Australia: The Land Where Time Began

A biography of the Australian continent 

Neanderthals and Modern Humans – Adaptive Introgression Driven by RNA Viruses

At least twice interbreeding took place between Neanderthals and modern humans over the last 100,000 years. Though there is evidence that most segments of introgressed DNA from Neanderthals to modern humans were removed by purifying selection, less is known about the adaptive nature of the introgressed sequences that were retained. It was hypothesised by Enard & Petrov that interbreeding between Neanderthals and modern humans resulted in:

1.     The exposure of each species to novel viruses, and

2.     The exchange of alleles that were adaptive which provided resistance to these viruses.

In this study it was found by Enard & Petrov that there were segments of Neanderthal ancestry, that were long, frequent – and more likely adaptive – in modern humans are enriched in proteins that interact with viruses (VIPs). They found that VIPs that interact with RNA viruses specifically were more likely to belong to segments that had been introgressed in modern Europeans. Their results showed that segments of Neanderthal ancestry can be used to detect ancient epidemics.

Modern humans and Neanderthals interbred a least twice following their divergence 500,000 BP to 800,000 BP: the first time ~100,000 BP (Kuhlwilm et al., 2016) and the 2^nd ~50,000 BP (Fu et al., 2015; Green et al., 2010; Pääbo, 2015; Sankararaman et al., 2012, 2014).

The first episode of interbreeding introgressed (IS) of modern ancestry within Neanderthal genomes (Kuhlwilm et al., 2016), as was revealed by analysis of ancient DNA obtained from a single Altai Neanderthal individual that had been sequenced (Prüfer et al., 2014). According to Enard & Petrov it appears not to have left any segments of Neanderthal ancestry that were detectable in extant modern human genomes (Kuhlwilm et al., 2016). Contrasting with this, detectable introgressed segments of Neanderthal ancestry within the genomes of non-African modern humans (Fu et al., 2015; Green et al., 2010; Prüfer et al., 2014; Sankararaman et al., 2014; Vernot & Akey, 2014).

The discovery that the majority of genomic segments that were initially introgressed from Neanderthals to modern humans were removed rapidly by purifying selection, has resulted from recent in the detection of introgression. It has been estimated (Harris & Nielsen, 2016) that the proportion of Neanderthal ancestry in modern human genomes fell rapidly from ~10% to the levels of the present of 2%-3% in modern Asians and Europeans (Fu et al., 2016); Juric et al., 2016).

Several important questions are raised by this history of interbreeding and purifying selection against introgressed segments.

1.     Frist, Can it be detected among the sequences that were introgressed that were ultimately retained which sequences persisted by chance because they were less deleterious or not at all deleterious to the recipient species and which persisted because of natural selection, and not in spite of it – that is, introgressed segments increased in frequency as a result of positive selection?

2.     Can it be determined which pressures in the environment drove this adaptation if any of the introgressed sequences were indeed driven into the recipient species due to positive selection?

Enard & Petrov found recently that proteins that interact with viruses (virus-interacting proteins [VIPs]) evolve under stronger purifying selection as well as tending to adapt at much high rates compared to similar proteins that do not interact with viruses (Enard & Petrov et al., 2016). It was estimated that ~30% of protein adaptation in the human lineage was accounted for by interaction with viruses (Enard & Petrov et al., 2016). Enard & Petrov hypothesised that some introgressed sequences might have provided a measure of protection against the exchanged viruses and were driven into a recipient species by positive directional selection, as it appears that viruses have driven so much adaptation in the human lineage, and because it is plausible that Neanderthals and humans exchanged viruses either directly by contact or via their shared environment, when they interbred. Several cases of likely adaptive introgression, which are consistent with this model (Gittelman et al., 2016; Racimo et al., 2015, 2017), from Neanderthals to modern humans involve immune genes that are specialised to deal with pathogens which includes viruses (Abi-Rached et al., 2011; Dannemann et al., 2016; Deschamps et al., 2016; Houldcroft & Underdown, 2016; Mendez et al., 2012, 2013; Nedelec et al., 2016; Quach et al., 2016; Sams et al., 2016).

In this study the hypothesis was tested by assessing whether VIPs are enriched in IS overall and, more specifically, in longer and more frequent IS that are more likely to have been driven into the recipient genome by positive selection. Introgression enrichments at VIPs were tested after controlling for stronger purifying selection at VIPs as well as many other potentially confounding factors, because of the probability of introgressed sequences being retained by chance is affected strongly by purifying selection.

