Content and data contained in this blogpost was in reference to a study which since been released after this webpage has been published. A major change should be considered when reading the following. OUB02 designation is that of the Epipaleothic culture, however the archeological site and date has been elucidated by the full article. She was discovered in the cave of Ifri Ouberrid in the village of Aïn Elleuh. Furthermore, though the AGRIWESTMED team were behind excavations of some of the samples in the article, OUB02 wasn’t one of them. Click here to read the full article.

A look at Moroccan aDNA from an upcoming study

In 2018 with the introduction of the Iberomaurasian genetic profile¹⁠, the public, for the first time was given a genetic snapshot of the Epipaleolithic on the African continent. A hint of what could be expected was revealed when two years prior, analysis of a relatively contemporaneous group of individuals, the Natufians, was published²⁠. This group resided in Israel, about 180 kilometers from the closest African region modern day Egypt, and highlighted a genetic component which haven’t been seen before them. Thus dubbing the “Natufian component.” The Iberomaurasians, who some thought to have been of Southern European {Iberian} origin^3–5⁠ were later revealed to be another example of ancestry not previously seen before. They were an unprecedented mixture of what we recognize as Sub Saharan African and Natufian^1,6⁠, a profile which set the tone for speculation of an autochthonous North African component⁷⁠. A few more samples of later periods starting with the Neolithic shown that the genetic composition of the Iberomaurasians had lasted for at least another 5,000 years in Morocco⁸⁠. Following on, neolithization likely preceded with the onset of migrations from Southern Europe and later the Near East and or North East Africa. This scenario is hinted at in an upcoming study in which 9 ancient samples have been uploaded to the European Nucleotide archive from Uppsala University, titled: “Genomic sequence data from Neolithic and Epipaleolithic individuals from northwestern Africa.”

The abstract reads as:

“ In northwestern Africa, lifestyle transitioned from foraging to food-production ~7,400 years ago, but what sparked that change remains unclear. Archaeological data supports conflicting views: that migrant European Neolithic farmers brought the new way of life to North Africa, or that local hunter-gatherers adopted technological innovations. The latter is also supported by archaeogenetic data. Here, we fill key chronological and archaeogenetic gaps for the Maghreb, from Epipalaeolithic to Middle Neolithic, by sequencing the genomes of nine individuals (to between 45.8 and 0.2× genome coverage). Remarkably, we trace 8,000 years of population continuity and isolation from the Upper Palaeolithic via the Epipaleolithic, to some Maghrebi Neolithic farming groups. However, remains from the earliest Neolithic contexts showed mostly European Neolithic ancestry. We suggest that farming was introduced by European migrants and then it was rapidly adopted by local groups. During the Middle Neolithic, a new ancestry from the Levant appears in the Maghreb, coinciding with the arrival of pastoralism in the region and all three ancestries blend together during the Late Neolithic. Our results reveal ancestry shifts in the Neolithization of northwestern Africa that likely mirrored a heterogeneous economic and cultural landscape, in a more multifaceted process than observed in other regions. “

Genomic sequence data from Neolithic and Epipaleolithic individuals from northwestern Africa. (Uppsala University) see ref 24

Of the 9 new North African samples one individual was likely dated to the Iberomaurasian period. Evidence of this possibility stems from the accession ID of the individual “oub02” which is likely an abbreviation of Oued Beth, an archeological region in central Morocco containing sites which was occupied from the Epipaleolithic to the Chalcolithic²⁵. One of the sites, a cave at Ifri n’Amr ou Moussa (IAM) housed previously published samples which represented the earliest Neolithic African genomes to date⁶⁠. It had been noted that the AGRIWESTMED project managed to sequence two ancient individuals from the caves.⁹ ⁠One representing the Neolithic and the other the Epipaleolithic. Within the newly uploaded samples, there is one sample accessioned with an ID “IAM04” which matches the previous Neolithic samples of region. Scientific convention allows it to assume this individual would be representative of the previously published early Neolithic Moroccan population as they share the same ID. Hence, if these two individuals Oub02 and IAM04 are the individuals sequenced by the AGRIWESTMED project, the former must then be of the Epipaleolithic Iberomaurasian and the latter of the early Neolithic.

Outside of the potential age of the sample, what drew my attention to Oub02 in particular was the high coverage. ~45% of her DNA was reassembled. With that in mind I sought to seek out analysis using diploid calls to see if any more inferences on the Iberomaurasian profile can be made. Differences in results of some analysis were seen in the past when Bayira a 4500 year old sample from Mota cave¹⁰⁠ in Ethiopia was re-observed with diploid calls¹¹⁠. Conventionally he was thought to be of an ancestral East African population with slight south African (khoisan) affinities. In the case of the reanalysis it was shown that his ancestry was likely up to 30% related to a ghost African population^11,12⁠. And though he was closely related to contemporary East Africans his line of Ancestry had a more minor contribution to contemporary East Africans which is a notable difference when trying to disentangle African prehistory. Being that for the last half decade, the previous Iberomaurasian samples published of Taforalt’s Grotte des Pigeons, or Taforalt for short, represented a cornerstone in African population history. It is only right that when given an opportunity we should see what conclusions can be made with diploid calls.

Cluster Analysis

With a new “high-coverage” individual in mind, I revisited ADMIXTURE¹⁴⁠ to infer just that; admixture. The goal was to start as blind as possible by not assuming Oub02’s ancestry composition. ADMIXTURE’s algorithms simply assigns probabilities that a group of SNP’s (genetic variants in the form of a single base pair polymorhism) were passed on by an individual or groups of individuals. It makes it one of the least aprioric approaches to initiate analysis on population history via genetics. Secondly, I wanted to see if there’s any notable differences between them and the previously published available Iberomaurasian genomes. So at least two separate analysis was ran each containing an initial and secondary run. One of which including the Taforalt genomes in the initial runs (simulation A) and the other simply projecting the Taforalt genomes in the secondary runs (simulation B). To explain further, aDNA not specified to be diploid called are typically “psuedo-haploid” where a random SNP is chosen at a site when the variants are discovered, imputed or assigned. Therefore there are no heterozygous calls and a higher likelihood of cultivating noise especially when analyzed along with diploid genomes. With ADMIXTURE, an output containing the ancestral likelihoods at the shared sites can be generated along with individual ancestry compositions. This was done in all of the initial runs. We could then use that output to infer ancestry in a greater set of individuals or those of lower coverage in the secondary run.

For all instances I limited the models to up to 12 population clusters (K’s) and at the least, 4. Right away when trying to attribute all the available samples to four lines of ancestry (K4) in each run, we can see variation among the Iberomaurasians. Other populations as expected were more or less identical between runs but that would observably change as the population clusters increase. Nonetheless Taforalt, Oub02 and IAM all show having ~68% Eurasian ancestry in lime green. The other three components are African components related to Southern African Huntergatherers (Pink), Western Africans(Yellow) and East Africans(red). Interestingly, Taforalt in particular shows quite a bit of variation between runs. For instance, when included in the initial run, they show no or negligable amounts of South African and East African related ancestry, however when projected they show non-negligable amounts of both ancestries. Oub02 and IAM, are rather consistent in showing West and East African related ancestry at similar amounts in both runs.

As the clusters increase things get more interesting due to the fact that the Iberomaurasians form their own component as early as K5. As a result, the ancestry compositions of all other populations become a little more unexpected due to the subtle and not-so subtle variation between Oub02 and Taforalt. In theory if the aforementioned two populations were the same, they’ll consistently form a tight cluster which would show predictable distribution between runs and through all K’s. But that’s not what occurred here. In the cases where the initial runs included Taforalt, they formed a cluster to which Oub02 and IAM shared most but not all of their ancestry. And when Taforalt was projected, Oub02 formed her own cluster where Taforalt and once again IAM share most but not all of their ancestry. IAM have been previously shown to be closely related to Taforalt and almost indistinguishable in ADMIXTURE^8,15,16⁠. In this instance, when Taforalt is included in the initial run, IAM shows an extremely similar composition to Oub02 with the exception of constantly showing elevated levels of West African-like ancestry. What’s also interesting is the fact that the West African related component is always almost completely absorbed by the newly formed Iberomaurasian components, but in all instances Taforalt shows evidence of having more than Oub02.

At K6 Bayira, labeled “Ethiopia 4500kya” forms his own component with the East African Hadza. His ancestry is pervasive in Eastern African populations in noticeable to small amounts but peaks in the ancestry composition of the Aari of Ethiopia at up to 24%. Taforalt ancestry (founded in the initial runs which included them, simulation A), seems as widespread in non-North African populations as Oub02’s Iberomaurasian cluster, but universally in smaller amounts. Also it seems that Oub02 has less of Taforalt’s Iberomaurasian component than vice versa. These patterns are probably testament to the latter lacking heterogeneity.

At K7 the data really begins to diverge. A seemingly Arabian/Northeast African component which peaks in an individual from Chad forms its own cluster in the chart which included Taforalt in the initial run. And a Somalian/Cushitic cluster forms itself in simulation B. In both instances, the Iberomaurasian component was reduced in almost every non-North African, indicating overlapping variants in the three components. Moreover, the Cushitic component appears to be more relegated to East African populations, reaching high estimates in the horn of Africa and near negligible amounts in some North Africans and Middle eastern populations. However, the North East African/Arabian component (in red), reaches as far west as Chad and maintains moderate estimates in the Near East and North Africa. At K8 a component which peaks in the Mbuti, a population referred to as Pygmies of Central Africa forms in simulation B. For simulation A, The previously established component in red splits into a Chadic component (red) and a North East African/Arab component (aquablue), peaking in the Maba of Chad and the Batahin of East Sudan respectively. Notably, at K9, not only does simulation A have the Mbuti forming their own cluster, the earlier established (since K4) West African component, splits into a yellow component that follows mostly Bantu speakers and a lime green component which peaks in Sahelian/Westernmost Africans. As a result of Chadic populations being simulated to have more of the latter component, the previous NorthAfrican/Arabian component returns. Though at K9 it peaks in the horn of Africa. Simulation B simply has Bayira forming his own isolated component.

K10 has simulation B finally a Northeast African component, however to a significantly lower amplitude. The component is highly localized in Sudan and reach minor frequencies in further populations. Bayira, no longer has his own component and the Eurasian component is split between a South Asian component (beige) and a West Eurasian component (lilac). On the other hand in simulation A there’s a Northwest African component which peaks in the Algerian samples and outside of North Africa it is the highest the Fulani of Cameroon. With the exception of a dedicated Arabian component, K’s 11-12 runs are essentially a mix up of previously unveiled components and splits previously seen. See appended figures below.

As it related to the Iberomaurasian clusters, there are a few key takeaways.

1. IAM tends to highly resemble the Iberomaurasian whom doesn’t form their own component.
2. There is no non-Iberomaurasian Northwest African component when Oub02 forms a component nesting Taforalt and IAM.
3. Reduction in Iberomaurasian related ancestry in populations with suggested north African ancestry gives way to presence of a Northeast African related component.
4. Oub02’s component has higher presence in contemporary North Africans. Coupling that with the fact that about 50% of Oub02’s ancestral composition can belong to a putative north African component, It can be suggested that when accounting for heterogeneity, these ancient North Africans might share more variation with modern North Africans than previously thought. That, and or there is detectable substructure among Epipaleolithic and even early Neolithic North Africans.

Any Differences With Formal Stats?

To get a better overview of what might be the case specifically, I chose to do more analysis in the form of f₄ stats. First it seemed appropriate to look into the amount of deep ancestry the Epipaleolithic and Early Neolithic sample(s) might have. A quick test being f₄(related Eurasian, test; chimp, African). For the Eurasian samples I chose Epipaleolithic samples: Villabruna¹⁷⁠, Loschbour, Natufian, Ust Ishim man, and Pınarbaşı¹⁸⁠. For the African samples I used a sample representing each of the major components which lack Eurasian correspondence in the Clustering analysis above. This includes the Yoruba of Nigeria, Bayira (and his psuedo-haploid variants for comparison), the Nuer of South sudan, the Maba of Chad, an ancient South African huntergatherer found at Faraoskop rockshelter¹⁹⁠, the Mbuti pygmies of the Congo and lastly the Laka of Chad for the simple fact that they looked to be similar to the African portion of Oub02’s ancestry at K4.

