The conventional approach to drug discovery shines a light on specific target proteins, using an arsenal of biochemical and virtual screening approaches to create tight binders. This reductionist, target-centric approach has yielded multiple successes, but the inherent problem is that you get what you screen. This can greatly limit the potential for unexpected and new solutions to the problem at hand.

Many allosteric compounds have been discovered through phenotypic screening, an approach that is agnostic to the desired target of interest. Rather than spotlighting a specific target, every protein in a cell is potentially fair game. Even though the target protein is unknown, assays are run within a relevant biological system or cellular signaling pathway to see whether any of the tested chemical substances lead to the desired phenotype. 

Phenotypic screening has been used in the past when molecular target information was limited or unknown. However, driven by new technologies for cell-based screening, phenotypic screening has experienced a renaissance [see Swinney figure below] as a method for the discovery of novel drug targets. It has proven to be especially useful within the area of small-molecule discovery, as shown by the discovery of small-molecule drugs such as Fingolimod, an immune-modulating drug used for the treatment of Multiple Sclerosis, or the chemotherapy-drug Trametinib.

Source: Swinney et Anthony. Nature Reviews Drug Discovery (2011) doi:10.1038/nrd3480

A major advantage of phenotypic screening is that the drug target does not need to be prespecified.  This excludes any bias attributed to the drug target or to its possible mechanism of action. This can enable the discovery of novel treatments for diseases for which the molecular cause is unknown. In contrast to target-driven methods, this more general approach allows complex or multifactorial biological mechanisms to be addressed. Further, in phenotypic screens we can be certain that the protein of interest is presented in the correct conformation and activation state, and the right co-factors are present. This is in stark contrast to the traditional reductionist approach where one clones, expresses and purifies a protein but then has little idea as to whether the protein is really the same as the one in the cell.

On the other hand, there are millions of potential targets in a cell. This means there is no guarantee that the generated molecules are going to be acting on a specific target or mechanism.  Thus, the deconvolution of the biological target and the identification of the mechanism of action can be the most challenging steps following the screening process. Moreover, this leads to the highly debated question of whether the target identification is always a requirement for the development of effective drugs, or whether the identity of the target can be confirmed further down the line, if at all. 

A recent Cell article by Kang et al demonstrates an elegant application of the phenotypic screening paradigm to identify selective inhibitors of mTORC1. The kinase complex mTOR is a central regulator of cellular metabolic processes and serves as the catalytic subunit of the two macromolecular signaling complexes mTORC1 and mTORC2. mTORC1 (mechanistic target of rapamycin complex 1) is the primary sensor and integrator of the cell’s response to nutrient availability, and dysregulation of mTORC1 activation is at the core of many age-related diseases. Whilst the molecule rapamycin demonstrates fascinating pharmacology and is being used in a variety of mTOR-related clinical applications, such as immunosuppression and anticancer drugs, it is not well tolerated over time. This is due to its poor selectivity – not only does it act as the desired inhibitor of mTORC1, but also of mTORC2. To date, the discovery of mTORC1-selective molecules is a challenge.

When investigating the mTORC1-pathway in their recent paper, the scientists from Navitor Pharmaceuticals, a company exclusively focused on the mTORC1-pathway, decided to use a cell-based phenotypic screen. Study author Dr. Eddine Saiah, explained:

“We had multiple approaches to modulate this pathway, many of them included specific discrete protein members of this pathway against which we screened. […] However, we decided to also include a cell-based phenotypic screen. Mainly to recognize that even though we think we know this pathway quite well, you have to admit there are also things you don’t know. And really admitting that was the motivation to include a cell-based screen, because now you are open to new biology, things that you don’t necessarily expect and which can lead you into a new interesting direction.”

Nutrients and growth factors, such as insulin, are key activators of the mTORC1-pathway. Their presence activates mTORC1, which leads to cell growth. Amongst others, this is effected through the regulation of protein translation in form of the phosphorylation of the downstream substances S6 Kinase 1 (S6K1) and 4E-BP1 [see figure 2]

Figure 2: The mTORC1-pathway. Source: Meng et al.Development (2018), doi:10.1242/dev.152595

Nutrient restriction or capping by S6K1 turns off mTORC1-signaling, which then leads to the inhibition of cell growth and stimulation of autophagy.

The initial approach of the Navitor scientists was to identify molecules whose selective inhibition of mTORC1 in cells was defined by a specific signaling signature characterized by an increased phosphorylation of Akt as well as by the reduced phosphorylation of 4E-BP1, an mTORC1 substrate which is not inhibited by rapamycin. As is typical for phenotypic screens, this required the design of a well thought out screening funnel which eliminated any potentially distracting outputs.

In order to do this, the researchers performed a high-throughput phenotypic screen in HeLa cells, which allowed them to narrow 265,000 initial hit compounds down to 70. These 70 top-scoring candidate compounds were then assessed in a high-content immunofluorescence mTOR lysosomal localization assay on amino acid stimulation, which is a specific marker for the nutrient-sensing mechanism of the mTORC1 pathway.

This counter screening for agents that prevent mTOR lysosomal membrane localization upon amino acid stimulation led to the elimination of compounds that did not selectively target the nutrient-sensing mTORC1 pathway, narrowing the number of compounds down to 15. [See figure 3A]

Figure 3A: A high-throughput screening assay cascade to identify selective mTORC1 inhibitors. Source: Kang et al. Cell Chemical Biology (2019), https://doi.org/10.1016/j.chembiol.2019.05.009

Following the initial screening funnel, the team completed hit confirmation and validation and then initiated hit to lead optimization efforts. By switching to pS6K1 as the primary mTORC1 substrate readout, they were able to streamline the new assay and identified the chemical scaffold NV-5440, a compound which showcased a selective mTORC1 inhibition profile [see figure 4].

Figure 4: Structure of the compound NV-5440, a potent and selective mTORC1 Inhibitor. Source: Kang et al. Cell Chemical Biology (2019), https://doi.org/10.1016/j.chembiol.2019.05.009

Utilizing Cellular Target Proteomic Profiling, the authors were able to determine rather unexpectedly that the intracellular target of NV-5440 is the glucose transporter GLUT1. Further characterization revealed that the cellular inhibition of glucose uptake by NV-5440 leads to the selective inhibition of mTORC1 activity, regardless of the presence of growth factors and other nutrients.  The complexity of deconvoluting the target should not be underestimated and the Navitor scientists explored a number of approaches before settling on the successful methodology described in the paper. 

These findings now open up opportunities for new biological research to elucidate the yet undiscovered link between GLUT1 and mTORC1. Potentially, the identified compounds might one day find pharmaceutical applications, for example, in the area of oncology, where the GLUT1 inhibitors might be able to prevent the glucose intake of cancer cells, enabling the shrinking of cancer tumors.

The paper showcases a successful application of phenotypic screening which, within three intense years, led to the identification of a completely novel mTORC1 inhibitory mechanism that is not directly on target.  This is a clear illustration of how phenotypic screening is different from the get-what-you-screen target approach to drug discovery.  An mTOR kinase screen, for example, would never have generated the novel mechanistic insights that Navitor have uncovered.

Despite being experts in mTOR biology, phenotypic screening enabled the Navitor scientists to uncover novel biology. The resulting compounds are drug like, readily optimized and showed clear structure-activity relationships.

But as is often the case in drug discovery, there is no free lunch because phenotypic screening does add a layer of complexity, cost and time. I asked Dr. Eddine Saiah what his conclusions were, looking back: 

“You can look at the paper and the amount of work that went into this program – putting it all together in terms of the HTS screening, the hit to lead optimization, demonstrating in vivo efficacy and then figuring out mechanistically how this compound worked – that is a meaningful and sustained dedication of resources. Phenotypic and cell-based screens certainly offer an excellent way to break new ground for small companies committed to exploring non target-centric approaches. But this also requires patience, dedication to the problem and a supportive board. Not knowing upfront which program will ultimately succeed, the allocation of resources across a number of approaches (including targeted screening and structure-based drug design programs) is key in balancing the portfolio. In Navitor’s case, we are fortunate to have all these ingredients and components in place.”

Navitor’s work illustrates that phenotypic screening requires a commitment of resources and time. It also shows that this screening method is often best combined as part of a portfolio of approaches to a given biology or pathway.  Phenotypic screening offers an orthogonal approach to traditional screening.  It has its strengths and weakness and requires different capabilities.  As noted by Jörg Eder et al. in their review of the origins of first-in-class drugs:

“In our view, the distinction between phenotypic screening and chemocentric drug discovery is not just a semantic one; rather, phenotypic screening as we define it here is a new discipline,”– nice quote from this paper.

Even though there is no guarantee of quick success, the work by the Navitor team shows us that phenotypic screening has the potential to open up new avenues in drug discovery, unlocking novel biology of which we are currently unaware.  As new techniques become available to rapidly deconvolute protein targets, I am certain that we will see phenotypic screening continue to increase in popularity.