The basic logic of this analysis is as follows. If there is positive directional selection soon after interbreeding, it is expected that adaptive Neanderthal introgressed haplotypes increase rapidly in frequency prior to being fragmented by recombination and therefore lead to the presence of long, frequent IS as a result. It is expected that recombination will break up introgressed segments over time, while deleterious alleles that hitchhiked with the adaptive variant(s) should be removed. Enard & Petrov suggest that as a result of this the signal should erode over time. As introgressed segments are scattered across multiple individuals by recombination can, however, be identified and aggregated into single contiguous genomic regions, as has been done (Sankararaman et al., 2014) the original introgressed segment that was adaptive of Neanderthal ancestry might still be identifiable as aggregated segments of Neanderthal ancestry. Also, the frequency and length of such retained regions of Neanderthal ancestry can be assessed.

In this study, a large data set of thousands of VIPs were gathered which showed that they are enriched within longer, more frequent introgressed segments of Neanderthal origin in human genomes, as well as in longer introgressed segments of modern human origin in genomes of Neanderthals. Also, it was found by Enard & Petrov that introgressed segments that react specifically with RNA viruses are particularly enriched in Neanderthal introgressed segments in the genomes of modern Europeans compared to VIPs that interact with DNA viruses. Enard & Petrov have provided a number of arguments which suggest that it is specifically adaption in response to viruses that drove these enrichments. Next, they identified several viruses as likely agents of selection, as well as a number of VIPs as likely targets of adaptation. Finally, they estimated that adaptation overall, and adaptation that was specifically in response to viruses, was an important force in the history of those Neanderthal introgressed segments that were ultimately retained in the genomes of modern humans.

VIPs are enriched in Introgressed Segments from Modern Humans to Altai Neanderthals

Next they tested for an excess of introgressions from modern humans to Neanderthals, using the data on introgressed genomic regions in a single Altai Neanderthal individual (Kuhlwilm et al., 2016). They estimated that the excess of segments of modern human ancestry in the genome of the single Altai Neanderthal individual at VIPs as a function of their size, because adaptive introgressed segments are expected to be longer than neutral ones. They found a large excess of long segments from modern humans at VIPs. Also, the excess is more pronounced in high recombination regions of the genome as had been predicted.

Identifying ancient viruses responsible for adaptive introgression

Next, they sought to find if it is possible to identify which ancient viruses might be responsible for the enrichments that had been observed. Such analyses in the direction from modern humans to Neanderthals is severely underpowered with only 19 VIPs found in introgressed segments over 100 kb in the Altai Neanderthal, in modern humans the number is much larger with 152 VIPs found in long (≥100 kb) and frequent (≥15%) Neanderthal introgressed segments.

Enard & Petrov used the 20 modern human viruses that interact with 10 or more VIPs as proxies for the ancient related viruses that affected humans at the time of interbreeding. These 20 viruses are distributed evenly between RNA viruses (2,684 VIPs) and DNA viruses (2,547 VIPs). 1,563 of the 2,684 RNA VIPs interact only with RNA viruses, while 1,426 out of 2,547 DNA VIPs interact only with DNA viruses.

First they investigated if ancient DNA or RNA viruses were more likely to be involved and they expected that the RNA viruses should be more likely to drive adaptive introgression as they are more likely to move from one species to another (Geoghegan et al., 2017; Kreuder Johnson et al., 2015). They used the bootstrap test in order to determine whether introgression was skewed towards either RNA or DNA viruses, comparing the number of introgressed segments at VIPs that interact with only 1 RNA virus with the number of introgressed segments at VIPs that interact with only 1 DNA virus and are located far from any RNA VIP (≥500 kb) (STAR Methods).