^{Weighted Z-scores as a result of f₄(related Eurasian, test; chimp, African). Population names above chart represents “related Eurasian.” Oued Beth EPip denotes psuedohaploid genotypes and Oued Beth Epip are the diploid genotypes.}

As expected, all runs produced a negative Z score signifying that these North Africans have more deep and or African ancestry than the Eurasians tested. Furthermore, western Africans once again seem to have the most positive score. Remarkably, the Laka appears to be the best fit overall for the African related ancestry in these samples. However, the Taforalt samples shows the most variance sometimes suggesting the Yoruba or the Maba to have the best fit. IAM and Oub02 seems to show more similarities with each other than either of the two with Taforalt, which can suggest that IAM could have been a more direct descendant of Oub02’s population. Her diploid and psuedo-haploid profile shows minor differences in which the latter is more shifted toward African populations overall. However, the pattern in relatedness remains somewhat identical. Bayira/Mota on the other hand might have a greater discrepancies between samples. This is indicated by the high variance in relatedness to these North African samples.

With the minor variance in mind I proceeded to look for more differences with more simple f4 stats. This time targeting a handful of various relevant populations from mostly Africa, the Near East and Europe. With the problem of f₄(test, reference; Oub02/Taforalt, Chimp) I can use the relationship of the test population and the reference population to examine how many unique alleles the test population shares with either Taforalt or Oub02. And then with both problems ran, I can use arithmetic to see the difference in possibly unique alleles shared. f₄((test, reference; Taforalt, Chimp) – (test, reference; Oub02, Chimp)), where a more negative value will result in relatedness to Oub02 and a more positive value will be relatedness to Taforalt. I explored these differences with four outgroups, Mbuti pygmies, Ballito Bay²⁰⁠ (southern African huntergatheres similar to the individuals of Faroaskop Rockshelter), A Chinese Paleolithic sample referred to as Tian Yuan Man²¹⁠ (田园人) and Ust Ishim. The two African samples are often used as outgroups due to having more basal ancestry^12,22⁠ and the two Eurasian samples represent Out of Africa populations not known to have contributed any geneflow to either the test samples or the North African samples.

^{Vertical line marks the midpoint in relatedness at 0 where as a positive score shows more relatedness to Taforalt and a negative Oub02}

Interestingly, when diploid samples are used as reference populations, the test samples are shifted more towards Oub02 though their sample selection might attribute that to coincidence. Some consistent outcomes to note is that IAM is always heavily shifted towards Oub02 which is a characteristic noticeable in the ADMIXTURE runs. Also, African populations with less Eurasian ancestry, Pastoral Neolithic populations of East Africa, and Near Eastern populations such as the Pre-pottery Neolithic samples of Jordan are relatively more shifted towards Taforalt.

To put more of a microscope on the data available I ran the same equations using Modern and known Ancient North African samples:

• f₄(test NA<North Afican modern/ancient>, Mbuti; Oub02/Taforalt, Chimp)
• f₄((test NA, Mbuti; Taforalt, Chimp) – (test NA, Mbuti; Oub02, Chimp))

*Kerkouane represents the “Saharan” individual, R11759, from this study.²⁷

Seeing that IAM consistently produces high positive f₄ stats makes it evident that they, Oub02 and Taforalt share more Alleles with each other than with any other population known. However, all of them consistently show evidence of being more related to Oub02 than Taforalt. Interestingly, the (likely to be) Skhirat²⁶ necropolis samples who were partially descendant from populations east, whether directly from the Near east or Egypt²⁴ aren’t significantly more related to Taforalt. Rather, a single sample (though that individual has relatively low coverage) is very much more shifted towards Oub02. Earlier Neolithic samples of Morocco, Kehf el baroud⁶ and Kehf taht el ghar²⁴ are more related to Taforalt.

With modern populations, it is evident that with the exception of the Berbers of Chenini, Tataouine, more local populations share more alleles with Iberomaurasians. That goes for the Berber and Arab populations of Morocco. Populations of North East Africa are unsurprisingly relatively further from the Iberomaurasians but something interesting happens when the two problems are compared.

Takeaways

It seems as if even though we have more available aDNA from Africa, somethings just remain poorly elucidated. With a high coverage sample from ancient Morocco to directly compare we can observe a few discrepancies with what we believe. The same can likely be said about the differences when looking at both Mota/Bayira and Shum Laka who has an individual with diploid genome calls, “Cameroon SMA.” At a quick glance it appears as though African populations with psuedo-haploid calls are more shifted towards more deeply diverged groups such as BallitoBay and possibly the Mbuti Pygmies.

