CRISPR/Cas9 genome editing has dramatically changed our way to perform biological experiments. While highly efficient and easy to use, one limitation with CRISPR/Cas9 mediated genome editing technology is the occurrence of off-target effects and the restriction of the PAM recognition sequence. Many modifications from the original system have been proposed to improve its efficiency, specificity and to avoid off-target effects. Recently a new system based on the bacteria Natronobacterium gregory Argonaute (NgAgo) was proposed as a serious alternative to CRISPR/Cas9. NgAgo is based on a DNA recognition pattern and unlike all the systems based on CRISPR doesn’t require a PAM recognition sequence. The target specificity is mediated from a phosphorylated oligonucleotide on the 5′ end. As it doesn’t require any cloning or in-vitro transcription, it was sought to be a serious alternative to the CRISPR-Cas9 system.
Recently an astonishing paper published in Nature Biotechnology from Chunyu Han’s group in China proposed NgAgo as a simple system to edit cell lines. As many, I was particularly interested to establish the protocol in my laboratory. The recent availability of the plasmid from Addgene encouraged us to establish this protocol and this is what I tried to do in the last two months or so. Below is a summary of my experience with NgAgo.
Reproducing Han’s paper results:
Firstly I decided to not repeat Han’s experiment stricto sensu as my group works primarily on mouse zygotes. The sequences targeted in this paper were all specific to the human genome. Instead I’ve chosen a gene that I’ve been working on for a very long time (Beta-spectrin) and used it to make my first CRISPR/Cas9 edited mouse line over 2 and 1/2 years ago. Usually to establish a new technique, I use a set of highly efficient sgRNA targeting this gene. These sgRNA are working extremely well and are extremely helpful to improve the technology in my hands.
We had a first attempts on Beta-spectrin gene by co-injecting the NLS-NgAgo-GK plasmid at 5 ng/µl with various concentrations of 5′ phosphorylated oligo (2.5, 25 and 50 ng/µl) purchased from IDT into the mouse zygote. After co-injection of the mix into the pronucleus, we cultured the zygotes for 4 days to blastocyst stage and extracted the DNA for PCR and Sanger sequencing.
Many extra bands on the gel electrophoresis:
The first results from our PCR are below (Figure 1) and were very exiting for us. It showed many extra-bands on the gel. I thought these were products of the edited genome as I see often with CRISPR/Cas9. At that time I was at the TAGC conference in Orlando, USA. I showed the results to my colleagues and after few discussions with them I decided to release this gel picture below (Figure 1) from my twitter account.
We then performed the T7 endonuclease assay on these PCR products (Figure 2) and surprisingly we couldn’t see a clear difference with the original PCR, which was very strange.
We then performed the T7 endonuclease assay on these PCR products (Figure 2) and surprisingly we couldn’t see a clear difference with the original PCR, which was very strange.
Interestingly at higher concentration of 5′ phosphorylated oligo and the same primer set, these extra bands almost disappeared (see Figure 3). We saw this with others genes too (Tet1 and Tet2).
Interestingly at higher concentration of 5′ phosphorylated oligo and the same primer set, these extra bands almost disappeared (see Figure 3). We saw this with others genes too (Tet1 and Tet2).
Meanwhile I discovered from many discussions on my Twitter account, at the TAGC meeting, emails I have received and from this interesting Google group discussion thread that many have tried to replicate Han’s results using his experimental setup, in human cell lines, mouse or zebrafish with NgAgo DNA, mRNA or protein. They all failed to edit the genome.
First Sanger sequencing results:
We then performed a first round of Sanger sequencing and the chromatograms were an absolute mess (Figure 4) to a point that we couldn’t properly identified any sequences (Except from the wild type allele) as many alleles were amplified. However, by matching the guide to the sequences, I had the suspicion that 2 samples were edited (from Figure 1, samples 3 and 8)
Second Sanger sequencing results:
We then performed again the PCRs and decided to cut every single extra band from the electrophoresis gel and send those to Sanger sequencing to determine whether these were sequences from the plasmid, from the edited beta-spectrin gene or primer dimers. Couple of discussions I had on twitter or elsewhere mentioned that the 5′ Phosphorylated oligo could act as a primer and amplify the genome, which is possible and I will come back to this later. The results are in Figure 5 and show convincingly that these extra bands were the amplification of random sequences.
I must make 2 important comments: 1) The primers are specific to the sequence of interest. we have performed tons of PCRs using this primer set and we never saw these extra bands. 2) This result is specific to the low 5′ phosphorylated oligo concentration setup and is almost nonexistent with 25 ng/µl of 5′ phosphorylated oligo.
