Application of Photoredox Catalysis for Late-stage Functionalization

You are here:
We recently explored photochemical methods to develop a small molecule library aimed at disrupting the PD-1/PD-L1 interaction. Utilizing MacMillan’s C(sp3)-C(sp3) photoredox cross-coupling reactions, we successfully created a diverse library of compounds through late-stage functionalization, achieving results previously unattainable. Our case study highlights the transformative potential of photochemical approaches in advancing medicinal chemistry projects.

We recently shared our own work towards identifying a small molecule which could disrupt the PD-1/PD-L1 interaction, with a focus on developing a compound with a property profile likely to result in an orally bioavailable molecule.

Through a computation-supported hit-to-lead program, we identified CMP-00003525 (Figure 1), which exhibited a significant increase in potency and good solubility compared to the original hit (primary amine structure redacted as RMe). The boc-protected 3-azetidine pyrrolidine required to make this compound is commercially available (CAS: 1251019-03-0). However, a supplier’s failure to deliver the requested amount before the project’s end prevented us from combining this moiety with a more potent core (RCl) identified later in the campaign.

Photoredox Catalysis_reaction Scheme

Figure 1. General scheme for the synthesis of CMP-00003525 from commercial 3-(1-Boc-3-azetidinyl)pyrrolidine, one of the lead compounds from the PD-L1 medicinal chemistry project program.

During the project, we synthesized multiple substituted pyrrolidine ureas,  exploring late-stage transformations such as amide couplings and reductive aminations along with synthesis of bespoke substituted pyrrolidines. However, the urea coupling reaction is often low yielding (less than 50%) due to issues like competing urea dimerization, inefficient carbamate formation and competing side reactions (especially when triphosgene was used instead of p-nitrophenol chloroformate). Additionally, the process requires twofold excess of pyrrolidine, which is impractical for expensive or precious starting materials. 


Having identified the benefit of introducing the pyrrolidine-azetidine moiety, we conducted molecular dynamics studies and hypothesized that introduction of sp3-rich groups in this region could produce potent molecules with favorable physicochemical properties. Even so, expanding the diversity of sp3-rich substituents on the pyrrolidine posed a significant challenge due to the need for multi-step synthesis of the building blocks. The organic chemist “tool-box” contains very few examples of C(sp3)-C(sp3) cross-coupling reactions compared to their C(sp2)-C(sp2) counterparts. While moving away from sp2-rich systems can enhance solubility and other physicochemical properties, fewer than 5% of published reactions couple two C(sp3)-hybridized centers, whereas 82% involve C(sp2)-C(sp2) cross-couplings (Figure 2).

Photoredox Catalysis_comparison of methodologies

Figure 2. Comparison of published methodologies for coupling C(sp2)-centers vs C(sp3)-centers.

Despite notable advancements in the field of C(sp3)-C(sp3) cross-coupling, their application in industrial medicinal chemistry remains constrained. These reactions frequently require large excesses of one coupling partner, posing challenges when a bespoke synthesis of the building block is needed. Additionally, these reactions often rely on non-abundant starting materials, such as air- and moisture-sensitive alkyl organometallics, necessitating expensive specialist equipment like glove boxes.

What We Did...

In April 2023, MacMillan published a paper entitled “Expedient Access to Underexplored Chemical Space: Deoxygenative C(sp3)–C(sp3) Cross-Coupling.” This study describes a photocatalyzed strategy to directly couple in situ NHC (N-heterocyclic carbene, 9) activated alkyl alcohols with alkyl bromides. The paper presents 54 examples (Figure 3), including three cases of expediated synthesis of APIs and functionalization of sugar and steroid analogs. We were particularly intrigued by Example 39, which featured the coupling of a bromo-azetidine amide 6 with a 3-hydroxy pyrrolidine 7.1

This example provided excellent precedent that the reaction could be applied to the synthesize the privileged pyrrolidine-azetidine building block we’d identified earlier in the project, and also potentially allowing late-stage functionalization of the fully elaborated pyrrolidine urea.

Photoredox Catalysis_scheme for NHC activated photocatalyzed

Figure 3. General scheme for NHC activated photocatalyzed C(sp3)−C(sp3) cross-coupling reaction of alkyl alcohols and alkyl bromides.

The Challenges of Translating Literature to Reality

We initially set out to apply the NHC activated photocatalyzed C(sp3)−C(sp3) cross-coupling to our own in-house PD-L1 leads (Figure 4). We utilized a homemade photoreactor, which had previously been successful in reproducing earlier MacMillan C(sp2)−C(sp3) cross-coupling photoredox chemistry, anticipating it would be suitable for this new project. In our enthusiasm, we attempted to synthesize compound CMP-00003525 via coupling boc-protected bromo-azetidine 13 with the elaborated pyrrolidine alcohol 12. However, this reaction initially failed. Believing we might have been too ambitious using the fully elaborated alcohol, we decided to reproduce an example from the original paper (Example 45), but this reaction also failed to yield any product.

Photoredox Catalysis_Initial trials of photoredox catalysed
Figure 4: Initial trials of photoredox catalyzed C(sp3)−C(sp3) cross-coupling.

Despite the initial disappointment, we pressed on to investigate why the reaction was failing. Our initial concern was whether the NHC-alcohol intermediate was forming correctly. To address this, we reproduced an NMR study from the MacMillan paper to identify the diagnostic chemical shift at around 6.5 ppm, corresponding to the proton between the nitrogen and oxygen of the NHC (Figure 5). We successfully confirmed that the formation of the NHC-alcohol intermediate 17, leading us to conclude that the problem was likely occurring during the photoredox reaction itself.

Photoredox Catalysis_NMR study

Figure 5. NMR study to prove formation of the NHC-alcohol intermediate.

Having established the formation of the NHC-alcohol intermediate, we turned our attention to the reaction profile and the by-products. According to the proposed mechanism, one expected by-product is the oxidized NHC 20, formed after single electron transfer (SET) to the excited photocatalyst, which generates the radical cation 18. This radical cation then loses a proton to give radical 19, followed by β-scission to generate the alkyl radical 21. However, our observations primarily showed the alcohol of the NHC 22. This implies that the NHC-alcohol intermediate is being quenched by water instead of undergoing SET transfer to the photocatalyst (Figure 6).

Photoredox Catalysis_mechanism of action

Figure 6. Proposed mechanism leading to productive turnover of the alcohol to the desired product vs competing hydrolysis of the NHC-alcohol intermediate.

To address this issue, we ensured that the glassware and solvent were stringently dry, but we still saw no improvement in the reaction profile. This led us to suspect that the hydrolysis of the NHC-alcohol intermediate occurs during UPLC sample preparation. Upon consulting with a member of the MacMillan group, they confirmed they had deliberately run the reaction with 5 equivalents of water present and observed only a reduction in yield, not a complete lack of product.

Eventually, we concluded that the iridium catalytic cycle was not initiating, likely due to degradation of the LED strip lights within the homemade photoreactor. To confirm this, we attempted to reproduce a reaction from the earlier C(sp2)−C(sp3) cross-coupling photoredox project, which also failed. This experience led us to conclude that, contrary to claims photoredox can be conducted with domestic lighting, high quality, high intensity lights are essential for reliable reproducibility of these reactions.

Following initial setbacks, the project gained momentum with the acquisition of a new photoreactor setup. After evaluating various options, we opted for a HepatoChem PhotoRedOx Box equipped with PR160L Kessil LED lamps in different wavelengths (Figure 7). This system appeared the most suitable due to its flexibility; unlike some other models that require specific glassware or accommodate only one vial size, the PhotoRedOx offered versatility in potential reaction scales and setups.

With the new photoreactor in hand, we revisited MacMillan’s Example 42 (16), successfully isolating the expected product in a 42% yield. To further validate the method, we replicated three additional examples from MacMillan’s paper (results not shown), achieving yields 34-50% before moving on to the PD-L1 series of interest.

Photoredox Catalysis_equipment

Figure 7. HepatoChem PhotoRedOx Box and PR160L Kessil LED lamp set up in-house.

Rapid Access to Substituted Pyrrolidines

After confirming the chemistry’s validity, we applied it to the synthesis of the earlier pyrrolidine-azetidine building-block (Figure 8). This also enabled us to establish the maximum batch-scale reaction feasible in the HepatoChem PhotoRedOx Box, which in the case of this reaction is 1.2 mmol (relative to the alkyl bromide 24). Initially conducted at trial scale, the reaction was successfully scaled-up to 1.2 mmol with yields remaining consistent. However, the larger scale required an extended reaction time. Following removal of the CBz group by hydrogenation, we isolated 119 mg of the pyrrolidine-azetidine 2 material.

Photoredox Catalysis_Synthesis of the pyrrolidine-azetidine building block

Figure 8. Synthesis of the pyrrolidine-azetidine building block.

The goal of this project was to demonstrate the feasibility of applying the C(sp3)-C(sp3) photoredox cross-coupling reaction to the late-stage functionalization of biologically active compounds. Initially, we derived the alcohol substrate from the PD-L1 core for synthetic convenience, as shown in Figure 9. Our first attempt involved coupling the boc-protected bromo-azetidine 13, to yield CMP0003525, previously synthesize through an alternative route. Yields ranged from 20-24%, which, although modest, are noteworthy given the typically low yields (<50%) associated with urea formation from pyrrolidine. 

Photoredox Catalysis_Synthesis of CMP-00003525

Figure 9. Synthesis of CMP-00003525 via photoredox catalysed C(sp3)−C(sp3) cross-coupling.

We were delighted by the success of this reaction as forming the C(sp3)-C(sp3) bond between the pyrrolidine and an alkyl group as the final step allows rapid access to a diverse range of substituted pyrrolidines from inexpensive, readily available starting materials. It is also worth noting that using the fully elaborated PD-L1 alcohol necessitated increasing the reaction time in the alcohol activation step.

Synthesis of a Compound Library

With the successful resynthesis of CMP-00003525, we designed a small library of compounds to test the scope of the chemistry. We focused on introducing sp3-rich groups onto the pyrrolidine. Although yields were generally low, they provided sufficient material for use in potential biological testing. Some reactions failed, potentially due unstable starting materials. Yields reported are isolated yields after purification by prep-HPLC (Figure 10).

Photoredox Catalysis_compound library
Figure 10. Library of PD-L1 compounds varying the alkyl bromide. Reactions were carried out on 0.2 mmol scale (relative to the alkyl bromide) using general conditions alcohol (1.55 eq.), NHC (1.4 eq.), pyridine (1.4 eq.) in MTBE (2 mL) and alkyl bromide (1.0 eq.), quinuclidine (1.1 eq.), potassium carbonate (1.0 eq.)*, Ir(ppy)2(dtbbpy)PF6 (0.01 eq.), NiBr2.glyme (0.075 eq.) and 2-(4-chloro-2-pyridyl)-4-isopropyl-4,5-dihydrooxazole (0.08 eq.) in DMA (1 mL). 440 nm LED Kessil lamp, 100% intensity, 20-22 hours. * LiBr (1.0 eq) used in place of potassium carbonate. † Reaction was carried out twice, best yield reported.

Optimization of Alkyl Bromide and Alcohol Variants

A portion of the alcohol material from the first round of experiments was converted to the corresponding bromide for a second round of experiments, varying the alkyl alcohol (Figure 11). This approach worked well, allowing access to a broader variety of potential compounds. The only failed reaction in this round was with an alcohol in the benzylic position of an aromatic heterocycle (Example 15). 

The original MacMillan paper contains only one example of a benzylic alcohol, leaving it unclear whether benzylic alcohols are generally not tolerated or were simply not investigated. Further experiments are needed to understand this limitation. For our study, we avoided aromatic groups, as SAR generated in the original PD-L1 work indicated that aromatic groups in this region are not advantageous. Isolated yields after purification by prep-HPLC are reported. Examples 12 and 13 experienced purification issues due to close-running impurities, resulting in diminished yields. 

Photoredox Catalysis_Library of PD-L1 compounds varying the alkyl alcohol
Figure 11. Library of PD-L1 compounds varying the alkyl alcohol. Reactions were carried out on 0.2 mmol scale (relative to the alkyl bromide) using general conditions alcohol (1.55 eq.), NHC (1.4 eq.), pyridine (1.4 eq.) in MTBE (2 mL) and alkyl bromide (1.0 eq.), quinuclidine (1.1 eq.), potassium carbonate (1.0 eq.), Ir(ppy)2(dtbbpy)PF6 (0.01 eq.), NiBr2.glyme (0.075 eq.) and 2-(4-chloro-2-pyridyl)-4-isopropyl-4,5-dihydrooxazole (0.08 eq.) in DMA (1 mL). 440 nm LED kessil lamp, 100% intensity, 5.5-22 hours.


The case study revolves around the quest to develop a small molecule capable of disrupting the PD-1/PD-L1 interaction, aiming for oral bioavailability. Lead optimization efforts yielded a lead-like compound with improved solubility and potency, characterized by a sp3-rich basic pyrrolidine urea.

However, challenges emerged from the complexity and scarcity of C(sp3)-C(sp3) cross-coupling reactions compared to their C(sp2)-C(sp2) counterparts, notably requiring large excesses of one coupling partner and non-abundant starting materials. In response, the study introduces the MacMillan group’s photoredox catalyzed C(sp3)-C(sp3) cross-coupling reaction, offering advantages like the use of readily available native alcohols and broad applicability.

The methodology, feasibility trials, and potential for rapid chemical space exploration are discussed. Despite challenges, including volatility and degassing issues, the study demonstrates promising pathways for medicinal chemistry campaigns targeting the PD-1/PD-L1 interaction.


  1. W. L. Lyon and D. W. C. MacMillan, J. Am. Chem. Soc. 2023, 145, 14, 7736–7742.