Like everywhere else in the world, ancient Morocco is no different in that people who occupied the region weren’t all clones of one another. However, some significant differences can be seen between Oub02 and Taforalt. Whether it’s due to substructure or Taforalt being pseudo-haploid hasn’t been fully resolved here. But there are some interesting results to take into consideration. One thing which I can likely attribute to the difference in ploidy is the ancestral clusters produced in ADMIXTURE. For one, the heterogeneity seen in modern day Saharan and coastal North Africans might be longstanding in the region. Evidence of this is how Oub02 consistently shows more overlap with other Africans when either forming their own component or being split between an Iberomaurasian and a generic North African component.

A complicated relationship with Eastern and West African populations

With the possible noise of pseudo-haploid groups considered, we can still identify some solid differences between the samples of the Epipaleolithic. The samples at Ifri n’Amr ou Moussa were heavily related to the Epipaleolithic sample of Oued Beth basin. It makes sense once we consider the proximity, being the caves at Ifri n’Amr ou Moussa are located within the basin. It also seems that Taforalt could be a bit more related to eastern populations and modern day West Africans. This trend can be seen when looking at their makeup at K4 and their deep ancestry often showing evidence of sharing more alleles with Yoruba and Maba. Their formal stats likely influenced by the noise of differences in ploidy show a pull towards Shum Laka, Kenya Pastoral Neolithic, Natufians and pre-pottery Neolithic samples of the Near east and modern North East African populations. It can also be the case, hinted via admixture, that the Epipaleolithic samples both have ancestry from a local Northwest African and Northeast African founding populations. In which case Taforalt might have slightly more eastern ancestry.

Tools n Methods

Sequencing information of the newly reported samples were found here. Admixture program can be downloaded here¹⁴⁠. Psuedo-haploid genotypes were called using PileupCaller and Diploid genotypes were called using a GATK/Beagle pipeline as seen in this study¹⁰. Admixture was ran on 3 datasets put through extensive preliminary QC procedure. I used plink²³⁠ to remove related individuals, alleles with minor allele frequency of 0.01, and pruned alleles based on linkage disequalibrium with a “indep-pairwise” using a window size of 200, a sliding window of 20 and a correlation threshold of 0.25. The three datasets had roughly 200K snps. Genotyping rates for pre-projected datasets were about 0.98 and the projected dataset ~0.90. For each K of every run in ADMIXTURE a minimum of 5 iterations were captured and the lowest cross validations of each were presented here. In the rare case when a run with a low CV error yields results which offer no resolution (like when single non ancient individual forms a solitary cluster) an exception was made and the run with the next lowest CV was used. Formal stats were calculated using Admixtools on a dataset sharing 1 million SNPs of the 1240K dataset.

Content and data contained in this blogpost was in reference to a study which since been released after this webpage has been published. A major change should be considered when reading the following. OUB02 designation is that of the Epipaleothic culture, however the archeological site and date has been elucidated by the full article. She was discovered in the cave of Ifri Ouberrid in the village of Aïn Elleuh. Furthermore, though the AGRIWESTMED team were behind excavations of some of the samples in the article, OUB02 wasn’t one of them. Click here to read the full article.

Don’t Be Afraid to Comment, Like and or Share!

References

1. Van De Loosdrecht, M. et al. Pleistocene north african genomes link near eastern and sub-saharan african human populations. Science (80-. ). 360, 548–552 (2018).

2. Lazaridis, I., Nadel, D., Rollefson, G., Merrett, D. C. & Rohland, N. The genetic structure of the world ’ s first farmers. Supplementary Information. (2016).

3. KEFI, R., STEVANOVITCH, A., BOUZAID, E. & BERAUD-COLOMB, E. Diversité mitochondriale de la population de taforalt (12.000 ans bp – Maroc) : Une approche génétique à l’étude du peuplement de l’Afriquedu nord. Anthropol. 43, 1–11 (2005).

4. Jones, W. O., Oliver, R., Gray, R. & Flint, J. E. The Cambridge History of Africa, Volume 3. Afr. Econ. Hist. i, 225 (1979).

5. Barton, R. N. E. et al. Origins of the iberomaurusian in NW Africa: New AMS radiocarbon dating of the middle and later stone age deposits at taforalt cave, Morocco. J. Hum. Evol. 65, 266–281 (2013).

6. Fregel, R. et al. Neolithization of North Africa involved the migration of people from both the Levant and Europe. bioRxiv 191569 (2017).

