The figure that cannot be replicated (taken from F. Gao et al.)

Last Wednesday (2nd August), Nature Biotechnology announced that they were officially retracting a paper published in May 2016, which claimed to demonstrate how a novel technique could perform gene editing in mammalian cells. The paper, written by a team from Hebei University of Science and Technology in Shijiazhuang, China, reported that the Argonaute nuclease from Natronobacterium gregoryi (NgAgo) had been used to successfully alter the genome of human cells for the first time.

However, since its publication, multiple independent laboratories from China, the USA, South Korea, and Germany have all been unable to reproduce the results of the paper. A patent for the system, which two of the researchers had applied for in 2015, was also withdrawn in January 2016 when they failed to correspond with the State Intellectual Property Office of China.

In response, Nature Biotechnology have said that they felt they had to retract the paper. They labelled the paper with a statement that read, in part, “Despite the efforts of many laboratories (Protein Cell 7, 913–915, 2016; Nat. Biotechnol. 35, 17–18, 2017; Cell Res. 26, 1349–1352, 2016; PLOS One 12, e0177444, 2017), an independent replication of these results has not been reported. We are therefore retracting our initial report at this time to maintain the integrity of the scientific record.”

In light of these revelations, we decided to take a look at the Ago system and why this story has generated so much attention.

The Argonaute Nucleases and NgAgo

When discussing the Ago system, most researchers will compare it directly to the CRISPR-Cas9 system, which has become so prevalent in genomics over the last few years. There are several good reasons for this comparison; the two systems share many similarities. Both have developed naturally within prokaryotic cells to act as a defence against invasive genomes, such as viruses. Both utilise a non-specific nuclease, Ago and Cas9, in conjunction with a guide oligonucleotide sequence (gDNA and gRNA respectively). However, the Argonaute system also possesses some properties that could make it a more desirable tool for gene editing than CRISPR.

The first is that the system doesn’t require a PAM sequence to bind to DNA, a limitation in the CRISPR system that restricts edits to specific sites. The Argonaute nuclease has also been shown to be capable of binding to gDNA without the need for a specific secondary structure, such as the loops necessary for gRNA-Cas9 binding. At 887 amino acids long, Ago is smaller than Cas9 (1,368 amino acids) and so offers more variety when trying to introduce the editing components into cells, such as certain viral vectors that are size limited.

Another difference between CRISPR and Ago is that the Ago system has persisted through the development and evolution of eukaryotic organisms. In fact, the nuclease can be found in one form or another in most living organisms and plays a key role in RNA silencing in healthy cells (when guided by single stranded (ss) RNA). It is only when the nuclease is guided by short 5’ phosphorylated ssDNA (gDNA) that it can cause double strand DNA breaks. As such guide molecules rarely occur naturally in mammalian cells, the Ago system is almost always relegated to RNA interference.

These properties, alongside its ability to perform site specific gene editing with high adaptability and efficiency has meant that the Ago system has been of great interest to genomicists. However, while it has potential as a means of gene editing, there is one important issue with the Ago system that stands in the way of gene editing in vivo: the system requires supraphysiological conditions. This primarily means that the system is only functional at high temperatures, exceeding the temperatures survivable by mammalian cells, and thus the Ago system has been significantly limited its applications. Even in the now retracted paper, the initial intention was to work with nucleases from Thermus thermophilus and Pyrococcus furiosus, but both were shown to only work at temperatures greater than 65 °C.

This limitation was the primary reason that researchers were so excited when the Chinese team, led by Chunyu Han, became the first – and only – group who were able to use the system in mammalian cells by utilising the N. gregoryi nuclease at 37 °C. Starting with in vitro work on Escherichia coli before moving onto human cells, the team reported that they were able to use gDNA with NgAgo to cause site-specific double strand breaks in DNA with high efficiency. Their initial results indicated that the system had a low tolerance for guide-target mismatches, ensuring a low occurrence of off-target activity, as well as a particularly high efficiency at G-C rich target loci. The team also observed that the time between transcribing the Ago nuclease and introducing gDNA impacted the ability for the two to bind; with a gap of 24 hours, guide-nuclease binding was almost entirely absent.

Using protocols developed over the course of their experiment, the team reported that they’d been able to use NgAgo-gDNA and Homologous Directed Repair to insert new DNA into human cells. The main problem that the team encountered was that they were still forced to raise the temperature to 55 °C to load the gDNA. Once cooled to 37 °C, the bound oligos could no longer be exchanged. However, the team hoped that with further investigation, they would be able to find a way of resolving their loading problems and thus, make the tool available at physiological temperatures.

When the paper was published, the reaction within the scientific community was overwhelmingly positive. The scientists involved received wide-spread recognition, there were promises of increased funding for genomics in China, and some people were discussing the possibility of Nobel prizes. If the results could be replicated by other teams, it was research that could completely change our ability to alter the genomes of living cells. The authors ended their publication writing, “Future studies will also likely reveal possible structural modification to enhance the precision and efficiency of NgAgo for genome editing.”

The Controversy

Problems with the paper first began to appear as early as November 2016, when papers from Lee, S.H., et al. and Burgess, S., et al. reported that they couldn’t reproduce the original results. They were later followed by a collection of papers in other journals from separate teams who all reported difficulties in replicating the results Han’s team had published.

The first to be published, Shawn Burgess et al. submitted a paper to Protein & Cell that reported their work testing NgAgo in mice and zebrafish. Initially using similar protocols to those applied to CRISPR-Cas9 experiments, they found that they were unable to achieve gene editing in any of their reactions. They then attempted to use an expression vector that Han had provided them with but they continued to find no evidence of gene editing occurring. The editorial letter accompanying the Burgess paper concludes with, “None of these studies proves that NgAgo has any genome editing activities.”

Several days later, the second paper from Seung Hwan Lee et al. was published in Nature Biotechnology. This paper reported the work of three independent teams who used identical protocols and genomic targets as the original paper, all of whom reported no evidence of gene editing. In a similar fashion to Burgess et al., the editorial letter ends by saying, “On the basis of the above data, we conclude that in conditions designed to replicate those in Gao et al., co-delivery of plasmid DNA encoding NgAgo and a 5′-phosphorylated single-strand gDNA alone is insufficient to induce gene editing at the indel frequencies in cultured human cells reported in the original study.”

Following the publication of these two papers, Nature Biotechnology added a correction to the original report, stating the uncertainty currently being felt. Then, as more papers were publishing highlighting similar difficulties, the editorial team at Nature Biotechnology decided to officially retract the paper.

Last week, following news of the retraction, Han’s team announced that they weren’t willing to give up on their work. They’ve theorised that the NgAgo-gDNA system is only effective when ‘key requirements’ are met, although for now they haven’t specified what those requirements may be.

“Our next step, in response to social concerns, is to do the relevant work in accordance with the school arrangements,” they said, “To select a third party laboratory, accompanied by peer experts to carry out experiments to verify the efficiency of NgAgo-gDNA gene-editing technology and publish the results.”

For now, we’ll have to wait and see if anyone can get the NgAgo system to work as the original paper implied that it could. If it does, it could change way we approach gene editing, and with CRISPR in the news so frequently over the last few months, it has the potential to completely alter the current genomic landscape. If it doesn’t, then we might all have been chasing a myth for over a year.