-TRKA in HEK293 cells with or without compound 9 and MG132 (to block protein degradation). TPM-TRKA was immunoprecipitated and the blot stained with anti-ubiquitin antibodies; TPM3-TRKA was poly-ubiquitinated to a greater extent in cells treated with compound 9 and MG132 compared with the cells treated with MG132 alone indicating that the compound promotes TMP3-TRKA poly-ubiquitination (Fig. 5C). Furthermore, we co-treated KM12 cells with compound 9 and chloroquine (50 μM), an autophagy and lysosome inhibitor. Chloroquine had no effect on compound 9 ability to degrade and, consequently, reduce autophosphorylation of TPM3-TRKA, indicating that the autophagic and lysosomal pathways were not involved in compound 9 activity (Fig. 5D).
Next, we explored the role of CRBN in compound 9-mediated TPM3-TRKA polyubiquitination. To this aim, KM12 cells were treated with compound 9 in the presence of an excess of pomalidomide to compete compound 9 binding to CRBN. In cells treated with compound 9, free pomalidomide blocked protein degradation in a dose-dependent manner; as a control, pomalidomide had no impact on TPM3-TRKA the protein level and, consequently, autophosphorylation (Fig. 5E).
CRBN is part of an E3 ligase complex composed by Cullin 4A, RBX1, DDB1 and CRBN itself, in which Cullin 4A is covalently modified by the attachment of NEDD8, a ubiquitin-like chain. Neddylation of Cullin 4A is necessary to recruit the E2 ubiquitin conjugating enzyme. Accordingly, in the not-neddlyated state, the Cullin C-terminal domain forms a groove in which the RING domain of Rbx1 is embedded. Such conformation restrains the movements of Rbx1 and sets the ubiquitin E2 away from its substrates. We transfected HEK293T cells with a myc-tagged Cullin 4A dominant negative construct (Cullin 4A ΔNEDD8), displaying deletion of neddylation site and therefore an impaired ability to recruit E2. Expression of Cullin 4A ΔNEDD8 strongly impaired compound 9 capability to degrade TPM3-TRKA (Fig. 5F). Accordingly, treatment with MLN4924, a neddylation inhibitor, also reduced compound 9 ability of causing TPM3-TRKA downregulation in KM12 cells (Fig. 5G). Autophosphorylation inhibition by compound 9 was not affected by MLN4924 since kinase inhibition was mediated by Pz-1 component of the PROTAC independently from protein polyubiquitination. All together, these results demonstrated that compound 9 degraded TPM3-TRKA by recruiting Cullin 4A/CRBN-containing E3 ligase complex, which, in turn, mediates its polyubiquitination and targeting to the proteasome.
We ought to verify whether Pz-1-based degraders were able to act on TMP3-TRKA in living animals. Due to the potential metabolic liability at the amide bond in the linker of compound 9, we selected compound 20, which does not contain it, for the in vivo testing (NMR spectra are reported in Supplementary information, Figs. S13, S14). We treated mice xenografted with KM12 cells with a single intraperitoneal dose of compound 20 (15 mg/kg) and verified its ability to induce TMP3-TRKA degradation after 12 h. Compound 20 demonstrated a robust efficacy in inducing degradation of TRKA, with a concomitant decrease of protein autophosphorylation (Fig. S15).