While DNA was successfully extracted from the finger bone and informed scientists that the Denisovans were genetically distinct from Neanderthals and modern humans, it did not contain enough detail to provide them with much information about their anatomy; it could only tell them that the child had dark hair, skin and eyes.
Now, however, scientists from the Hebrew University of Jerusalem have used data obtained through something called ‘DNA methylation’ to successfully reconstruct an image of the Denisovans.
By comparing DNA methylation activity between Denisovans, Neanderthals and modern humans, the researchers found areas in the genes that were different between the species. They then looked for evidence about what the Denisovans might have looked like, based on what’s known about human disorders when those same genes lose their function.
“By doing so, we can get a prediction as to what skeletal parts are affected by differential regulation of each gene, and in what direction that skeletal part would change – for example, a longer or shorter femur,” said David Gokhman, who co-authored the research.
In total, the researchers discovered 56 features – 34 of them in the skull – in which Denisovans were different from Neanderthals and/or humans. They found that the Denisovans probably shared Neanderthal traits, like a longer face and wider pelvis, yet also had differences unique to the species, such as a larger dental arch and expanded skull.
“Studying Denisovan anatomy can teach us about human adaptation, evolutionary constraints, development, gene-environment interactions, and disease dynamics,” said Liran Carmel, who co-authored the research. “At a more general level, this work is a step towards being able to infer an individual’s anatomy based on their DNA.”
Sequencing the Neanderthal genome (Green et al., 2010, Prüfer et al., 2014), the Denisovan genome (Reich et al., 2010), and several early modern human genomes from Eurasia (Fu et al., 2014, Fu et al., 2015) has confirmed that archaic hominins left their mark in the genomes of modern humans (Plagnol and Wall, 2006, Sankararaman et al., 2014, Vernot and Akey, 2014, Vernot et al., 2016). Present-day individuals in Eurasia inherited ∼2% of their genome from Neanderthals (Green et al., 2010), and individuals from Oceania inherited ∼5% of their genome from Denisovans (Reich et al., 2010). Suggestive evidence indicates that admixture from other unidentified hominin species occurred in Africa (Hammer et al., 2011, Hsieh et al., 2016, Lachance et al., 2012, Plagnol and Wall, 2006, Wall et al., 2009).
To understand the functional, phenotypic, and evolutionary consequences of archaic admixture, it is necessary to identify the specific haplotypes and alleles that were inherited from archaic hominin ancestors (Huerta-Sánchez et al., 2014, Juric et al., 2016, Sankararaman et al., 2014, Simonti et al., 2016, Vernot and Akey, 2014). Approaches to identifying introgressed haplotypes include methods that specifically incorporate reference archaic hominin genome sequences and reference-free methods that do not utilize such information. An example of the former category is the method of Sankararaman et al. (2014), which identifies archaic haplotypes by comparing modern human haplotypes to a reference archaic sequence. The latter category of methods include the S∗ statistic (Plagnol and Wall, 2006), which searches for the mutational signature that ancient admixture leaves in the genomes of present-day humans.
The S∗ approach is powerful for finding introgressed haplotypes in the absence of an archaic reference genome because it leverages the unusual mutational characteristics of introgressed haplotypes. Because of the long divergence time between Neanderthals and modern humans, Neanderthals carry many alleles that are specific to their lineage. Such alleles are present on introgressed haplotypes but are absent or rare in African genomes. Further, based on the recent timing of admixture, introgressed haplotypes are expected to be maintained without recombination over distances of approximately 50 kb on average (Sankararaman et al., 2012), resulting in high levels of linkage disequilibrium (LD) between Neanderthal-specific alleles in non-African human genomes.
In this study, we develop an S∗-like method that has increased power and is suitable for large-scale genome-wide data. We apply the method to large sets of sequenced data from Eurasia and Oceania and identify putative archaic-specific alleles. We examine the rate at which these alleles match the sequenced archaic genomes and the role of the genes containing these alleles, to obtain insights into the history of the admixture events and their impact on modern human genomes.
Figure 3 Detection and Match Rates in 1000 Genomes Populations
In the UK10K study (UK10K Consortium et al., 2015), we find 304 Mb of the genome covered by one or more detected segments, and the average proportion of haplotypes carrying a detected segment at a position is 0.63%. This is lower than in the 1000 Genomes European populations. The lower detection rate in the UK10K may reflect characteristics of the methods used to generate this dataset.
Papuans, in addition to Neanderthal ancestry, harbor significant amounts of Denisovan ancestry. In the Papuans from the Simons Genome Diversity Project (SGDP) (Mallick et al., 2016), we find 239 Mb of the genome covered by one or more detected segments, and the average proportion of haplotypes carrying a detected segment at a position is 1.48%.
In the 1000 Genomes Eurasian populations, the detected putative introgressed haplotypes have median lengths ranging from 59 kb in Bengalis (BEB) to 71 kb in Finns (FIN). Due to tiling across individuals, the full segments that our method reports can be much longer (Figure 2). In the Eurasian 1000 Genomes populations, the median segment length varies from 205 kb in Iberians (IBS) to 239 kb in Telugus (ITU). The longest detected segment is 7.9 Mb.
Simulation Results We first verify the accuracy of our method and compare it with two previous versions of S∗ designed for windowed analysis of genome-wide data (Vernot and Akey, 2014, Vernot et al., 2016). The main differences between these two versions of S∗ is in the number of target individuals analyzed simultaneously. The 2014 version of S∗ analyzes subsets of 20 individuals (Vernot and Akey, 2014), similar to the original gene-based S∗ (Plagnol and Wall, 2006), whereas the 2016 version analyzes one individual at a time; doing so avoids potential effects of population structure (Vernot et al., 2016). For both methods, we use a sliding window of 50 kb, with a step size of 10 kb. Previous analyses using these methods also used sliding windows of size 50 kb and step sizes of 10 or 20 kb (Vernot and Akey, 2014, Vernot et al., 2016). We see that our method (Sprime) has a much better trade-off of detection frequency to accuracy than these previous versions (Figure 1). Several factors contribute to Sprime’s superior performance. One factor is that our method avoids windowing, which allows it to build up power across larger regions of tiled introgression (Figure 2). When we apply our method in 50-kb windows, its performance drops considerably but still remains higher than that of the previous S∗ versions. Another factor is our simultaneous analysis of a larger number of individuals, although the difference in performance between analyzing 100 individuals and analyzing 15 individuals with our method is not very large. Other likely contributing factors for the difference in performance include our method’s allowance for migration from introgressed populations to the simulated African outgroup, and our method’s different scoring function.
Figure 1 Comparison of Our Method (Sprime) with a Previous Method (S∗) on Simulated Data
We verified that our method is robust to variability in mutation rate, allele frequency, and demographic history. Accuracy remains high across a range of mutation rates, provided the sample size is at least 15 (Figure S1). The accuracy of reported putative archaic-specific alleles is over 93% across the range of allele frequencies, with highest accuracy (over 98%) for the lower frequency alleles (frequency < 0.02). Detection frequency varies with mutation rate. With a constant mutation rate of Math Eq per base pair per meiosis, approximately half of the introgressed material is detected. The remaining introgressed material cannot be confidently identified because the introgressed segments are too short and/or the local mutation rate is too low. Across a wide range of demographic histories, with differing archaic-human split times and admixture times, accuracy remains at least 93% (Figure S2).
Figure 2 Tiling of Introgressed Haplotypes
We analyzed each non-African population from the 1000 Genomes Project (1000 Genomes Project Consortium, 2015) (Table 1). Across the 19 European, Asian, and American populations, we find 1.36 Gb of the genome covered by putative introgressed segments. In individual populations, coverage ranges from 382 Mb in Peruvians (PEL) to 655 Mb in Bengalis (BEB), and the average proportion of haplotypes carrying a detected segment at a position ranges from 0.80% in Puerto Ricans (PUR) to 1.23% in Han Chinese (CHB and CHS) (Figure 3). These detection rates are around half of the estimated introgression proportions obtained using Math Eq-ratio statistics (Prüfer et al., 2017). This is in line with the simulation results, in which half the introgressed material can be detected, whereas the other introgressed segments are too short for confident detection. East Asian populations have higher introgression detection rates than do European populations, consistent with previous reports of higher Neanderthal introgression rates in East Asians than in Europeans (Meyer et al., 2012, Sankararaman et al., 2014, Vernot and Akey, 2014, Wall et al., 2013). The South Asian populations and European populations have similar rates of detected introgression, a fact that has been previously reported (Vernot et al., 2016).
Figure S1 Power and Accuracy in Simulated Data as a Function of Mutation Rate and Sample Size, Related to Figure 1
Figure S2 Detection Frequency and Accuracy in Simulated Data as a Function of Population Split and Admixture Times, Related to Figure 1
Table 1 Samples Analyzed Identifier Description 1000 Genomes Region Number of Individuals BEB Bengali from Bangladesh South Asia 86 CDX Chinese Dai in Xishuangbanna, China East Asia 93 CEU Utah Residents (CEPH) with Northern and Western European Ancestry Europe 99 CHB Han Chinese in Beijing, China East Asia 103 CHS Southern Han Chinese East Asia 105 CLM Colombians from Medellin, Colombia America 94 FIN Finnish in Finland Europe 99 GBR British in England and Scotland Europe 91 GIH Gujarati Indian from Houston, TX, USA South Asia 103 IBS Iberian Population in Spain Europe 107 ITU Indian Telugu from the United Kingdom South Asia 102 MXL Mexican ancestry from Los Angeles, CA, USA America 64 JPT Japanese in Tokyo, Japan East Asia 104 KHV Kinh in Ho Chi Minh City, Vietnam East Asia 99 PEL Peruvians from Lima, Peru America 85 PJL Punjabi from Lahore, Pakistan South Asia 96 PUR Puerto Ricans from Puerto Rico America 104 STU Sri Lankan Tamil from the United Kingdom South Asia 102 TSI Toscani in Italia Europe 107 YRI Yoruba in Ibadan, Nigeria Africa 108 UK10K TwinsUK and Avon Longitudinal Study of Parents and Children (ALSPAC) – 3,781 Papuans From SGDP. Sampling location is 4S 143E (East Sepik province of Papua New Guinea) – 15 SGDP Africans From SGDP – 44
Comparison to Sequenced Archaic Genomes Our method infers putative archaic-specific alleles. If an archaic reference sequence exists, we can determine the proportion of putative archaic alleles that match the reference sequence. Some putative archaic-specific alleles cannot be compared to an archaic genome because of the masking filters that we applied (see the STAR Methods) to eliminate questionable regions due to factors such as low coverage or poor mappability. The match rate that we report is the proportion of matches for unmasked alleles.
In the 1000 Genomes European populations the overall match rate to the sequenced Altai Neanderthal genome is 0.719 (Figure 3). By considering the larger UK10K sample, we can investigate the effect of allele frequency in detail. In the UK10K analysis, the rate of matching of the detected alleles to the Altai Neanderthal is fairly constant across the full range of allele frequencies, with an overall rate of 0.743. In contrast, randomly selected alleles that, like the putatively archaic-specific alleles, are at frequency < 0.01 in the West African outgroup have a very low rate (0.034) of matching to the Altai Neanderthal (Figure S3). This demonstrates that the match rate achieved by our method is much higher than would be found if a high proportion of the putative archaic-specific alleles were false positive. The match rate to the Altai Neanderthal and Altai Denisovan genomes is lower in the American populations than in the other 1000 Genomes populations (Figure 3). This is likely because the American populations are admixed and thus have higher background levels of LD that could cause false-positive results.
Figure S3 Rate of Matching to the Altai Neanderthal Genome as a Function of Allele Frequency in the UK10K Data, Related to Figure 3
In order to look more closely at the Neanderthal and Denisovan ancestry in present-day humans, we plot two-way densities of match rate to the Altai Neanderthal and Altai Denisovan genomes for segments with at least ten positions that can be compared to the Altai Neanderthal and at least ten positions that can be compared to the Altai Denisovan (Figure 4). In each population, we see a large cluster of segments with high matching to the Altai Neanderthal and low matching to the Altai Denisovan. This cluster corresponds to segments introgressed from Neanderthals. In each population the mode of matching to the Altai Neanderthal for this cluster is approximately 0.8, whereas the mode of matching to the Altai Denisovan genome is approximately 0.2. Thus approximately 20% of the archaic-specific variants introgressed from Neanderthals are also carried by the Altai Denisovan due to the relatedness of the Neanderthal and Denisovan populations, whereas 80% of the archaic-specific variants introgressed from Neanderthals are present in the Altai Neanderthal. In each population we also see a small cluster of segments with almost no matching to the Altai Neanderthal or to the Altai Denisovan; these are likely to be false-positive results that do not correspond to archaic introgression. In the Asian and Papuan populations we see a third cluster of segments. The segments in this third cluster have high matching to the Altai Denisovan and low matching to the Altai Neanderthal. This cluster corresponds to segments introgressed from Denisovans and confirms the previous finding of Denisovan admixture in Papuans and in Asians (Prüfer et al., 2014, Qin and Stoneking, 2015, Sankararaman et al., 2016, Skoglund and Jakobsson, 2011). Figures 4 and 5 also indicate that several other populations may carry a small proportion of segments introgressed from Denisovans. These include the Finns, who are estimated to have obtained around 7% of their ancestry from East Asia (Sikora et al., 2014), and admixed American populations whose Native American ancestors are related to East Asians (Gutenkunst et al., 2009).
Figure 4 Contour Density Plots of Match Proportion of Introgressed Segments to the Altai Neanderthal and Altai Denisovan Genomes
In the Japanese and Chinese (Dai, Beijing, and Southern Han) populations we see that the Denisovan cluster of segments has a wide and bimodal distribution of match rates to the Altai Denisovan genome (Figure 4). A test for two distinct components of Denisovan ancestry (see the STAR Methods) is statistically significant (p < 0.05 after adjusting for multiple testing) in each of these four populations (Table 2) but is not significant in the other 1000 Genomes populations. The fitted two-component mixture has approximately one-third of the Denisovan segments in the Japanese and Chinese populations coming from the component with higher affinity to the Altai Denisovan genome. The putative archaic-specific alleles in the high-affinity component have a match rate of around 80% to the Altai Denisovan genome, which is similar to the match rate of putative archaic-specific alleles in Neanderthal introgressed segments with the Altai Neanderthal, whereas the putative archaic-specific alleles in the other (moderate-affinity) component have a match rate of around 50% to the Altai Denisovan genome.
Table 2 Two-Component Mixtures for Denisovan-Related Introgression
Population Math Eq Math Eq Math Eq Math Eq Math Eq p value Southern Han Chinese 0.82 0.46 0.08 0.12 0.42 0.00002 Han Chinese (Beijing) 0.84 0.50 0.08 0.14 0.36 0.00021 Chinese Dai 0.86 0.52 0.04 0.18 0.20 0.00069 Japanese (Tokyo) 0.86 0.52 0.06 0.18 0.26 0.00143 Finnish 0.84 0.50 0.04 0.14 0.22 0.00348 Punjabi (Pakistan) 0.82 0.48 0.10 0.12 0.10 0.04589
To check that the moderate-affinity component is not due to segments that are a mosaic of Neanderthal and Denisovan ancestry, we reran the two-component mixture test excluding segments containing any Neanderthal-specific alleles (putative archaic-specific alleles matching the Neanderthal genome but not the Denisovan genome). We find that the same four populations (the three Chinese populations and the Japanese population) still have statistically significant p values for a two-component mixture after adjusting for multiple testing (Math Eq), and the estimated mixture parameters are essentially unchanged.
Based on the mode of matching to the Denisovan genome, most of the Denisovan ancestry in the South Asian and Papuan populations is from the archaic component with moderate affinity to the Altai Denisovan (Figure 4). This is consistent with previous work that noted that the Altai Denisovan is significantly more distantly related to the introgressing Denisovans compared to the relationship between the Altai Neanderthal and the introgressing Neanderthals (Prüfer et al., 2014).
Figure 5 Mean Amounts of Detected Introgressed Material per Individual, Classified by Affinity to the Altai Neanderthal and Altai Denisovan Genomes
To facilitate further analyses, we extracted subsets of segments based on their affinity to the Altai Neanderthal and to the Altai Denisovan (see the STAR Methods). We performed several analyses to check for possible confounders of match rate to the Denisovan genome. We checked whether the divergence between the Altai Neanderthal and Altai Denisovan differs between regions covered by the moderate-affinity Denisovan introgression and the high-affinity Denisovan introgression in case such differences could account for the two components. In the East Asian data, the mean relative divergence (number of homozygous discordances between the Altai Neanderthal and Altai Denisovan divided by the number of 1000 Genomes variants) per segment was 1.65 (SE 0.26) for high-affinity Denisovan segments and 2.51 (SE 1.00) for moderate-affinity Denisovan segments. The difference is not statistically significant (p > 0.05). We also investigated the average density of putative archaic-specific variants in segments attributed to the different components. We adjusted for length of the detected segments, because the power to detect segments increases with both length and the density of archaic-specific variants. In the East Asian data, the adjusted mean inverse density (bp per archaic-specific variant) was 103 (SE 440) for the high-affinity Denisovan segments, 395 (SE 464) for the moderate-affinity Denisovan segments, and 1,164 (SE 72) for the Neanderthal segments. The difference is not statistically significant (p > 0.05). Thus we do not find confounding by divergence or by density of archaic-specific alleles.
We investigated the lengths of haplotypes within segments attributed to the different components in order to investigate potential differences in admixture time between components. We analyzed haplotype lengths in units of centimorgans (cM) rather than base pairs because centimorgan distances reflect recombination and are thus less variable. We adjusted for frequency and overall segment length because high frequency and high segment length increase power to detect a segment and are correlated with haplotype length. In the East Asian data, the mean adjusted haplotype length was 0.066 (SE 0.014) cM for Neanderthal segments, 0.19 (SE 0.13) cM for high-affinity Denisovan segments, 0.072 (SE 0.13) cM for moderate-affinity Denisovan segments, and 0.13 (SE 0.06) cM for Denisovan segments overall. These are not significantly different. We also checked for differences in Europeans, in South Asians, in Asians overall (East and South), and in Papuans, again finding no significant differences. While it is probable that the Neanderthal admixture and the two waves of Denisovan admixture occurred at distinct times, there is insufficient information in the data to determine the ordering of these events.
Overall, East Asians and South Asians carry similar amounts of detected Denisovan ancestry, while Papuans carry much more detected Denisovan ancestry (Figure 5). Approximately one-third of the Denisovan ancestry segments in the East Asians are from the high-affinity component (Table 2), whereas very little of the Denisovan ancestry in the South Asians and Papuans is from the high-affinity component (Figure 4). A possible scenario consistent with this pattern would have the high-affinity component introgressing into East Asia after the split between East and South Asia. Because the Papuans have a much higher frequency of the moderate-affinity Denisovan component than other populations, it may be that this component was primarily introgressed into the ancestors of Papuans after they split from Asia, and arrived in Asia via migration from the ancestors of Papuans; however, other scenarios are also possible (Prüfer et al., 2014, Sankararaman et al., 2016).
Lack of Evidence for Multiple Waves of Neanderthal Ancestry The frequency of Neanderthal introgression is substantially higher (∼30%) in East Asians than in Europeans (Meyer et al., 2012, Wall et al., 2013). This difference cannot be explained by differential effects of selection, but could be due to an additional Neanderthal admixture event into the ancestors of East Asians after the Europe-Asia split (Kim and Lohmueller, 2015, Vernot and Akey, 2015). Another possible explanation would be dilution of Neanderthal admixture in Europe due to migration from a population without Neanderthal admixture (Meyer et al., 2012, Vernot and Akey, 2015).
In our results, the Neanderthal-introgressed segments in East Asians and in Europeans show indistinguishable levels of similarity to the Altai Neanderthal (Figure 4). There is also no clear difference between East Asians and Europeans in the similarity of their Neanderthal-introgressed segments to the Vindija 33.19 Neanderthal (Figure S4). Thus, if the ancestors of East Asians received a large pulse of Neanderthal admixture after splitting from Europeans, then the original (shared Eurasian) and additional (East Asian-specific) admixing populations must have been closely related.
Figure S4 Contour Density Plots of Match Proportion of Introgressed Segments to the Altai Neanderthal and Vindija 33.19 Neanderthal, Related to Figure 4
We looked for introgressed segments with highest frequency in 1000 Genomes populations. Specifically we found in each population the two regions of highest frequency that had high matching to the Altai Neanderthal or Altai Denisovan genome (see the STAR Methods). Table S1 lists the regions. All these regions appear to have been introgressed from Neanderthals rather than Denisovans. Several of the positively selected regions have been described previously, including BNC2, POU2F3, and KRT71, which are involved in skin and hair traits (Sankararaman et al., 2014, Vernot and Akey, 2014). Genomic regions introgressed from Neanderthals and under positive selection have been shown to be enriched for genes involved in pigmentation and immunity (Deschamps et al., 2016, Gittelman et al., 2016, Racimo et al., 2015, Sankararaman et al., 2014, Sankararaman et al., 2016, Vernot and Akey, 2014, Vernot et al., 2016).
In addition to the regions that have been extensively described in previous studies of positively selected archaic introgression, our results include two immunity-related regions, which we highlight here. The first of these immunity-related regions is on chromosome 3p21.31. This region was included in a supplementary table of high-frequency introgressed haplotypes in Gittelman et al. (2016), but was not discussed in that work. The introgressed alleles at this locus are at high frequency in South Asia (0.38). The region contains CCR9 (C-C motif chemokine receptor 9) and CXCR6 (C-X-C motif chemokine receptor 6), which are chemokine receptors involved in immunity (Papadakis et al., 2000, Paust et al., 2010, Zlotnik and Yoshie, 2000).
The second of these immunity-related regions is on chromosome 14q32.33. The introgressed alleles in this region are at very high frequency throughout Eurasia. This region is located in the immunoglobulin heavy locus, which contains multiple genes that code for antibodies (Schroeder and Cavacini, 2010). Immunoglobulin heavy genes contained within the high-frequency region are IGHA1, IGHG1, and IGHG3. The most highly conserved introgressed position is rs10144746 (PhyloP score 4.1) and is an expression quantitative trait locus (eQTL) for IGHG4 and several other immunoglobulin heavy genes in various tissues including esophagus and liver. The high-frequency introgression is in a region with significant masking of the Altai Neanderthal and Altai Denisovan genomes due to poor quality sequence. For example, for the segment found in the Southern Han Chinese (CHS) population, 119 of the 145 putatively introgressed alleles are filtered in the Altai Neanderthal genome (see the STAR Methods). Of the 26 unfiltered alleles, 22 match the Altai Neanderthal genome. Thus this region appears to be derived from Neanderthal admixture, but would be difficult to find using a reference-based approach.
Discussion We applied a new method for detecting archaic introgressed segments to worldwide non-African populations from the 1000 Genomes project, Papuans from the SGDP and individuals from the UK10K project. Our method is reference-free, which means that it can detect introgression from archaic admixing populations without a reference sequence. We show that when a reference sequence exists, comparison of the detected segments to the reference genome can lead to new insights into population history.
We found evidence that Asians carry Denisovan introgression, confirming previous reports that used alternative methods (Prüfer et al., 2014, Qin and Stoneking, 2015, Sankararaman et al., 2016, Skoglund and Jakobsson, 2011). Further, we found evidence for two waves of Denisovan admixture, one from a population closely related to the Altai Denisovan individual, and one from a population more distantly related to the Altai Denisovan. The component closely related to the Altai Denisovan is primarily present in East Asians, whereas the component more distantly related to the Altai Denisovan forms the major part of the Denisovan ancestry in Papuans and South Asians. The East Asian populations are the only populations with relatively equal and non-negligible contributions from both components, and it is in these populations that the two waves of Denisovan admixture are most evident.
In contrast, we did not find evidence for two or more waves of Neanderthal admixture from diverged Neanderthal populations. The higher rates of Neanderthal introgression in East Asians relative to Europeans may be due to dilution of Neanderthal admixture in Europeans as a result of migration from a population without Neanderthal admixture (Meyer et al., 2012, Vernot and Akey, 2015). If there was an additional pulse of Neanderthal admixture into East Asians after the Europe-Asia split, then it was from a population closely related to the primary admixing Neanderthals.
We found a number of high-frequency introgressed haplotypes that appear to have been subject to positive selection. Two of these regions are involved in immunity, containing the immunoglobulin heavy locus and a cluster of chemokine receptors. These regions, in addition to previous reports of positively selected introgressed haplotypes in histocompatibility leukocyte antigen (HLA) genes (Abi-Rached et al., 2011), Toll-like receptors (Deschamps et al., 2016), and many other immunity genes (Abi-Rached et al., 2011, Deschamps et al., 2016, Quach et al., 2016, Racimo et al., 2015) underscore the crucial role that Neanderthal introgression played in adapting the human immune system to the pathogenic landscape of Eurasia.
Cell, VOLUME 173, ISSUE 1, P53-61.E9, MARCH 22, 2018