They failed to detect any skew in favour of RNA virus VIPs in East Asia. In Europe, in contrast, they detected a strong bias of RNA virus VIPs in long, high frequency introgressed segments. In the regions of high recombination this pattern was more pronounced. The enrichment of Neanderthal RNA VIPs was even more pronounced when they repeated the comparison after excluding genes that were known to interact with bacteria, Plasmodium (Ebel et al., 2017) had been excluded, and immune genes annotated as such by the gene ontology databases (The Gene Ontology consortium, 2017). Other pathways therefore appear to be unable to explain the signal at RNA VIPs. Also, the background selection at RNA VIPs that was slightly stronger than at control DNA VIPs in East Asia as well as Europe (7% stronger in both cases, p<10^-3) makes the comparison conservative. RNA VIPs have slightly fewer segregating variants (9% less in Europe, p<^-3) and therefore slightly more sites that are under strong purifying selection than control DNA VIPs, which is again conservative. The enrichment at RNA VIPs was confirmed further by using only LT-VIPs or a different recombination map.

Next, they attempted to identify which families of ancient RNA viruses might explain the skew towards RNA VIPs that was observed in Europeans. Included among the 11 RNA viruses in this analysis, influenza A virus (IAV, an orthomyxovirus) and hepatitis C virus (ICV, a flavivirus) have by far the highest number of VIPs. They found that it appears that HIV-only and IAV only VIPs  were each associated with a large excess of high frequency, long adaptive introgressed segments in modern humans from Europe compared to VIPs that interact with only 1 DNA virus. For HIV only and IAV only VIPs within regions of high recombination the excess was particularly strong. Specifically, it was found that 7 (versus 0.29 expected) high frequency (≥15%) introgressed segments overlapping HIV only VIPs (p<10^-3).

A significant enrichment at IAV only LT-VIPs was not detected, though these results were robust when restricting to HIV only LT-VIPs. Enard & Petrov suggest it is possible that the smaller number of IAV only LT-VIPs (56 overall versus 195 for HIV and only 15 regions of high recombination) provided insufficient power to detect a significant excess of introgression. The power was reduced enough to eliminate statistical significance by subsampling the HIV only LT-VIPs to the number of IAV only LT-VIPs (bootstrap test p>0.05 for all LT-IAV only LT-VIPs and all 10 random samples and all introgression lengths and frequencies.

Though no significant enrichment on introgressed segments at HCV only VIPs were detected, it is possible that this might also be an issue of statistical power though HCV has far fewer unique VIPs than HIV and IAV (157 versus 405 and 490, respectively. Subsampling of HIV only and IAV only VIPs has the result that there is insufficient power to detect any excess due to a small sample size, which leaves open the possibility that HCV-like viruses might also been involved.

VIPs and specific host functions

Statistical associations are represented by the enrichments of VIPs in introgressed segments, and in order to determine causality more evidence is required. A particular worry is that biological functions of the host that are enriched significantly among VIPs might also be enriched in adaptive introgressed segments independently of their interactions with viruses. The observed enrichment in any such case of adaptive introgression at VIPs would not be due to interactions with viruses, but it would be due to the uneven representation of diverse host functions between VIPs and non-VIPs instead.

Consider the hypothetical example in which genes which have the biological Gene Ontology (GO) function “cell cycle” tend to be enriched in introgressed segments independently of their interactions with viruses. This could by itself lead to the enrichments of VIPs in the introgressed segments. Enard & Petrov tested this possibility by assessing if the biological functions overrepresented among VIPs (≥1.5 enrichment, permutation test p ≤ 0.05) are enriched within introgressed segments overall or within introgressed segments that are particularly long and/or frequent introgressed segments. If they are not, then this is an unlikely explanation for the results obtained by Enard & Petrov. If they are, then this would be a worry and would require the carrying out of additional analyses to control for this bias. There is no evidence that VIPs show any biased signals (Wilcoxon rank sum test p > 0.05 in all cases) in their presence in introgressed segments.

Crucial immune and proviral VIPs are overrepresented in introgressed segments

Enard & Petrov assumed that the viruses were responsible for the enrichments that were observed; they then asked if VIPs in introgressed segments tend to possess any particular functions. They estimated the overrepresentation of GO annotations of biological functions at VIPs inside introgressed segments compared to VIPs outside introgressed segments by using a permutation test (STAR Methods. They used the introgressed segments of at least the length and the frequency they first started observing highly significant (bootstrap test p < 10^-3 in Europe and East Asia) enrichments in introgressed segments at VIPs (longer than 50 kb and at frequencies that were higher than 5%).

Multiple functions related to immune response or to crucial steps of the viral replication cycle were significantly over-represented among VIPs with the introgressed segments in Europe, though not in East Asia, which is in line with a possible causal role for viruses. When using only the manually curated LT-VIPs, this pattern was particularly pronounced, which may reflect the fact that LT-VIPs are annotated better. The GO annotation “immune effector process”, e.g., was one of the most strongly overrepresented in Europe (all VIPs: 57 VIPs versus 34 expected by chance, permutation test p = 5 x 10⁴; LT-VIPs: 49 verses 27.5 expected by chance, p = 10^-4). It is important that the “immune effector process” function was not over represented among non-VIPs within introgressed segments (permutation test p > 0.05 for all GO annotations in Table S6).

According to Enard & Petrov it is interesting that the GO function for a crucial early step of infection, “virion attached to host cell,” was among the most strongly overrepresented of GO functions among all VIPs, in Europe (5 VIPs instead of 0.8 that expected by chance, permutation test p < 10^-7), and in East Asia (4 VIPs versus 0.75 expected by chance, p = 5.10^-3). All virion attachment VIPs are LT-VIPs. Of the 5 attachment VIPs 4 were shown to be involved in RNA virus attachment (ICAM1, CD209/DCSIGN, HSP90AB1, SLEC4M) and 1 in DNA virus attachment (PVERL2). 3 of the 4 virion attachment VIPs that interact with RNA viruses interact with HCV VIPs (CD209/DCSIGN, HSP90AB1, CLEC4M; HCV VIPs: 3 versus 0.4 expected virion attachment VIPs, p = 7.10^-4.).

“Viral genome replication” is another GO annotation category representing an important step in the viral replication cycle, was also overrepresented strongly particularly at LT-VIPs within introgressed segments in Europe (all VIPs: 147 VIPs versus 7.8 expected, p = 0.029 and LT-VIPs: 17 LT-VIPs versus 6.9 expected, p = 1.6 x 10^-3). Enard & Petrov consider it intriguing that a large number of these “viral genome replication” VIPs were also VIPs that interact with HCV as in the case for virion attachment (LTF, EIF2AK2/PKR, ATG5, MAVS, CD209/DCSIGN, CLEC4M, VAPB; All HCV VIPs: 7 versus 1.7 viral genome replication VIPs, p = 0.009; HCV LT-VIPs: 7 viral genome replication VIPs versus 1.5 expected, p = 10^-4).

Therefore the representation of the RNA and DNA viruses within the function that is overrepresented matches the previous findings by Enard & Petrov of a stronger effect of RNA viruses in Europe and suggests strongly a causal role of viruses.

Discussion

In this paper Enard & Petrov present evidence that a substantial proportion of introgressed segments from Neanderthals to modern humans and vice versa are enriched strongly for proteins that interact with viruses. Also, they detected a particularly strong signal in Europeans for VIPs interacting with RNA viruses. The annotations of host-virus interactions that were more comprehensive that Enard & Petrov used as well as the controls for multiple genomic factors explain why such signals were not noticed in earlier functional analysis of introgressed segments (STAR methods).

Altogether, their results indicate that adaptive introgression in response to viruses might have been more prevalent than was known previously based on the small number of examples that have been published (Abi-Rached et al., 2011; Dannemann et al., 2016; Mendez et al., 2012, 2013; Nedelec et al., 2016; Quach et al., 2016; Sams et al., 2016). Enard & Petrov estimate that f all the long (≥100 kb) and high frequency (≥ 15%) introgressed segments from Neanderthals to modern humans, 32% (54 of 171) in Asians and 25% (27 of 105) in Europeans might have been selected positively in response to viruses.

Note that the model of positive directional selection of introgressed segments was specifically tested in this study soon after interbreeding which is expected to drive long introgressed segments of Neanderthal ancestry to high frequency into the population. Enard & Petrov did not exclude the possibility that the integrated segments were maintained subsequently by balancing selection of selection that was frequency dependent of some kind. This possibility is intriguing, especially if the increased genetic variability that was introduced by introgression was advantageous in variable environments.

Overall, preliminary support for the “poison-antidote” hypothesis is provided by these results under which each species was exposed to novel viruses between modern humans and Neanderthals while a measure of resistance was provided by gene flow by allowing VIPs that were already adapted to the presence of specific viruses in the donor species to cross species boundaries and provide adaptive function in the recipient species.

