Insights into the Degradation of Polymer–Drug Conjugates by an Overexpressed Enzyme in Cancer Cells

Intensive efforts have been made to provide better treatments to cancer patients. Currently, nanoparticle-based drug delivery systems have gained propulsion, as they can overcome the drawbacks of free drugs. However, drug stability inside the nanocapsule must be ensured to prevent burst release. To overcome this, drugs conjugated to amphiphilic copolymers, assembled into nanoparticles, can provide a sustained release if endogenously degraded. Thus, we have designed and assessed the drug release viability of polymer–drug conjugates by the human Carboxylesterase 2, for a targeted drug activation. We performed molecular dynamics simulations applying a quantum mechanics/molecular mechanics potential to study the degradation profiles of 30 designed conjugates, where six were predicted to be hydrolyzed by this enzyme. We further analyzed the enzyme–substrate environment to delve into what structural features may lead to successful hydrolysis. These findings contribute to the development of new medicines ensuring effective cancer treatments with fewer side effects.


Table of Contents
. Active site pocket reference structures of the lowest energy stationary points RC, TI1, and EAM of 1a (A, B, and C, respectively), 1b (D, E, and F, respectively), and 1c (G, H, and I, respectively), where key distances are given in Å, and the free gemcitabine drug here shortened to GEM.

S3
Starting with derivatives 1a-c, the RC structures have the substrate CT atom below 3.3 Å from the OS201, where the corresponding hydrogen is below 2.2 Å from the NɛH448 (Figures S3A, S3D, and S3G). The angles are aligned to move the reaction to TS1, with the worst result for 1a (140.1º, Figure S3A). The free energy barriers for the forward reaction deviate from 20.0 to 27.7 kcal mol -1 (Figures S2). The TI1 is 29.0 kcal mol -1 exergonic in relation to the RC ( Figure S2) and in this latter structure, all three hydrogen bonds of the oxyanion hole are stabilizing the negative charge ( Figures S3B, S3E, and S3H). The reaction evolves to the EAM through the TS2, which has a ΔG ‡ 29.3, 32.2, and 34.9 kcal mol -1 for 1a-c, respectively ( Figure   S2). After drug release (EAM, Figures S3C, S3F, and S3I), the newly generated hydroxyl interacts with the NɛH448, as the drug leaves the active site, and the polymeric chain attached to S201 keeps interacting with the oxyanion hole residues (even though already in the form of carbonyl).
Then, we screened the other derivatives (2a-c, Figure 3). These have the functionalized hydroxyl directly linked to the oxolane ring. These RC structures are in general below 3.0 Å from S201 and, as the substrate prepares to react with the enzyme, the oxyanion hole residues start to interact with the oxygen atom that will develop the negative charge ( Figures 5, S4A, and S4D). Proton transfer from S201 to H448 will occur with the TS1 that is located at 22.0 and 26.3 kcal mol -1 for 2a and 2c, respectively, while 2b is located slightly higher (30.5 kcal mol -1 , Figure S2). Moving to the TS2, the intermediate 2a has a ΔG ‡ of 22.2 kcal mol -1 , while in 2b-c this barrier amounts to 29.3 and 26.1 kcal mol -1 , respectively ( Figure S2). The acylation step ends with the drug release (EAM, Figures 5C, S4C, and S4F), which has lower energy than that of TI1. Alike the 1a-c, the released gemcitabine drug interacts with H448 as it moves out of the reactive center. From these gemcitabine-based conjugates, we have calculated one with a barrier below our threshold (2a) with 24.8 kcal mol -1 for the TS2 in the PES ( Figure S2). The FEL calculations for this conjugate showed a decrease in ΔG ‡ of 2.8 kcal mol -1 and similar energy for the TS1 in both PES and FEL calculations (21.9 versus 22.0 kcal mol -1 , Figure S2). Considering the FEL for 2a and PES for 2b-c, the ΔG ‡ are similar in all three cases ( Figure S2). Figure S3. Active site pocket reference structures of the lowest energy stationary points RC, TI1, and EAM of 2b (A, B, and C, respectively) and 2c (D, E, and F, respectively), where key distances are given in Å, and the free gemcitabine drug shortened to GEM.

