I and others have been talking to @sheri_fink from the NYT over the last couple of days trying to trace the beginnings of Ebola in Sierra Leone. Sheri and her team has done a tremendous amount of investigations in West Africa over the last couple of months and their phenomenal article was just published:
This puts some of our earlier findings into perspective. E.g. we report in our Gire et al. paper that the outbreak started in Sierra Leone around mid-May (funeral of traditional healer), but the NYT article clearly shows that Ebola had been lingering undetected for quite some time. This is very disconcerting and the article describes in great details how this could have happened. For the purpose of this post, I’m interested in discussing some of the scientific/transmission implications.
One of the main findings we saw in the first dataset was the presence of two viral lineages in Sierra Leone in mid/late May. We speculated how this could be linked to the presence of two lineages in the traditional healer, or the presence of other infected individuals at the funeral. Given NYT’s story, it’s clear to me that these events are independent - the funeral was in April and much too early to have been directly responsible for the late-May cases. These cases must have been a result of a couple of already existing transmission chains, which also explains the presence of viral lineages (clades) 1 and 2.
I don’t think this influences our dating estimates (the individuals we initially sequenced were presenting around the 25th, which we use), but calculating in incubation time, can we use all this information to estimate when clades 1 and 2 split? A back-of-the-envelope calculation gives me something around early-mid April, which seems to fit very well with the article. The presence of lineage 2 in Liberia also fits very well with our findings.
Given new and previous data can we get better estimates of these events so we can understand the transmission events in more detail?
An excellent article. I agree that the scenario of circulating virus in Sierra Leone back into April makes much more sense to me than an individual with multiple lineages transmitting to others (or other lineages circulating at the funeral given the dead individual is the only likely source of transmission). I will think about it some more but if I remember correctly, the upper CI of the date of divergence between lineages 1 and 2 was early April. With the new data the rate is becoming more precise and centering on the low part of the Gire et al range which could push this back further.
@arambaut, sounds great. It’s also important to remember that our dates are all ‘diagnostic testing dates’, so extrapolating from that to infections dates (which is hard and filled with problems, but still) will probably subtract another week or two from our estimates of lineage divergence.
I’ve been getting some new independent new info that’s consistent with the NYTimes investigation. The new sequence data just got a whole lot more important, if thats even possible.
Whats it going to take to show that that the newest Mali and Liberian samples are indeed cluster 2.
Also,can you say that cluster 3 disappeared? I’m especially interested in the 3c that looked as though it was spreading widely in mid june. Did 3 die out?
Far from it - everything we have recently sequence from Sierra Leone is clade 3, so clade 2 seems to have ‘disappeared’ from this country - at least based on our sequence coverage of the country. @rfgarry, do you have a figure with 3c displayed? I can’t remember exactly what that sequence corresponds to - if we have the figure, then I’m sure @dpark could take a quick look at the alignment.
Looking at Figure 4A in Gire 2014 I infer that cluster 1 was the first cluster to be derived from a Guinean isolate.
Cluster 2 came from cluster 1 and cluster 3 came from cluster 2.
Cluster 3 is now apparently the only cluster in Sierra Leone.
Unless the seemingly cluster 2 mutations in the Liberian isolate arose independently by chance or are some sort of assembly artifact that sequence is pretty good evidence that Sierra Leone cluster 2 at least partly reseeded the outbreak in Liberia.
Need new sequences from Guinea , NIgeria, Liberia and Mali
We haven’t yet been able to rule out assembly artifacts in the Liberian assembly, but looking at all the data it seems plausible that clade 2 was indeed introduced into Liberia. Convergent evolution (independent mutation) is highly unlikely in my opinion. I have an earlier discussion on that here (before the NYT article):
Based on all this I think that clade 2 likely seeded both Liberia (as a second wave) and Mali. Hopefully we’ll get data from Nigeria pretty soon - overall though, we need more data from all involved countries.
A BEAST analysis confirms the split between clade 1 and 2 to be mid April (mean 16th April but with a 44 day HPD range). So this does suggest that both lineages do originate at the funeral but the individuals sampled in late May were not directly infected at the funeral (even if they had attended) but rather had been infected by a chain initiated by (2?) other individuals who were infected at the funeral.
Just to follow up on this, a paper just came out in Lancet ID where they mention a secondary introduction into Guinea from Sierra Leone around end of June / early July. They don’t provide any details, but there’s a pretty nice graph. The paper can be accessed here: