The dose makes the poison

Alle Dinge sind Gift, und nichts ist ohne Gift, allein die Dosis macht dass ein Ding kein Gift ist. (The dose makes the poison.)

Aled Edwards, CEO of the Structural Genomics Consortium.

The Chemical Probes Portal was formed to provide expert advice to scientists about the potency and selectivity of molecules reported in the literature, as well as on their proper use, in particular the right concentration to use in a cellular experiment. A new paper by a team from UCSF and Novartis on Covid drug discovery provides an excellent case study of what happens when good probes are used badly.

This is a Covid drug discovery story that started when members of the team used each of the proteins from the SARS-CoV-2 virus as molecular hooks to catch all the human proteins that interact with them in the cell. They had hoped to identify a human protein that was essential for the virus to replicate and, by chance, also was a target of a known drug. Their hypothesis was that the approved drug (targeting the human protein) would also act as an anti-viral. 

The first part of the experiment went well – they discovered many potential drug targets among the human proteins that interacted with SARS-CoV-2 proteins. These included the sigma receptors, which are like molecular chaperones found in the endoplasmic reticulum and for which there are many potent inhibitors, including some chemical probes.

The second part of the story started off well too, with their preliminary observation that a few of the drugs targeting the sigma receptors reduced viral replication in cells. This “chemical validation” supported many cell biology studies that pointed to the sigma receptors as candidate therapeutic targets for Covid. It was the icing on the cake (and the obligate “Figure 4” in the one-word journal paper).

Here is where the story takes a turn. Rather than march into clinical trials, as many would have, the UCSF/Novartis team took the road that is (unfortunately) less travelled – they took a step back and carried out a range of more detailed studies to try to invalidate their hypothesis. And they did.

Among the most critical of their experiments was to explore the relationship between the chemical structure of the drugs and their activity on their molecular target (sigma receptors) and on viral replication – the so-called “structure-activity relationship” or SAR. And they found no correlation. Whatever these drugs were doing to the virus, it had nothing to do with the sigma receptors. It turns out that at the concentration they used in cells in their first studies, the drugs were subtly poisoning the cells, and also viral growth, by a toxic mechanism called phospholipidosis. They then went on to show that many of the drugs reported by others to inhibit SARS-CoV-2 in cells – many of which did advance to clinical studies – also were non-specific, and unlikely to work in humans at the doses recommended for the drug. And so far, their hypothesis has been borne out – in over 300 clinical trials, not a single drug identified in these types of cell-based anti-viral screens has shown efficacy. 

At the Portal, the reviewers take care to recommend the appropriate concentration of the chemical probe. It’s important to pay attention to this. Using compounds at a higher concentration could produce effects unrelated to the target of interest. Compounds that only work at high concentrations should be a red flag – more characterization might be needed to confirm what the probe or drug is doing in a cell.


Drug-induced phospholipidosis confounds drug repurposing for SARS-CoV-2. Tia A. Tummino, Veronica A. Rezelj, Benoit Fischer et al. Science 10.1126/science.abi4708

No shortcuts to SARS-CoV-2 antivirals. Aled Edwards and Ingo V. Hartung. Science 10.1126/science.abj9488


Back to News