In this respect, it is particularly interesting that the greater enrichment of Neanderthal ancestry at RNA VIPs is restricted to Europe. It has been concluded by several authors that there were multiple pulses of interbreeding between Neanderthals and modern humans, at least one pulse prior to the split between Asian and European populations of modern humans, and multiple independent pulses following the split of these populations (Kim & Lohmueller, 2015; Vernot & Akey, 2015; Villanea & Schraiber, 2018). It is then suggested by the strong enrichment at RNA VIPs in Europe though not in East Asia that the difference arose as the result of interbreeding pulses after the split between Europeans and Asians. It remains an open question whether adaptive introgression at DNA VIPs in East Asians compensated for a previous bias towards RNA VIPs that are still visible in Europeans, or adaptive introgression at RNA VIPs occurred specifically after an independent pulse of interbreeding in Europe.

Enard & Petrov believe that the evidence that has been gathered in favour of the poison-antidote scenario is preliminary, in spite of the strong statistical signals. Though the enrichments they describe are defined rigorously, they represent only statistical associations. They believe that to establish the causal impact of the virus-host interactions on the patterns of adaptive introgression more functional work will be required. Such evidence is, intriguingly, beginning to accumulate. It has recently been found (Quach et al., 2016) that regulatory variants that affect host gene expression during IAV infection are enriched particularly strongly in Neanderthal ancestry in Europeans. Also, it has been found recently (Sams et al., 2016) that flaviviruses – a class of RNA virus that include HCV, Dengue, as well as other viruses – might have been driving selection at the adaptively introduced Neanderthal haplotype at the OAS1 locus.

According to Enard & Petrov the analysis presented in this paper opens the door to more functional studies of this kind. Some tantalising patterns have been revealed by this cursory look at the functional data. Several of the RNA virus VIPs that have a low FDR for adaptive introgression are known to affect specific steps of the cycle of viral replication, e.g., and the frequency of specific functional variants that plausibly increased resistance to viruses, was increased by introgression in some cases. It has been shown, e.g., that IAV VIP PPIE (aka cyclophilin E) inhibits the viral ribonucleoprotein complex that is required for replication of IAV RNA (Wang et al., 2011b). In Europeans introgression at the PPIE locus has increased the frequency of allelic variants of expression quantitative trait loci (eQTL) SNPs that are associated with high expression of PPIE across many tissues, including the lungs (GTXe Consortium, 2015). According to Enard & Petrov it is possible that such an increased expression results in greater inhibition of the viral nucleoprotein complex and therefore increased resistance to viruses.

Toll-like receptor 2 TLR2, a HIV VIP is another noteworthy example where introgression increased the frequency of alleles at multiple linked eQTL SNPs associated with higher TLR2 expression in many tissues (GTXe Consortium, 2015). It has now been shown that HIV protein ENV binds and inhibits TLR2 activity (Reuven et al., 2014), which suggests that increased expression might have prevented, at least partially, such inhibition for related lentiviruses.

Also Enard & Petrov suggest it will be interesting to study whether the presence of Neanderthal ancestry at VIPs leads to variability to modern viruses in modern humans. One of the 5 virion attachment HCV VIPs that has been found to be highly enriched in introgressed segments in Europe, the C-type lectin receptor CD209/DCSIGN) Crucial Immune and Proviral VIPs are Over-Represented in the Introgressed Segments, in results), interacts with HCV and Dengue virus. At this locus introgression from Neanderthals has affected the frequency of alleles at a well-known variant (rs4804803) within the promotor of DCSIGN with an effect that has been documented on CD209/DCSIGN expression and on HCV and dengue virus infection severity (Ryan et al., 2010; Wang et al., 2011a).

Though these scenarios are clearly speculative, they demonstrate the power of the enrichment approach of Enard & Petrov to systematically formulate plausible hypotheses that are testable for specific adaptive introgression candidates. Many of the genes that are identified are at low enough FDR that it is sensible to carry out functional studies on the introgressed variants.

Finally, it is suggested by these results that genomes and other species contain signatures of arms races from the past with diverse viruses and other pathogens, which makes it possible to use host genomic signatures to study ancient interactions with viral and other pathogens that are ever present and ever shifting.

Sources & Further reading

Enard, D. and D. A. Petrov (2018). "Evidence that RNA Viruses Drove Adaptive Introgression between Neanderthals and Modern Humans." Cell 175(2): 360-371.e313.