7. Pagani, L. & Crevecoeur, I. What is Africa? A human perspective. Mod. Hum. Orig. Dispersal 15–24 (2019).

8. Fregel, R. et al. Ancient genomes from North Africa evidence prehistoric migrations to the Maghreb from both the Levant and Europe. Proc. Natl. Acad. Sci. U. S. A. 115, 6774–6779 (2018).

9. Martínez-Sánchez, R. M., Vera-Rodríguez, J. C., Pérez-Jordà, G., Peña-Chocarro, L. & Bokbot, Y. The beginning of the Neolithic in northwestern Morocco. Quat. Int. 470, 485–496 (2018).

10. Gallego Llorente, M. et al. Ancient Ethiopian genome reveals extensive Eurasian admixture throughout the African continent. Science (80-. ). 350, 820–822 (2015).

11. Lipson, M. et al. Ancient DNA and deep population structure in sub-Saharan African foragers. Nature 603, 290–296 (2022).

12. Lipson, M. et al. Ancient West African foragers in the context of African population history. Nature 577, 665–670 (2020).

13. Lazaridis, I. et al. Paleolithic DNA from the Caucasus reveals core of West Eurasian ancestry. bioRxiv (2018) doi:10.1101/423079.

14. Alexander, D. H., Novembre, J. & Lange, K. Fast model-based estimation of ancestry in unrelated individuals. Genome Res. 19, 1655–1664 (2009).

15. Serra-Vidal, G. et al. Heterogeneity in Palaeolithic Population Continuity and Neolithic Expansion in North Africa. Curr. Biol. 29, 3953-3959.e4 (2019).

16. Olalde, I. et al. The genomic history of the Iberian Peninsula over the past 8000 years. Science (80-. ). 363, 1230–1234 (2019).

17. Vercellotti, G., Alclati, G., Richards, M. P. & Formlcola, V. The Late Upper Paleolithic skeleton Villabruna 1 (Italy): A source of data on biology and behavior of a 14.000 year-old hunter. J. Anthropol. Sci. 86, 143–163 (2008).

18. Feldman, M. et al. Late Pleistocene human genome suggests a local origin for the first farmers of central Anatolia. Nat. Commun. 10, 1–10 (2019).

19. Skoglund, P. et al. Reconstructing Prehistoric African Population Structure. Cell 171, 59-71.e21 (2017).

20. Schlebusch, C. M. et al. Ancient genomes from southern Africa pushes modern human divergence beyond 260,000 years ago. bioRxiv (2017) doi:10.1101/145409.

21. Yang, M. A. et al. 40,000-Year-Old Individual from Asia Provides Insight into Early Population Structure in Eurasia. Curr. Biol. 27, 3202-3208.e9 (2017).

22. Lazaridis, I. et al. Ancient human genomes suggest three ancestral populations for present-day Europeans. Nature 513, 409–413 (2014).

23. Chang, C. C. et al. Second-generation PLINK: Rising to the challenge of larger and richer datasets. Gigascience 4, (2015).

24. Simões, L. G.. Uncovering the Past through ancient DNA : The Fate and Legacy of the last hunter-gatherers in Western Europe and Northwestern Africa (PhD dissertation, Acta Universitatis Upsaliensis) (2023).

25. Schwarzhans, W. Geology and stratigraphy of the Neogene section along the Oued Beth between Dar bel Hamri and El Kansera (Rharb Basin, northwestern Morocco) and its otolith-based fish fauna: a faunal inventory for the Early Pliocene remigration into the Mediterranean. Swiss J Palaeontol 142, 4 (2023)

26. Lacombe, J.P. & Daugas, J. et al. La nécropole Néolithique de Skhirat (Maroc) : caractérisation de la population inhumée et organisation spatiale. Bulletins et mémoires de la Société d’anthropologie de Paris. 26. (2013) 10.1007/s13219-013-0082-x.

27. Moots, H. M. et al. A Genetic History of Continuity and Mobility in the Iron Age Central Mediterranean. 1–18 (2022).

28. Simões, L.G., Günther, T., Martínez-Sánchez, R.M. et al. Northwest African Neolithic initiated by migrants from Iberia and Levant. Nature 618, 550–556 (2023). https://doi.org/10.1038/s41586-023-06166-6