Developing Novel PD-L1 Modulators

You are here:

This article describes Charnwood Discovery’s approach to small molecule drug discovery, centered on our PD-L1 program. Herein, we address the complexities of developing potent small molecule inhibitors, whilst concurrently optimizing the properties required for oral bioavailability, detailing a systematic approach aimed at optimizing compound efficiency. Importantly, effective cross-functional collaboration and communication enabled the efficient discovery of differentiated lead compounds, exemplifying the indispensable role of interdisciplinary teams in driving the success of complex drug discovery programs.

Challenges

Programmed cell death 1 protein (PD-1) is a co-inhibitory receptor expressed on the surface of T-cells. PD-1 terminates T-cell mediated anti-tumour responses by binding with PD-L1. Blocking the PD-1/PD-L1 interaction has been proven to reactivate T-cell mediated anti-tumour immunity. Consequently, generating durable clinical responses, and prolonging patient survival rate [1,2], with several monoclonal antibodies (mAbs) approved in oncology indications.[3]

Design make test understand cycle

Small molecule PD-L1 modulators have not yet emulated the success of their mAb counterparts in the clinic , largely because the current state-of-the-art in small molecule PD-L1 inhibitors have sub-optimal properties for the development of oral drugs. Generally, these molecules have high lipophilicity, three or more aromatic rings and very poor aqueous solubility.[4]  Our aim was to identify novel PD-L1 inhibitors with a focus on improving upon the physicochemical and DMPK properties of existing compounds.

What We Did...

As part of an internal drug discovery program Charnwood Discovery identified novel chemical equity as inhibitors of the PD-L1:PD-1 axis. The hit molecules identified were pursued in a hit expansion/hit-to-lead program. 

The initial hits were characterized by “traditional” PD-L1 small molecule inhibitor properties, specifically high aromatic ring count, high lipophilicity, and low aqueous solubility.  

CMP-00002802 (2802) was an archetypal compound of the initial hit series and exhibited weak affinity for PD-L1 by SPR. Coupled with a CHILogD of 3.42, this compound was characterized as an inefficient PD-L1 binder with a low lipophilic ligand efficiency (LLE). The metabolic stability of 2802 was measured in microsomes and was found to be ver low. Reducing clearance through a combination of blocking metabolic hotspots and modulating the global properties (e.g. LogD, etc) of the molecule therefore became important feature of our design strategy for the synthesis of further analogues.

The Permeability of 2802 was found to be moderate in Caco-2 cells. Our strategy of decreasing LogD, as outlined above, therefore carried the risk of reducing the permeability, which could become a limiting parameter with respect to oral bioavailability. The identification of a LogD window in which we could achieve highly efficient binding alongside acceptable solubility and permeability was therefore key consideration.

Developing novel pd-l1 modulators

Figure 1: Hypothesized binding pose of compound CMP-00002802. Monomer 1 shown as orange cartoon, Monomer 2 shown as red cartoon. Ligand represented in bubble form, biphenyl motif and linker shown in yellow, terminal aryl binding group shown in cyan.

The modeled binding mode and pharmacophore of the hit compounds shared many similarities with previous PD-L1 small molecule inhibitors. The deep hydrophobic pocket formed by the homodimer was occupied by a bi-aryl motif, which was linked to an aromatic moiety occupying the region around tyrosine 56.

The overarching aim of the initial project phases was to focus rigorously on improving the physiochemical properties of the series, whilst increasing target affinity, using LLE as a key metric to gauge the quality of our analogues.[5] Alongside lipophilicity modulation, we were keen to investigate the tolerance of the series to the removal of at least one aromatic group, as it’s well known that aromatic rings carry a penalty in terms of optimal drug-like properties in excess to their contribution to high lipophilicity alone.[6,7] Based on our knowledge of the binding modes of different PD-L1 inhibitors and the molecular modelling of our hit compounds, we identified an aromatic group at the opening of the PD- L1 homodimer interface as being amenable to substitution with a non-aromatic moiety (cyan in Figure 1).