SN38-based conjugates
For the allylic derivatives (3a-c) in the RC ( Figures S6A, S6D, and S6G), the carbonyl oxygen atom of the drug's dihydropyran ring is pointing to the amide groups of G122, G123, and A202 oxyanion hole residues. This may be responsible for placing the 3a-c carbonyl from the ester group in a favorable position for nucleophilic attack. The distance from the S201 hydroxyl hydrogen to the H448 rounds 2.2 Å with angles above 158.0º, while the carbonyl carbon atom that is going to be attacked and the S201 hydroxyl oxygen are in general above 3.0 Å apart. After the nucleophilic attack, the TI1 is established and located more than 5.4 kcal mol -1 below the RC ( Figure S5). In the functionalized benzylic derivatives (4a-c, Figures 6D, 6F, and S7A), the substrate CT atom and the serine hydroxyl are more than 2.8 Å apart, whereas the S201-H448 hydrogen bond maintains regular in the three moieties (1.8-2.3 Å, Figures 6D, 6F, and S7A).
The TS2 barrier amount to more than 30.2 kcal mol -1 and the hydrolysis of these conjugates is expected to take too long to occur. However, for PDCs 4a-b, we have calculated energy barriers below our threshold and submitted them to further characterization. Small deviations were observed between PES and FEL for 4a-b. For 4a, these deviations amounted to only 0.1-0.2 kcal mol -1 and for 4b, in particular the TS2, this barrier increased by 3.0 kcal mol -1 . Nonetheless, derivatives 4a-b have rate-limiting steps below 22.3 kcal mol -1 ( Figure S5) and can be classified as candidates for hydrolysis by the hCE2.   The first tested doxorubicin derivatives (5a-c, Figure 3) are under 3.0 Å from the S201 hydroxyl oxygen in their RC complex ( Figures S9A, S9D, and S9G). The substrates' approximation results in hydrogen bonds with the oxyanion hole residues, namely A202 (in all three cases), G123 (in 5b), and G122 (in 5c).
The S201 sidechain proton is below 2.0 Å from the NɛH448 atom with angles for the latter bond above 155.0º. After the achievement of TS1, the system will eventually generate the TI1 structure after releasing around 17.0 kcal mol -1 , locating itself above the RC.
As previously seen for the other drugs, the HɛH448 is much closer to the OS210 rather than Odrg atom at the TI1. Also, the angles describe a better position of the system for the reverse reaction (to RC): 101.1º and 116.5º for 5b-c, respectively ( Figures S9E and S9H). In the 5a TI1, the angle for the forward reaction is higher but near the threshold for a hydrogen bond, as well as the distance between Odrg-HɛH448 (138.2º, 2.9 Å, Figure S9B).
For the hydroxyl directly linked to the allylic ring 6a-c (Figure 3), the RC structures have the substrate CT atom more than 3.0 Å distanced from the OS201 and the corresponding hydrogen atom 1.9/1.8 Å from

S8
NɛH448. The angles are well aligned for the reaction to TS1, ranging from 148.8º to 160.5º (Figures 7A,   7D, and 7G). Figure S8. Active site pocket reference structures of the lowest energy stationary points RC, TI1, and EAM of 5a (A, B, and C, respectively), 5b (D, E, and F, respectively), and 5c (G, H, and I, respectively), where key distances are given in Å, and the free doxorubicin drug shortened to DOX.
The systems evolve to the TI1 (Figures 7B, 7E, and 7H), releasing more than 18.8 kcal mol -1 , being the formation of the TI1 exergonic concerning the RC ( Figure S8). In 6a and 6c, the oxyanion hole is stabilizing the negative charge by three hydrogen bonds and the HɛH448 is closer to the Odrg atom ( Figures   7B and 7H). By contrast, in 6b, the G122 is not interacting with the negatively charged oxygen and the HɛH448 atom is equidistant to OS201 and Odrg (1.9 Å, Figure 7E). The system is better positioned to move the reaction forward, as the angle towards the Odrg atom is higher (149.9º versus 136.9º, Figure 7E). The released doxorubicin (EAM, Figures 7C, 7F, and 7I) is energetically below the TI1 with more than 4.5 kcal mol -1 in the three derivatives ( Figure S8).

S9
Comparing the PES and FEL for these systems, we notice small changes in their barriers. For instance, the TS1 barrier increased around 2.0 kcal mol -1 for 6a-b and 4.1 kcal mol -1 in the case of 6c, which gives us an average barrier of 20.0 kcal mol -1 for the TS1. Concerning the TS2, lower barriers were calculated for 6a-b (2.5 and 1.2 kcal mol -1 , respectively), while in 6c we observed a small increase of 0.8 kcal mol -1 , for an average barrier of 22.0 kcal mol -1 . This is also related to energetic changes in the TI1 structure, which decreased for conjugates 6a-b and increased for 6c. Overall, the rate-limiting step for these conjugates (6a-c) was predicted to be below 23.7 kcal mol -1 ( Figure S8) and capable of hydrolysis by the hCE2. The 7a-c derivatives were then screened (Figure 3). The RC structures have the S201 and the substrate distanced around 2.9 Å for 7a-b, and 3.5 Å for 7c. As the substrate enters the active site, the hCE2 oxyanion hole residues start interacting with the oxygen atom that will develop a negative charge (Figures S10A, In the 8a-c (hydroxyl based on the cyclohexane-oxetane fused rings, Figure 3) RC structure, the substrate's carbonyl carbon atom, and the OS201 are distanced more than 3.0 Å, and the S201 hydroxyl hydrogen is 2.0-2.3 Å from the NεH448. The TS1 is located 24.3 and 23.8 kcal mol -1 above the RC structure for 8a-b, respectively, and a lower ΔG ‡ of 15.9 kcal mol -1 was calculated for 5c ( Figure S11). This will lead to the formation of the TI1 structure, where we observed a pattern concerning the distance between OS201 and HɛH448. In the three 8a-c cases, it is smaller than the Odrg-HɛH448 distance, and the angles favor the backward reaction (Figures S12B, S12E, and S12H). Here, the oxyanion hole residues weakly interact with the negatively charged oxygen: the amide bond of A202 is stabilizing this negative charge in the three cases, and the G123 amide hydrogen is only interacting with this substrate atom in 8a and 8c, rather than interacting with an oxygen atom other than the negatively charged.
We have also explored the functionalization of the hydroxyl function of the cyclooctane (9a-c, Figure 3) and the one located near the benzamide group (10a-c, Figure 3), both of which were unable to retrieve a good active site conformation in the cMD simulations. In the case of PDC 9a, simulations revealed a distortion of the oxyanion hole loop. Concerning PDC 10a, changes in the oxyanion hole loop structure were also observed. Interactions between the alcohol and amide groupsoriented towards the oxyanion hole loop and residues G122 and G123were preventing the interaction of the latter with the substrate's negatively charged atom ( Figure 8A). The amide group of G122 is facing and interacting with residues of the loop, while the G123 interacts with the oxygen atom of the substrate amide group. The only active oxyanion hole hydrogen bond is being performed by the A202 amide group (2.5 Å, Figure 8A). aMD simulations for these conjugates were also unable to capture a conformation where the glycine residues from the oxyanion hole are facing and interacting with the negatively charged oxygen atom. Aside from these issues in the negative charge stabilization, this hydroxyl position ( Figure 8B) exhibits a stereo effect S11 executed by the two methyl groups nearby (placed between the Odrg and HεH448), which may block the proton transfer that needs to occur. Although the OS201 and HεH448 atoms are 2.4 Å apart with an angle of 125.9º, the Odrg and HεH448 atoms are much more distanced (3.3 Å) with a worst angle of 88.9º ( Figure   8A). This may reflect the VdW forces exercised by the two methyls that are pushing the H448 away from the reactive oxygen atoms. Once these groups are near the Odrg, a greater effect can be observed in this oxygen atom, when compared to the more distanced one (OS201). Figure S11. Active site pocket reference structures of the lowest energy stationary points RC, TI1, and EAM of 5a (A, B, and C, respectively), 5b (D, E, and F, respectively), and 5c (G, H, and I, respectively), where key distances are given in Å and the free paclitaxel drug here shortened to PTX. Figure S12. FEL maps of 11a for the deacylation step. The energetic values were calculated with B3LYP-D3/6-31++G(d,p)/MM 25,51 and are given in kcal mol -1 . Table S1. Selected aMD parameters based on: average total potential energy threshold (αdih), inverse strength boost factor for the total potential energy (αpot), average dihedral energy threshold (Edih), and inverse strength boost factor for the dihedral energy(Epot). The parameter values are given in kcal mol -1 .