Clustal alignment:
Initially I thought these sequences were random but I wasn’t quite sure. To test this hypothesis, I aligned all these sequences together using Clustal to see whether I could identify a common pattern. The results are presented in Figure 6 using the results from Sanger sequencing (forward primers). The results are similar for the Reverse primers and I won’t show it here.
There is clearly a common pattern which doesn’t match at all the 5′ phosphorylated oligo. However it matches with the sequences from the Forward and Reverse primers but quite imperfectly and I will come back to this later. The first hypothesis that came into my mind is my primers are not specific enough. Although it didn’t explain 1) Why at 25 ng/µl of 5′ Phosphorylated oligo I don’t see this pattern, 2) I should have for a long time noticed this given I have genotyped and Sanger sequenced over 100 CRISPR/Cas9 edited mice using these primers and 3) the initial PCRs (Figure 1,2 and 3) showed no extra-bands for the B6 (C57BL/6) DNA control or the water.
To investigate this further, I hypothesised that a foreign DNA sequence (plasmid or other nucleotides from the mouse genome) integrated to these amplified sequences. To test this, I Blast searched the sequences to the mouse genome and the primer pairs for each sequence that were cut from the gel. One example is presented in Figure 7. I found the same pattern for the Forward and Reverse primers for all samples that I have tested.
Figure 7 shows two features. Firstly the first 6 to 9 nucleotides from the Forward and the Reverse primers match perfectly with the endogenous sequence. Secondly the remaining 13 to 16 nucleotides from the primer pairs were added to the endogenous sequence. This explains the amplification of these extra bands on the gel (Figure 1). This primer pair was not phosphorylated and no ligase was added to the PCR and sequencing reactions.
NgAgo: A ligase enzyme?
From these results, my hypothesis is as following: The NgAgo plasmid was injected into the zygotes and NgAgo was transcribed and translated into a protein, possibly at zygote stage. The enzyme certainly persisted to blastocyst stage at 37ºC and remained intact after DNA isolation from the blastocysts. The PCR reaction certainly activated the NgAgo enzyme, which functioned as ‘a ligase’ under the classical PCR conditions and added the 10 to 15 nucleotides to the endogenous sequences that were matched with the first 6 to 10 nucleotides of the primer pairs. Interestingly this ‘ligase’ activity from NgAgo seems to be inhibited at high concentration of 5′ Phosphorylated oligo. My hypothesis is this might have degraded the NgAgo enzyme.
My Hypothesis on how NgAgo function:
After these series of experiments, these are my thoughts on NgAgo. Firstly, as many elsewhere found, I have found strictly NO EVIDENCE for a genome editing with NgAgo after multiple attempts with various settings and 3 different genes. Secondly I found instead a ‘Ligase’ like activity of NgAgo under normal PCR conditions, which has strictly nothing to do with the endonuclease activity claimed in Han’s paper. It seems to me that the NgAgo enzyme needs to be heated over 50ºC to function, which is in direct contradiction to the Han’s paper.
My take on all these failed experiments trying to reproduce Han’s paper is basically the incubation temperature of the cells is too low for the enzyme to function or the enzyme/5′ phosphorylated oligo complex is rapidly degraded within the cells explaining possibly why nobody has been able to reproduce Han’s experiment. NgAgo may or may not have an endonuclease activity creating a double strand break but under so specific conditions that they are almost impossible to reproduce and too restrictive for a broad use of this system if this is real. Additionally I do have some serious doubt on NgAgo over its endonuclease activity. Nature Biotechnology should ask Han to release all his raw data + experimental condition to the public. This is a duty of care from the journal. Finally I do believe strongly that whatever happens with NgAgo. the CRISPR/Cas9 system will be there for a very long time and NgAgo will be rapidly abandoned after such failed attempts from everyone in the genome editing field. There is clearly no bright future for NgAgo.
My view on Open Science:
Finally I would like to conclude my post by acknowledging all the people in my laboratory, on Twitter and elsewhere that have contributed to this story. It was my first open science experience and I found the discussion with my peers highly stimulating. I think rather than to chase high impact publications and be secretive, we should be more open and share our results to avoid everyone wasting their time on results that are irreproducible and pointless. In my opinion this is the way Science should work.
Group Leader: Genetics of Host-pathogens interactions and Genome editing
Head of the Transgenesis Facility
ANU College of Medicine, Biology and Environment.
The Australian National University
GPO box 334, Canberra, ACT 2600, Australia.
Why would his product for figure 1 be the result of DSB anyways? PCR wouldn’t amplify a band if it was cleaved, thats not how PCR works. You don’t get two fragments when you try to amplify a fragment that has been cut, you just don’t get a product at all.
Could you please provide Figure 4’s abl file and related control WT abl sequencing file?
We are currently trying the NgAgo approach in C.elegans, will see how it goes but all the recent post about it are not making me fell confident about it…