Early hit expansion work performed by the medicinal chemistry team demonstrated that the replacement of the aromatic ring at the opening of the PD-L1 homodimer interface was tolerated, affording compounds with comparable LLE values. Pleasingly, this change led to improved aqueous solubility in several analogues, despite the lipophilicity of these compounds remaining similar to that of 2802. Based on these improvements in solubility, further analogues within the two aromatic ring series were pursued resulting in the identification of CMP-00002794 (2794) which inspired further rounds of optimization work (Table 1, 2794 annotated in Figure 2).

CMP-00002802

CMP-00002794

SPR kd (µM)

11.8

5.1

CHILogD

3.42

2.72

LLE (SPR)

1.51

2.57

Aromatic ring count

3

2

PFI (chromLogD)

9

6.8

Kinetic solubility (µM)

13

169

Mics (µL/min/mg) r/h

N/A

181 / <8

Caco-2 (AtoB/BtoA/ER)

N/A

11.4 / 7.9 0.7

Table 1: Profiles of an initial hit compounds 2802 and 2794, identifying 2794 as the more suitable hit to progress into further rounds of optimization due to improvements in several PhysChem and DMPK parameters. 

Following the identification of molecules such as 2794, further rounds of optimization focused on the two aromatic ring series. At this point in the project a TR-FRET biochemical assay came online and correlated well with SPR, with only a handful of outliers to the trendline. All following affinity data is derived from TR-FRET.

To further improve 2794 we pursued a strategy of targeting binding interactions with polar residues to increase binding affinity, whilst maintaining favorable PhysChem properties. The PD-L1 homo-dimer has aspartic acid (ASP-122) and lysine (LYS-124) residues at the exit of the deep hydrophobic tunnel created by its dimerization.[8] These residues were targeted in PD-L1 ligand development in several previous inhibitor series.

Our hypothesis was to target these residues with either hydrogen bond donors, hydrogen bond acceptors or cation forming groups. At this stage in the project, collaboration with computational chemistry was key in allowing us to target residues of the homodimer interface to pick up directional and specific polar interactions. We utilized several techniques but notably molecular dynamics (MD) simulations allowed the most insight into the subtleties of the binding mode of these molecules.

Results and Discussion

The targeting of specific residues with polar interacting groups allowed the team to rapidly identify molecules with increased affinity for PD-L1. This round of screening identified a sub-series of molecules that was clustered at the middle left of the LLE plot, with CHILogD values between 1 and 2, and micromolar binding affinity. Furthermore, the cluster of molecules all exhibited high aqueous solubility as measured by a kinetic solubility assay utilizing DMSO stock solutions (Trajectory A, Figure 2).

Whilst pleased by the improvements in potency and favorable LLE values obtained by the design hypothesis outlined above, we were aware from previous reports that there was the potential to improve potency by modifying the central core of the ligands.

Several analogues of the cluster identified by trajectory A were prepared including modifications to their central core in accordance with this strategy. Pleasingly, this change led to a significant boost in the potency yielding compounds such as CMP-00003824 (3824) (Trajectory B, Figure 2). Our rigorous focus on property-based design allowed the rapid improvement in both potency and DMPK parameters.

Developing novel pd-l1 modulators

Figure 2: LLE plot depicting the optimization journey of molecule CMP-00002794. A rigorous focus on efficient binding and physiochemical property optimization facilitated rapid optimization to deliver a highly efficient lead molecule (CMP-00003824).

Presented in Table 2 is a summary of the profiles of 2794 and 3824. Improvements were made across the board including a drastic reduction in clearance in human microsomes. However, rat microsomal clearance is high and further work is required to understand this species disconnect. Alternatively work in other rodent and non-rodent species could also allow a path for development of this series. 

As expected with the introduction of polar groups, we began to observe slight reductions in permeability and hints of efflux. the synthesis of further analogues allowed us to identify the lower LogD limits of this series, beyond which, loss of permeability becomes a major issue (data not shown). During optimization within any series it is important to develop an understanding of how high the polarity can be pushed before a lack of permeability becomes prohibitive. In the absence of any limiting factor, increasing the polarity (as far as it can be tolerated) is generally considered to be a viable optimization strategy (i.e. driving LLE as a key optimization parameter).

CMP-00002794

CMP-00003824

TR-FRET IC50 (µM)

9.1

0.08

CHILogD

2.72

1.24

LLE (TR-FRET)

2.57

5.84

Aromatic ring count

2

2

PFI (chromLogD)

6.8

4.4

Kinetic solubility (µM)

169

>200

Mics (µL/min/mg) r/h

396 / 263

181 / >8

Caco-2 (AtB/BtoA/ER)

11.4 / 7.9 / 0.7

6.8 / 11.8 / 1.7

Table 2: Profiles of 2794 and 3824, highlighting the improved profile of 3824 in terms of both its potency for PD-L1 and DMPK properties, which are more in line with those required for an orally administered drug candidate.

Extensive computational work was carried out on this series of molecules. MD simulations had a significant impact on the project, enabling the team to rationalizing emerging SAR and guide further compound design. The predicted binding mode of 3824 indicated H-bonding with Asp-122, which was postulated to be a key contributor to the affinity boost relative to 2794. (Figure 3). This binding model was used in further optimization rounds to propose an additional vector to further boost potency for the PD-L1 dimer (data not shown).

Developing novel pd-l1 modulators_4

Figure 3: Proposed binding mode of compound CMP-00003824, obtained from MD simulations. Monomer 1 shown as orange cartoon, Monomer 2 shown as red cartoon. Ligand represented in bubble form, biphenyl motif and linker shown in yellow, terminal non-aryl binding group shown in cyan. Polar contact depicted as magenta dashes.

Summary

  • Utilizing our in-house integrated drug discovery platform, we rapidly optimized hit molecules  into CMP-00003824 (molecule 60 synthesized in the 2 aromatic rings series)
  • Focusing on multiparameter optimization and utilizing metric driven approaches afforded lead molecules with a well-rounded profile.
  • CMP-0003824 gave an LLE of ~6 and showed good aqueous solubility, representing an attractive PD-L1 lead with good differentiation from existing series.
  • Although the microsomal clearance of CMP-0003842 was found to be below the limit of quantification in human, it was very high in rat Clint. Future work will focus on obtaining clearance levels suitable for efficacy studies in animal models.
  • Key to this success was maintaining the aromatic ring count at no more than two rings, as well as focusing on achieving potency boosts via specific-directional interactions with PD-L1.
This project demonstrates Charnwood Discovery’s integrated drug discovery platform and our commitment to efficient hypothesis-driven drug design. A key pillar of our philosophy is to collaborate closely with our clients to define project objectives, functioning as your critical friend to help advance drug discovery projects towards candidate nomination. 

This case study highlights the importance of effective cross-functional collaboration to the successful execution of small molecule drug discovery projects. Computational work on vHTS, with input from the “Medicinal chemist’s eye” was key to the identification of putative hits, which were rapidly profiled using biophysical methods. Following Hit ID, collaboration between DMPK, biology and CADD allowed the medicinal chemistry team to prosecute the design, make, test, understand cycle as efficiently as possible to identify 3824 as a promising lead compound targeting PD-L1.   

References

  1. A Salmaninejad et al, J Cell Physiol, 2019; 234:16824–37. doi.org/10.1002/jcp.28358
  2. KC Ohaegbulam, Trends Mol Med. 2015; 21: 24–33. doi.org/10.1016/j.molmed.2014.10.009
  3. HO Alsaab, Front Pharmacol. 2017; 8:561. doi.org/10.3389/fphar.2017.00561
  4. M.Qin et al, Bioorganic & Medicinal Chemistry, 2021; 33: 116038. doi.org/10.1016/j.bmc.2021.116038
  5. R. Young & P. Leeson, J. Med. Chem. 2018; 61, 15, 6421–6467. doi.org/10.1021/acs.jmedchem.8b00180
  6. T. Ritchie et al, Drug Discovery Today, 2011; 16: 164-71. doi.org/10.1016/j.drudis.2010.11.014
  7. R. Young et al, Drug Discovery Today, 2010; 15: 648-55. doi.org/10.1016/j.drudis.2010.05.016
  8. T. Holak et al, Oncotarget, 2016; 7: 30323-30335. doi.org/10.18632%2Foncotarget.8730
Share this post:
LinkedIn
WhatsApp

From the same category: