Ecology

These authors contributed equally: Marc Palahí, Ruben ValbuenaEuropean Forest Institute, Joensuu, FinlandMarc Palahí, Lauri Hetemäki, Pieter Johannes Verkerk & Minna KorhonenSchool of Natural Sciences, Bangor University, Bangor, UKRubén ValbuenaEcosystem Dynamics and Forest Management Group, Technical University of Munich, Munich, GermanyCornelius Senf & Rupert SeidlSchool of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham, UKNezha Acil, Thomas A. M. Pugh & Jonathan SadlerBirmingham Institute of Forest Research, University of Birmingham, Birmingham, UKNezha Acil, Thomas A. M. Pugh & Jonathan SadlerDepartment of Physical Geography and Ecosystem Science, Lund University, Lund, SwedenThomas A. M. PughDepartment of Geographical Sciences, University of Maryland, College Park, MD, USAPeter PotapovInstitut Européen de la Forêt Cultivée, Cestas, FranceBarry GardinerDipartimento di Scienze e Tecnologie Agrarie, Alimentari, Ambientali e Forestali, Università degli Studi di Firenze, Florence, ItalyGherardo Chirici & Saverio FranciniDipartimento per l’Innovazione dei Sistemi Biologici, Agroalimentari e Forestali, Università degli Studi della Tuscia, Viterbo, ItalySaverio FranciniFaculty of Forestry and Wood Sciences, Czech University of Life Sciences Prague, Prague, Czech RepublicTomáš Hlásny & Róbert MarušákWageningen Environmental Research, Wageningen University and Research, Wageningen, The NetherlandsBas Jan Willem Lerink & Gert-Jan NabuursDepartment of Forest Resource Management, Swedish University of Agricultural Sciences (SLU), Umeå, SwedenHåkan Olsson & Jonas FridmanJoint Research Unit CTFC – AGROTECNIO, Solsona, SpainJosé Ramón González OlabarriaDepartment of Agricultural, Forest and Food Sciences, University of Turin, Grugliasco, ItalyDavide AscoliNatural Resources Institute Finland, Joensuu, FinlandAntti AsikainenAlbert-Ludwigs-University of Freiburg, Freiburg, GermanyJürgen Bauhus & Marc HanewinkelDepartment of Space, Earth and Environment, Chalmers University of Technology, Gothenburg, SwedenGöran BerndesLatvian State Forest Research Institute Silava, Salaspils, LatviaJanis DonisINRAE, University of Bordeaux, BIOGECO, Cestas, FranceHervé JactelEuropean Forest Institute, Bonn, GermanyMarcus LindnerUniversity of Molise, Campobasso, ItalyMarco MarchettiForest Ecology and Forest Management Group, Wageningen University and Research, Wageningen, The NetherlandsDouglas Sheil & Gert-Jan NabuursCentro de Estudos Florestais, Instituto Superior de Agronomia, Universidade de Lisboa, Lisbon, PortugalMargarida ToméForest Science and Technology Centre of Catalonia, CTFC, Solsona, SpainAntoni TrasobaresM.P. and G.-J.N. conceived and initiated the study. R.V., C.S., N.A., T.A.M.P., G.C., S.F., T.H., B.J.W.L. and D.A. ran different parts of the analyses and demonstrations. M.P., R.V., G.-J.N., C.S., T.A.M.P., J.S., R.S., B.G. and L.H. drafted the initial version of the manuscript. P.P. provided first-hand experience of the algorithms involved in the production of GFC data. All authors offered insights from their own national statistics and local knowledge, which focused the analyses and the argumentation, and contributed critically to the interpretation of the results, revising and approving the final version of the manuscript. More

Identification of ultra-small cells associated with blooms of anoxygenic phototrophic gammaproteobacteriaThe Salada de Chiprana (NE Spain) is the only permanent athalassic hypersaline lake in Western Europe. It harbors thick, conspicuous microbial mats covering its bottom (Fig. 1a, b) and exhibits periodic stratification during which the deepest part of the water column becomes anoxic and sulfide-rich, favoring the massive development of sulfide-dependent anoxygenic photosynthetic bacteria16. We collected microbial mat fragments that were maintained in culture in the laboratory. After several weeks, we observed a bloom of anoxygenic photosynthetic bacteria containing numerous intracellular sulfur granules (Fig. 1c). Many of these bacterial cells showed one or several much smaller and darker non-flagellated cells attached to their surface (Fig. 1d, e). The infected photosynthetic cells were highly mobile, swimming at high speed with frequent changes of direction (Supplementary Movie 1), in contrast with the non-infected cells that displayed slower (approximately half speed) and more straight swimming. Sometimes, two or more photosynthetic cells were connected through relatively short filaments formed by stacked epibiont cells (Fig. 1f). Although the photosynthetic cells carrying these epibionts were often actively swimming, in some cases the epibionts were associated to empty ghost cells where only the photosynthesis-derived sulfur granules persisted. Closer scrutiny of the epibionts revealed that they actually consisted of piles of up to 10 very flattened cells of 550 ± 50 nm diameter and 220 ± 20 nm height (n = 100). These characteristics (size, morphology, and specific attachment to sulfide-dependent anoxygenic photosynthetic bacteria) perfectly matched the morphological description of the genus Vampirococcus observed over forty years ago in sulfidic freshwater lakes15.Fig. 1: Sampling site and microscopy observation of Vampirococcus cells.a General view of the microbial mat covering the shore of the Salada de Chiprana lake. b Closer view of a microbial mat section. c Natural population of blooming sulfide-dependent anoxygenic photosynthetic bacteria in waters of microbial mat containers after several weeks of growth in the laboratory; note the conspicuous refringent intracellular sulfur inclusions. d–f Closer microscopy view of anoxygenic photosynthetic bacteria infected by epibiotic Vampirococcus cells and few-cell filaments (indicated by yellow arrows). g Scanning electron microscopy image of a host cell infected by two stacking Vampirococcus cells (yellow arrow). h Transmission electron microscopy (TEM) image of a thin section of a host cell infected by Vampirococcus (yellow arrow). i Closer TEM view of a thin section of Vampirococcus cells, notice the fibrous rugose cell surface and the large space separating contiguous cells. Scale bars: 5 cm (b), 5 µm (c), 1 µm (d–h), 0.5 µm (i).Full size imageSince the first Vampirococcus description included transmission electron microscopy (TEM) images, to further ascertain this identification we examined our Chiprana Lake samples under TEM and scanning electron microscopy (SEM). SEM images confirmed the peculiar structure of the epibionts, with multiple contiguous cells separated by deep grooves (Fig. 1g). Thin sections observed under TEM confirmed that the cells were actually separated by a space of ~20–50 nm filled by a fibrous material (Fig. 1h, i). The space between epibiont and host cells was larger (~100 nm) and also filled by dense fibrous material (Fig. 1h). The sections also showed that, in contrast with the typical Gram-negative double membrane structure of the host, the epibiont cells had a single membrane surrounded by a thick layer of fibrous material that conferred a rugose aspect to the cells (Fig. 1i). In sharp contrast with the often highly vacuolated cytoplasm of the host, the epibiont cells showed a dense, homogeneous content. These observations were also in agreement with those published for Vampirococcus, reinforcing our conclusion that the epibionts we observed belonged to this genus, although most likely to a different species, as the first described Vampirococcus occurred in a non-hypersaline lake15.Using a micromanipulator coupled to an inverted microscope, we collected cells of the anoxygenic photosynthetic bacterium carrying Vampirococcus attached to their surface (Supplementary Fig. 1) and proceeded to amplify, clone, and sequence their 16 S rRNA genes. We were able to obtain sequences for both the epibiont and the host for ten infected cells and, in all cases, we retrieved the same two sequences. The host was found to be a Halochromatium-like gammaproteobacterium (Supplementary Fig. 2). Phylogenetic analysis of the epibiont sequence showed that it branched within the CPR radiation close to the Absconditabacteria (Supplementary Fig. 3), previously known as candidate phylum SR12. Since all host and epibiont cells we analyzed had identical 16 S rRNA gene sequences, suggesting that they were the result of a clonal bloom, we collected three sets of ca. 10 infected cells and carried out whole genome amplification (WGA) before sequencing (Illumina HiSeq; see Methods). This strategy allowed us to assemble the nearly complete genome sequence of the Vampirococcus epibiont (see below). In contrast with the completeness of this genome, we only obtained a very partial assembly (~15%) of the host genome, probably because of the consumption of the host DNA by the epibiont. To make more robust phylogenetic analyses of Vampirococcus, we retrieved the protein sequence set used by Hug et al. to reconstruct a multi-marker large-scale phylogeny of bacteria2. The new multi-gene maximum likelihood (ML) phylogenetic tree confirmed the affiliation of our Vampirococcus species to the Absconditabacteria with maximum support, and further placed this clade within a larger well-supported group also containing the candidate phyla Gracilibacteria and Peregrinibacteria (Fig. 2a and Supplementary Fig. 4). Therefore, our epibiotic bacterium represents the first characterized member of this large CPR clade and provides a phylogenetic identity for the predatory bacterial genus Vampirococcus described several decades ago. We propose to call this new species Candidatus Vampirococcus lugosii (see Taxonomic appendix).Fig. 2: Phylogeny and global gene content of the Vampirococcus genome.a Maximum likelihood phylogenetic tree of bacteria based on a concatenated dataset of 16 ribosomal proteins showing the position of Vampirococcus lugosii close to the Absconditabacteria (for the complete tree, see Supplementary Fig. 4). Histograms on the right show the proportion of genes retained in each species from the ancestral pool inferred for the last common ancestor of Absconditabacteria, Gracilibacteria and Peregrinibacteria. b Percentage of Vampirococcus genes belonging to the different Clusters of Orthologous Groups (COG) categories. c Genes shared by Vampirococcus and the three Absconditabacteria genomes shown in the phylogenetic tree. COG categories are: Energy production and conversion [C]; Cell cycle control, cell division, chromosome partitioning [D]; Amino acid transport and metabolism [E]; Nucleotide transport and metabolism [F]; Carbohydrate transport and metabolism [G]; Coenzyme transport and metabolism [H]; Lipid transport and metabolism [I]; Translation, ribosomal structure and biogenesis [J]; Transcription [K]; Replication, recombination and repair [L]; Cell wall/membrane/envelope biogenesis [M]; Secretion, motility and chemotaxis [N]; Posttranslational modification, protein turnover, chaperones [O]; Inorganic ion transport and metabolism [P]; General function prediction only [R]; Function unknown [S]; Intracellular trafficking, secretion, and vesicular transport [U]; Defense mechanisms [V]; Mobilome: prophages, transposons [X]; Secondary metabolites biosynthesis, transport and catabolism [Q]. Source data are provided as a Source Data file.Full size imageGenomic evidence of adaptation to predatory lifestyleWe sequenced DNA from three WGA experiments corresponding each to ~10 Halochromatium-Vampirococcus consortia. Many of the resulting (57.2 Mb) raw sequences exhibited similarity to those of available Absconditabacteria/SR1 metagenome-assembled genomes (MAGs) and, as expected, some also to Gammaproteobacteria (host-derived sequences) as well as a small proportion of potential contaminants probably present in the original sample (Bacillus- and fungi-like sequences). To bin the Vampirococcus sequences out of this mini-metagenome, we applied tetranucleotide frequency analysis on the whole sequence dataset using emergent self-organizing maps (ESOM)6. One of the ESOM sequence bins was enriched in Absconditabacteria/SR1-like sequences and corresponded to the Vampirococcus sequences, which we extracted and assembled independently. This approach yielded an assembly of 1,310,663 bp. We evaluated its completeness by searching i) a dataset of 40 universally distributed single-copy genes17 and ii) a dataset of 43 single-copy genes widespread in CPR bacteria8. We found all them as single-copy genes in the Vampirococcus genome, with the exception of two signal recognition particle subunits from the first dataset which are absent in many other CPR bacteria18. These results supported that the Vampirococcus genome assembly was complete and did not contain multiple strains or other sources of contamination. Manually curated annotation predicted 1151 protein-coding genes, a single rRNA gene operon, and 38 tRNA coding genes. As already found in other Absconditabacteria/SR1 genomes19, the genetic code of Vampirococcus is modified, with the stop codon UGA reassigned as an additional glycine codon.A very large proportion of the predicted proteins (48.9%) had no similarity to any COG class and lacked any conserved domain allowing their functional annotation (Fig. 2b). Thus, as for other CPR bacteria, a significant part of their cellular functions remains inaccessible. A comparison with three other Absconditabacteria genomes revealed a very small set of only 390 genes conserved in all them (Fig. 2c), suggesting a highly dynamic evolution of gene content in these species. Comparison with more distantly related CPR groups (Gracilibacteria and Peregrinibacteria) showed that gene loss has been a dominant trend in all these organisms, which have lost 30–50% of the 1124 genes inferred to have existed in their last common ancestor (Fig. 2a). Nevertheless, this loss of ancestral genes was accompanied by the acquisition of new ones by different mechanisms, including horizontal gene transfer (HGT). In the case of Vampirococcus, we detected, by phylogenetic analysis of all individual genes that had homologs in other organisms, the acquisition of 126 genes by HGT from various donors (Supplementary Data 1).The set of genes that could be annotated provided interesting clues about the biology and lifestyle of Vampirococcus. The most striking feature was its oversimplified energy and carbon metabolism map (Fig. 3). ATP generation in this CPR species appeared to depend entirely on substrate-level phosphorylation carried out by the phosphoenolpyruvate kinase (EC 2.7.1.40). In fact, Vampirococcus only possesses incomplete glycolysis, which starts with 3-phosphoglycerate as first substrate. This molecule is the major product of the enzyme RuBisCO and, therefore, most likely highly available to Vampirococcus from its photosynthetic host. Comparison with nearly complete MAG sequences available for other Absconditabacteria/SR1 showed that Vampirococcus has the most specialized carbon metabolism, with 3-phosphoglycerate as the only exploitable substrate, whereas the other species have a few additional enzymes that allow them to use other substrates (such as ribulose-1,5 P and acetyl-CoA) as energy and reducing power (NADH) sources (Supplementary Fig. 5). This metabolic diversification probably reflects their adaptation to other types of hosts where these substrates are abundant. Vampirococcus also lacks all the enzymes involved in some Absconditabacteria/SR1 in the 3-phosphoglycerate-synthesizing AMP salvage pathway20, including the characteristic archaeal-like type II/III RuBisCO21.Fig. 3: Metabolic and cell features inferred from the genes encoded in the Vampirococcus genome.The diagram shows the host cell surface (bottom) with two stacking Vampirococcus cells attached to its surface (as in Fig. 1h).Full size imageThe genomes of Absconditabacteria/SR1 and Vampirococcus encode several electron carrier proteins (e.g., ferredoxin, cytochrome b5, several Fe-S cluster proteins) and a membrane F1FO-type ATP synthase. However, they apparently lack any standard electron transport chain and, therefore, they seem to be non-respiring19,20,22. The electron carrier proteins may be related with the oxidative stress response and/or the reoxidation of reduced ferredoxin or NADH20. In the absence of any obvious mechanism to generate proton motive force (PMF), the presence of the membrane ATP synthase is also intriguing. It has been speculated either that CPR bacteria might tightly adhere to their hosts and scavenge protons from them or that the membrane ATP synthase might work in the opposite direction as an ATPase, consuming ATP generated by substrate level phosphorylation to extrude protons and drive antiporters9. However, in the case of Vampirococcus the direct transport of protons from the host is unlikely since, as observed in the TEM sections (Fig. 1h), it seems that there is no direct contact with the host cell membrane. In fact, parasite and host cell membranes are separated by a relatively large space of ~100 nm, which would be largely conducive to proton diffusion and inefficient transfer between cells. Alternatively, since the Chiprana lake has high Na+ concentration (1.6 g l−1), it might be possible that the ATP synthase uses Na+ instead of protons. However, the Na+-binding domain of the subunit c of typical Na+-dependent ATP synthases exhibited several differences with that of Vampirococcus (Supplementary Fig. 6). Similar differences have been considered indicative of the use of protons instead of Na+ in other organisms23. Although the proton/cation antiporters (e.g., for Na+, K+, or Ca2+) encoded by Vampirococcus and the other Absconditabacteria/SR1 may serve to produce some PMF, it is improbable that this mechanism represents a major energy transducing system as cells would accumulate cations and disrupt their ionic balance; these antiporters are most likely involved in cation homeostasis.These observations prompted us to investigate other ways that these cells might use to generate PMF usable by their ATP synthase. We found a protein (Vamp_33_45) with an atypical tripartite domain structure. The N-terminal region, containing 8 transmembrane helices, showed similarity with several flavocytochromes capable of moving electrons and/or protons across the plasma membrane (e.g., 24). The central part of the protein was a rubredoxin-like nonheme iron-binding domain likely able to transport electrons. Finally, the C-terminal region, containing an NAD-binding motif, was similar to ferredoxin reductases involved in electron transfer25. This unusual Vampirococcus 3-domain protein is well conserved in the other Absconditabacteria/SR1 genomes sequenced so far, suggesting it plays an important function in this CPR phylum. Its architecture suggests that it can transport electrons and/or protons across the membrane using ferredoxin as electron donor and makes it a strong candidate to participate in a putative new PMF-generating system. Alternatively, this protein could play a similar role to that of some oxidoreductases in the strict anaerobic archaeon Thermococcus onnurineus, including a thioredoxin reductase, which couple reactive oxygen species detoxification with NAD(P) + regeneration from NAD(P)H to maintain the intracellular redox balance and enhance O2-mediated growth despite the absence of heme-based or cytochrome-type proteins26.Although our Vampirococcus genome sequence appears to be complete, genes encoding enzymes involved in the biosynthesis of essential cell building blocks such as amino acids, nucleotides and nucleosides, cofactors, vitamins, and lipids are almost completely absent (Fig. 2b). Therefore, the classical bacterial metabolic pathways for their synthesis27 do not operate in Vampirococcus. Such simplified metabolic potential, comparable to that of intracellular parasitic bacteria such as Mycoplasma28, implies that Vampirococcus must acquire these molecules from an external source and supports the predatory nature of the interaction with its photosynthetic host. An intriguing aspect of this interaction concerns the transfer of substrates from the host to Vampirococcus, especially considering that, despite examination of several serial ultrathin sections, the cell membranes of these two partners do not appear to be in direct contact (Fig. 1h). Vampirococcus encodes several virulence factors, including divergent forms of hemolysin and hemolysin translocator (Vamp_11_169 and Vamp_9_166, respectively), a phage holin (Vamp_5_129), and a membrane-bound lytic murein transglycosylase (Vamp_144_2). These proteins are likely involved in the host cell wall and membrane disruption leading to cell content release. Hemolysin has also been found in Saccharibacteria (formerly candidate phylum TM7), the only CPR phylum for which an epibiotic parasitic lifestyle has been demonstrated so far13,14. Recent coupled lipidomic-metagenomic analyses have shown that CPR bacteria that lack complete lipid biosynthesis are able to recycle membrane lipids from other bacteria29. In Vampirococcus, also devoid of phospholipid synthesis, a phospholipase gene (Vamp_34_196) predicted to be secreted and that has homologs involved in host phospholipid degradation in several parasitic bacteria30, may not only help disrupting the host membrane but also to generate a local source of host phospholipids that it can use to build its own cell membrane. Two Vampirococcus peptidoglycan hydrolases (Vamp_68-56_103 and Vamp_145_30), also predicted to be secreted, most probably contribute to degrade the host cell wall. The Vampirococcus genome also encodes two murein DD-endopeptidases (Vamp_311_38 and Vamp_41_33). As in other predatory bacteria, such as Bdellovibrio, one probably acts to degrade the prey cell wall whereas the other is involved in self-wall remodeling31. Despite their high sequence divergence, we could align both Vampirococcus sequences with those of Bdellovibrio and other bacteria (Supplementary Fig. 7). Both sequences conserved the characteristic active site serine residue of DD endopeptidases and, in contrast with the Bdellovibrio “predatory” enzymes, also the regulatory domain III. The deletion of this regulatory domain has been associated with the capacity of the Bdellovibrio “predatory” DD endopeptidases to act promiscuously on a wide variety of peptidoglycan substrates31. This difference most likely reflects that, whereas Bdellovibrio is able to prey on very diverse bacteria, Vampirococcus is a specialized predator of Chromatiaceae that has evolved specialized enzymes to degrade the wall of its particular prey. The Vampirococcus enzymes could also be aligned with the region where the ankyrin-repeat-containing self-protective regulatory inhibitor Bd3460 binds the Bdellovibrio “predatory” DD endopeptidases32, although only partially for the C-terminal part of Vamp_41_33, like in the self-wall Bdellovibrio enzyme Bd3244 (Supplementary Fig. 7). In that sense, the Vamp_311_38 enzyme seems more similar to the “predatory” Bdellovibrio ones. Interestingly, the “predatory” endopeptidase Bd3459 and the regulatory inhibitor Bd3460 are contiguous in the genomes of Bdellovibrio and other periplasmic predators but not in epibiotic predators32. Vampirococcus confirms this pattern since, although it possesses several ankyrin-repeat-containing proteins, none of them is encoded adjacent to the DD endopeptidase genes.Vampirococcus also possesses a number of genes encoding transporters, most of them involved in the transport of inorganic molecules (Fig. 3). One notable exception is the competence-related integral membrane protein ComEC (Vamp_67_106)33 which, together with ComEA (Vamp_21_186) and type IV pili (see below), probably plays a role in the uptake of host DNA that, once transported into the epibiont, can be degraded by various restriction endonucleases (five genes encoding them are present) and recycled to provide the nucleotides necessary for growth (Supplementary Fig. 8). These proteins are widespread in other CPR bacteria where they may have a similar function34. Vampirococcus also encodes an ABC-type oligopeptide transporter (Vamp_40_40) and a DctA-like C4-dicarboxylate transporter (Vamp_41_97), known to catalyze proton-coupled symport of several Krebs cycle dicarboxylates (succinate, fumarate, malate, and oxaloacetate)35. The first, coupled with the numerous peptidases present in Vampirococcus, most likely is a source of amino acids. By contrast, the role of DctA is unclear since Vampirococcus does not have a Krebs cycle.In sharp contrast with its simplified central metabolism, Vampirococcus possesses genes related to the construction of an elaborate cell surface, which seems to be a common theme in many CPR bacteria9,12. They include genes involved in peptidoglycan synthesis, several glycosyltransferases, a Sec secretion system, and a rich repertoire of type IV pilus proteins. The retractable type IV pili are presumably involved in the tight attachment of Vampirococcus to its host and in DNA uptake in cooperation with the ComEC protein. Other proteins probably play a role in the specific recognition and fixation to the host, including several very large proteins. In fact, the Vampirococcus membrane proteome is enriched in giant proteins. The ten longest predicted proteins (between 1392 and 4163 aa, see Supplementary Table 1) are inferred to have a membrane localization and are probably responsible of the conspicuous fibrous aspect of its cell surface (Fig. 1i). Most of these proteins possess domains known to be involved in the interaction with other molecules, including protein-protein (WD40, TRP, and PKD domains) and protein–lipid (saposin domain) interactions and cell adhesion (DUF11, integrin, and fibronectin domains). Two other large membrane proteins (Vamp_6_203, 2368 aa, and Vamp_19_245, 1895 aa) may play a defensive role as they contain alpha-2-macroglobulin protease-inhibiting domains that can protect against proteases released by the host. Several other smaller proteins complete the membrane proteome of Vampirococcus, some of them also likely involved in recognition and attachment to the host thanks to a variety of protein domains, such as VWA (Vamp_41_85) and flotillin (Vamp_11_100). We did not detect genes coding for flagellar components, confirming the absence of flagella observed under the microscope (Fig. 1).New CRISPR-Cas systems and other defense mechanisms in Vampirococcus
Although most CPR phyla are devoid of CRISPR-Cas36, some have been found to contain new systems with original effector enzymes such as CasY37. In contrast with most available Absconditabacteria genomes, Vampirococcus possesses two CRISPR-Cas loci (Fig. 4a and Supplementary Fig. 9). The first is a class II type V system that contains genes coding for Cas1, Cas2, Cas4, and Cpf1 proteins associated to 34 spacer sequences of 26–32 bp. Proteins similar to those of this system are encoded not only in genomes of close relatives of the Absconditabacteria (Gracilibacteria and Peregrinibacteria) but in many other CPR phyla. These sequences form monophyletic groups in phylogenetic analyses (e.g., Cas1, see Fig. 4b), which suggests that this type V system is probably ancestral in these CPR. The second system found in Vampirococcus belongs to the class I type III and contains genes coding for Cas1, Cas2, Csm3, and Cas10/Csm1 proteins associated to a cluster of 20 longer (35–46 bp) spacers. In contrast with the previous CPR-like system, the proteins of this second system did not show strong similarity with any CPR homolog but with sequences from other bacterial phyla, suggesting that they have been acquired by HGT. Phylogenetic analysis confirmed this and supported that Vampirococcus gained this CRISPR-Cas system from different distant bacterial donors (Supplementary Fig. 10). Interestingly, these two CRISPR-Cas systems encode a number of proteins that may represent new effectors. A clear candidate is the large protein Vamp_48_93 (1158 aa), located between Cpf1 and Cas1 in the type V system (Fig. 4a), which contains a DNA polymerase III PolC motif. Very similar sequences can be found in a few other CPR (some Roizmanbacteria, Gracilibacteria, and Portnoybacteria) and in some unrelated bacteria (Supplementary Fig. 11). As in Vampirococcus, the gene coding for this protein is contiguous to genes encoding different Cas proteins in several of these bacteria, including Roizmanbacteria, Omnitrophica, and the deltaproteobacterium Smithella sp. (Supplementary Fig. 11). This gene association, as well as the very distant similarity between this protein and Cpf1 CRISPR-associated proteins of bacterial type V systems, supports that it is a new effector in type V CRISPR-Cas systems. Additional putative new CRISPR-associated proteins likely exist also in the Vampirococcus type III system (Fig. 4a). Three proteins encoded by contiguous genes (Vamp_21_116, Vamp_21_127, and Vamp_21_128) exhibit very distant similarity with type III-A CRISPR-associated Repeat Associated Mysterious Proteins (RAMP) Csm4, Csm5, and Csm6 sequences, respectively, and most probably represent new RAMP subfamilies. To date, Absconditabacteria38 and Saccharibacteria39 are the only CPR phyla for which phages have been identified. Because of its proximity to Absconditabacteria, Vampirococcus is probably infected by similar phages, so that the function of its CRISPR-Cas systems may be related to the protection against these genetic parasites. Nevertheless, we did not find any similarity between the Vampirococcus spacers and known phage sequences, suggesting that it is infected by unknown phages. Alternatively, considering that Vampirococcus -as most likely many other CPR bacteria- seems to obtain nucleotides required for growth by uptaking host DNA, an appealing possibility is that the CRISPR-Cas systems participate in the degradation of the imported host DNA.Fig. 4: CRISPR-Cas systems in Vampirococcus.a Genes in the two systems encoded in the Vampirococcus lugosii genome, elements common to the two systems are highlighted in blue. b Maximum likelihood phylogenetic tree of the Cas1 protein encoded in the class II type V system, numbers at branches indicate bootstrap support.Full size imageAlthough CPR bacteria have been hypothesized to be largely depleted of classical defense mechanisms40, we found that Vampirococcus, in addition to the two CRISP-Cas loci, is endowed with various other protection mechanisms. These include an AbiEii-AbiEi Type IV toxin-antitoxin system, also present in other CPR bacteria, which may offer additional protection against phage infection41 and several restriction-modification systems, with three type I, one type II and one type III restriction enzymes and eight DNA methylases. In addition to a defensive role, these enzymes may also participate in the degradation of the host DNA. As in its sister-groups Absconditabacteria and Gracilibacteria5,19,20,42,43, Vampirococcus has repurposed the UGA stop codon to code for glycine. The primary function of this recoding remains unknown but it has been speculated that it creates a genetic incompatibility, whereby these bacteria would be “evolutionarily isolated” from their environmental neighbors, preventing their potential competitors from acquiring their genomic innovations by HGT19. However, the opposite might be argued as well, since the UGA codon reassignment can protect Vampirococcus from foreign DNA expression upon uptake by leading to aberrant protein synthesis via read-through of the UGA stop with Gly insertion. This can be important for these CPR bacteria because they are not only impacted by phages38 but they most likely depend on host DNA import and degradation to fulfill their nucleotide requirements. In that sense, it is interesting to note that the Vampirococcus ComEA protein likely involved in DNA transport44 is encoded within the class I type III CRISPR-Cas system (Fig. 4a). More

Invertebrate composition effects on primary productionTo evaluate the effects of fiddler crabs, marsh crabs, and mussels on benthic algae and cordgrass production, the dietary sources for fiddler and marsh crabs, respectively27,28, we measured benthic diatom biomass and cordgrass stem density every 4–6 weeks and quantified cordgrass biomass and grazing damage at the conclusion of the experiment in August 2017. Diatom biomass was enhanced in enclosures with mussels and/or marsh crabs relative to enclosures with only fiddler crabs or no invertebrates, and relative to all ambient plots (F36, 200 = 1.5; P = 0.04; Tukey’s HSD, all P  More

1.Elser, J. J. et al. Global analysis of nitrogen and phosphorus limitation of primary producers in freshwater, marine and terrestrial ecosystems. Ecol. Lett. 10, 1135–1142 (2007).Article 

Google Scholar 
2.Mengel, K. & Kirkby, E. A. Principles of Plant Nutrition (Kluwer Academic Publishers, 2001).Book 

Google Scholar 
3.Reich, M., Aghajanzadeh, T. & De Kok, L. J. Physiological basis of plant nutrient use efficiency—Concepts, opportunities and challenges for its improvement. In Nutrient Use Efficiency in Plants: Concepts and Approaches (eds Hawkesford, M. J. et al.) (Springer, 2014).
Google Scholar 
4.Agren, G. I. Ideal nutrient productivities and nutrient proportions in plant growth. Plant Cell Environ. 11, 613–620 (1988).Article 

Google Scholar 
5.Weih, M., Hamner, K. & Pourazari, F. Analyzing plant nutrient uptake and utilization efficiencies: Comparison between crops and approaches. Plant Soil 430, 7–21 (2018).CAS 
Article 

Google Scholar 
6.Sterner, R. W. & Elser, J. J. Ecological stoichiometry: The biology of elements from molecules to the biosphere (2002).7.Reich, P. B. et al. Evidence of a general 2/3-power law of scaling leaf nitrogen to phosphorus among major plant groups and biomes. Proc. R. Soc. B Biol. Sci. 277, 877–883 (2010).CAS 
Article 

Google Scholar 
8.Hutchinson, G. E. Population studies—Animal ecology and demography—Concluding remarks. Cold Spring Harbor. Symp. Quant. Biol. 22, 415–427 (1957).Article 

Google Scholar 
9.Agren, G. I. & Weih, M. Multi-dimensional plant element stoichiometry-looking beyond carbon, nitrogen, and phosphorus. Front. Plant Sci. 11, 23 (2020).Article 

Google Scholar 
10.Niklas, K. J. Plant allometry, leaf nitrogen and phosphorus stoichiometry, and interspecific trends in annual growth rates. Ann. Bot. 97, 155–163 (2006).CAS 
Article 

Google Scholar 
11.Hou, E. et al. Global meta-analysis shows pervasive phosphorus limitation of aboveground plant production in natural terrestrial ecosystems. Nat. Commun. 11, 637 (2020).ADS 
CAS 
Article 

Google Scholar 
12.Ryan, P. R. et al. Early vigour improves phosphate uptake in wheat. J. Exp. Bot. 66, 7089–7100 (2015).CAS 
Article 

Google Scholar 
13.Wiel, CCMvd., Linden, CGvd & Scholten, O. E. Improving phosphorus use efficiency in agriculture: Opportunities for breeding. Euphytica 207, 1–22 (2016).Article 

Google Scholar 
14.Bilal, H. M., Aziz, T., Maqsood, M. A., Farooq, M. & Yan, G. Categorization of wheat genotypes for phosphorus efficiency. PLoS ONE 13, e0205471 (2018).Article 

Google Scholar 
15.Wang, Z. et al. Magnesium fertilization improves crop yield in most production systems: A meta-analysis. Front. Plant Sci. 10, 1727 (2020).Article 

Google Scholar 
16.Hauer-Jakli, M. & Traenkner, M. Critical leaf magnesium thresholds and the impact of magnesium on plant growth and photo-oxidative defense: a systematic review and meta-analysis from 70 years of research. Front. Plant Sci. 10, 766 (2019).Article 

Google Scholar 
17.Chawade, A. et al. A transnational and holistic breeding approach is needed for sustainable wheat production in the Baltic Sea region. Physiol. Plant. 164, 442–451 (2018).CAS 
Article 

Google Scholar 
18.Weih, M., Pourazari, F. & Vico, G. Nutrient stoichiometry in winter wheat: Element concentration pattern reflects developmental stage and weather. Sci. Rep. 6, 35958–35958 (2016).ADS 
CAS 
Article 

Google Scholar 
19.Hamner, K., Weih, M., Eriksson, J. & Kirchmann, H. Influence of nitrogen supply on macro- and micronutrient accumulation during growth of winter wheat. Field Crop Res. 213, 118–129 (2017).Article 

Google Scholar 
20.Jia, X., Liu, P. & Lynch, J. P. Greater lateral root branching density in maize improves phosphorus acquisition from low phosphorus soil. J. Exp. Bot. 69, 4961–4970 (2018).CAS 
Article 

Google Scholar 
21.Kumar, A. et al. Root trait plasticity and plant nutrient acquisition in phosphorus limited soil. J. Plant Nutr. Soil Sci. 182, 945–952 (2019).CAS 
Article 

Google Scholar 
22.Lynch, J. P. Root phenes for enhanced soil exploration and phosphorus acquisition: Tools for future crops. Plant Physiol. 156, 1041–1049 (2011).CAS 
Article 

Google Scholar 
23.Lynch, J. P. Steep, cheap and deep: An ideotype to optimize water and N acquisition by maize root systems. Ann. Bot. 112, 347–357 (2013).CAS 
Article 

Google Scholar 
24.Lambers, H., Shane, M., Cramer, M., Pearse, S. & Veneklaas, E. Root structure and functioning for efficient acquisition of phosphorus: Matching morphological and physiological traits. Ann. Bot. 98, 693–713 (2006).Article 

Google Scholar 
25.Trachsel, S., Kaeppler, S. M., Brown, K. M. & Lynch, J. P. Maize root growth angles become steeper under low N conditions. Field Crop Res 140, 18–31 (2013).Article 

Google Scholar 
26.Jobbagy, E. G. & Jackson, R. B. The distribution of soil nutrients with depth: Global patterns and the imprint of plants. Biogeochemistry 53, 51–77 (2001).CAS 
Article 

Google Scholar 
27.Sun, B. R., Gao, Y. Z. & Lynch, J. P. Large crown root number improves topsoil foraging and phosphorus acquisition. Plant Physiol. 177, 90–104 (2018).CAS 
Article 

Google Scholar 
28.Weih, M., Asplund, L. & Bergkvist, G. Assessment of nutrient use in annual and perennial crops: A functional concept for analyzing nitrogen use efficiency. Plant Soil 339, 513–520 (2011).CAS 
Article 

Google Scholar 
29.Malhi, S. S., Johnston, A. M., Schoenau, J. J., Wang, Z. H. & Vera, C. L. Seasonal biomass accumulation and nutrient uptake of wheat, barley and oat on a Black Chernozern soil in Saskatchewan. Can. J. Plant Sci. 86, 1005–1014 (2006).Article 

Google Scholar 
30.Maeoka, R. E. et al. Changes in the phenotype of winter wheat varieties released between 1920 and 2016 in response to in-furrow fertilizer: Biomass allocation, yield, and grain protein concentration. Front. Plant Sci. 10, 1786 (2020).Article 

Google Scholar 
31.Pourazari, F., Vico, G., Ehsanzadeh, P. & Weih, M. Contrasting growth pattern and nitrogen economy in ancient and modern wheat varieties. Can. J. Plant Sci. 95, 851–860 (2015).Article 

Google Scholar 
32.Rietra, R. P. J. J., Heinen, M., Dimkpa, C. O. & Bindraban, P. S. Effects of nutrient antagonism and synergism on yield and fertilizer use efficiency. Commun. Soil Sci. Plant Anal. 48, 1895–1920 (2017).CAS 
Article 

Google Scholar 
33.Pedro, A., Savin, R. & Slafer, G. A. Crop productivity as related to single-plant traits at key phenological stages in durum wheat. Field Crop Res. 138, 42–51 (2012).Article 

Google Scholar 
34.Cakmak, I. & Yazici, A. M. Magnesium: A forgotten element in crop production. Better Crops Plant Food 94, 23–25 (2010).
Google Scholar 
35.Lancashire, P. D. et al. A uniform decimal code for growth-stages of crops and weeds. Ann. Appl. Biol. 119, 561–601 (1991).Article 

Google Scholar 
36.Trachsel, S., Kaeppler, S. M., Brown, K. M. & Lynch, J. P. Shovelomics: High throughput phenotyping of maize (Zea mays L.) root architecture in the field. Plant Soil 341, 75–87 (2011).CAS 
Article 

Google Scholar 
37.Colombi, T. & Walter, A. Root responses of triticale and soybean to soil compaction in the field are reproducible under controlled conditions. Funct. Plant Biol. 43, 114–128 (2016).Article 

Google Scholar  More

We asked whether the mating of male and female fruit flies would be affected by the presence of parasitoid wasps. We placed a pair of D. melanogaster flies in a small Petri dish, either with or without parasitoid wasps (Fig. 1a). In an initial experiment we used the wasp Leptopilina boulardi, which specializes on D. melanogaster and on closely related fly species14.Fig. 1: Exposure of Drosophila to wasps accelerates sexual behavior.a Courtship arena containing a male and virgin female fly with (left) and without (right) two wasps, one male and one female. b Copulation latency of D. melanogaster. p  More

Now that China has finally updated its List of Wildlife under Special State Protection, a more nimble and responsive approach is needed to aid conservation. The list should be reviewed every year, as well as subjected to the planned five-yearly updates. Species can quickly become endangered in times of rapid development.The latest additions are the first in more than 30 years (see go.nature.com/2q7sfga). During that time, China has changed profoundly, but the list of protected species has not kept pace. This lag has been disastrous for some animals that were not given the protection they needed.At least 33 species became extinct in China and many more are critically endangered (Y. Xie & W. Sung Integr. Zool. 2, 26–35; 2007; Z. Jiang et al. Biodivers. Sci. 24, 500–551; 2016).An independent government committee should be created to oversee amendments. When making decisions, it could refer to appendices of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) and the ‘red lists’ of threatened species curated by the Chinese Academy of Sciences and the International Union for Conservation of Nature (IUCN). These steps would build on the more forceful approach to managing wildlife that China has taken since the start of the COVID-19 pandemic. More

1.Whyte, R. O. Grassland and Fodder Resources of India Revised. (Indian Council of Agricultural Research, 1964).
Google Scholar 
2.Misra, R. The vegetation of the Indian Savannas. In Tropical Savannas (ed. Bourliere, F.) 151–166 (Elsevier, 1983).3.Behrensmeyer, A. K. et al. The structure and rate of late Miocene expansion of C4 plants: evidence from lateral variation in stable isotopes in paleosols of the Siwalik Group, northern Pakistan. GSA Bull. 119, 1486–1505 (2007).CAS 
Article 

Google Scholar 
4.Champion, H. G. & Seth, S. K. A Revised Survey of the Forest Types of India (Government of India Press, 1968).
Google Scholar 
5.Mani, M. S. The Flora. In Ecology and Biogeography in India (ed. Mani, M. S.) 159–177 (Dr. W. Junk b.v. Publishers, 1974).6.Ratnam, J., Tomlinson, K. W., Rasquinha, D. N. & Sankaran, M. Savannahs of Asia: antiquity, biogeography, and an uncertain future. Philos. Trans. R. Soc. B 371, 20150305 (2016).Article 
CAS 

Google Scholar 
7.Blasco, F. The transition from open forest to Savanna in continental Southeast Asia. In Tropical Savannas (ed. Bourliere, F.) 167–182 (Elsevier, 1983).8.Puri, G. S., Meher Homji, V. M., Gupta, R. K. & Puri, S. Forest Ecology. Phytogeography and Conservation Vol. 1 (Oxford & IBH Publishing, 1983).
Google Scholar 
9.Fuller, D. Q. & Korisettar, R. The vegetational context of early agriculture in South India. Man Environ. 29, 7–27 (2004).
Google Scholar 
10.Fuller, D. Q. Finding plant domestication in the Indian subcontinent. Curr. Anthropol. 52, S347–S362 (2011).Article 

Google Scholar 
11.Lehmann, C. E. R., Archibald, S. A., Hoffmann, W. A. & Bond, W. J. Deciphering the distribution of the savanna biome. New Phytol. 191, 197–209 (2011).PubMed 
Article 
PubMed Central 

Google Scholar 
12.Staver, A. C., Archibald, S. & Levin, S. A. Tree-cover in sub-Saharan Africa: rainfall and fire constrain forest and savanna as alternative stable states. Ecology 92, 1063–1072 (2011).PubMed 
Article 
PubMed Central 

Google Scholar 
13.Bond, W. J. What limits trees in C4 grasslands and savannas?. Annu. Rev. Ecol. Evol. Syst. 39, 641–659 (2008).Article 

Google Scholar 
14.Hirota, M., Holmgren, M., Van Nes, E. & Scheffer, M. Global resilience of tropical forest and savanna to critical transitions. Science 334, 232–235 (2011).ADS 
CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
15.Staver, A. C., Archibald, S. & Levin, S. A. The global extent and determinants of savanna and forest as alternative biome states. Science 334, 230–232 (2011).ADS 
CAS 
PubMed 
MATH 
Article 
PubMed Central 

Google Scholar 
16.Mayle, F. E. & Power, M. J. Impact of a drier early–mid-Holocene climate upon Amazonian forests. Philos. Trans. R. Soc. Lond. B Biol. Sci. 363, 1829–1838 (2008).PubMed 
PubMed Central 
Article 

Google Scholar 
17.Ngomanda, A. et al. Western equatorial African forest-savanna mosaics: a legacy of late Holocene climatic change?. Clim. Past 5, 647–659 (2009).Article 

Google Scholar 
18.Metwally, A. A., Scott, L., Neumann, F. H., Bamford, M. K. & Oberhänsli, H. Holocene palynology and palaeoenvironments in the Savanna Biome at Tswaing Crater, central South Africa. Palaeogeogr. Palaeoclimatol. Palaeoecol. 402, 125–135 (2014).Article 

Google Scholar 
19.Kuper, R. & Kröpelin, S. Climate-controlled Holocene occupation in the Sahara: motor of Africa’s evolution. Science 313, 803–807 (2006).ADS 
CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
20.Mayewski, P. A. et al. Holocene climate variability. Quat. Res. 62, 243–255 (2004).Article 

Google Scholar 
21.Wanner, H. et al. Mid- to late Holocene climate change: an overview. Quat. Sci. Rev. 27, 1791–1828 (2008).ADS 
Article 

Google Scholar 
22.Kathayat, G. et al. The Indian monsoon variability and civilization changes in the Indian subcontinent. Sci. Adv. 3, e1701296 (2017).ADS 
PubMed 
PubMed Central 
Article 

Google Scholar 
23.Shinde, V. The origin and development of the Chalcolithic in Central India. Indo-Pac. Prehist. Assoc. Bull. 19, 125–136 (2000).
Google Scholar 
24.Fuller, D. Q. Agricultural origins and frontiers in South Asia: a working synthesis. J. World Prehist. 20, 1–86 (2006).Article 

Google Scholar 
25.Fuller, D. Q., Boivin, N. & Korisettar, R. Dating the Neolithic of South India: new radiometric evidence for key economic, social and ritual transformations. Antiquity 81, 755–778 (2007).Article 

Google Scholar 
26.Johansen, P. G. Landscape, monumental architecture, and ritual: a reconsideration of the South Indian ashmounds. J. Anthropol. Archaeol. 23, 309–330 (2004).Article 

Google Scholar 
27.Fuller, D. Q. Asia, South: Neolithic cultures. In Encyclopedia of Archaeology (ed. Pearsall, D.) 756–768 (Springer, 2008).
Google Scholar 
28.Asouti, E. & Fuller, D. Q. Trees and Woodlands of South India: Archaeological Perspectives (Left Coast Press, 2008).
Google Scholar 
29.Singh, G., Joshi, R. D., Chopra, S. K. & Singh, A. B. Late quaternary history of vegetation and climate of the Rajasthan desert, India. Philos. Trans. R. Soc. Lond. B Biol. Sci. 267, 467–501 (1974).ADS 
Article 

Google Scholar 
30.Singh, I. B. Quaternary palaeoenvironments of the Ganga plain and anthropogenic activity. Man Environ. 30, 1–35 (2005).
Google Scholar 
31.Clarkson, C. et al. The oldest and longest enduring microlithic sequence in India: 35 000 years of modern human occupation and change at the Jwalapuram locality 9 rockshelter. Antiquity 83, 326–348 (2009).Article 

Google Scholar 
32.Riedel, N. et al. Modern pollen vegetation relationships in a dry deciduous monsoon forest: a case study from Lonar Crater Lake, central India. Quat. Int. 371 (2015).33.Sarkar, S. et al. Monsoon source shifts during the drying mid-Holocene: biomarker isotope based evidence from the core ‘monsoon zone’ (CMZ) of India. Quat. Sci. Rev. 123, 144–157 (2015).ADS 
Article 

Google Scholar 
34.Chakraborty, A., Joshi, P. K., Ghosh, A. & Areendran, G. Assessing biome boundary shifts under climate change scenarios in India. Ecol. Indic. 34, 536–547 (2013).Article 

Google Scholar 
35.Rasquinha, D. N. & Sankaran, M. Modelling biome shifts in the Indian subcontinent under scenarios of future climate change. Curr. Sci. 111, 147–156 (2016).Article 

Google Scholar 
36.Berkelhammer, M. et al. An abrupt shift in the Indian monsoon 4000 years ago in Climates, Landscapes, and Civilizations (eds. Giosan, L. et al.) 75–88 (American Geophysical Union, 2013).37.Fleitmann, D. et al. Holocene ITCZ and Indian monsoon dynamics recorded in stalagmites from Oman and Yemen (Socotra). Quat. Sci. Rev. 26, 170–188 (2007).ADS 
Article 

Google Scholar 
38.Sinha, A. et al. A global context for megadroughts in monsoon Asia during the past millennium. Quat. Sci. Rev. 30, 47–62 (2011).ADS 
Article 

Google Scholar 
39.Berkelhammer, M. et al. Persistent multidecadal power of the Indian Summer Monsoon. Earth Planet. Sci. Lett. 290, 166–172 (2010).ADS 
CAS 
Article 

Google Scholar 
40.Laskar, A. H., Yadava, M. G., Ramesh, R., Polyak, V. J. & Asmerom, Y. A 4 kyr stalagmite oxygen isotopic record of the past Indian Summer Monsoon in the Andaman Islands. Geochem. Geophys. Geosyst. 14, 3555–3566 (2013).ADS 
CAS 
Article 

Google Scholar 
41.Thamban, M., Kawahata, H. & Rao, V. P. Indian summer monsoon variability during the Holocene as recorded in sediments of the Arabian Sea: timing and implications. J. Oceanogr. 63, 1009–1020 (2007).Article 

Google Scholar 
42.Ponton, C. et al. Holocene aridification of India. Geophys. Res. Lett. 39, L03704 (2012).ADS 
Article 

Google Scholar 
43.Deblauwe, V. et al. Remotely sensed temperature and precipitation data improve species distribution modelling in the tropics. Glob. Ecol. Biogeogr. 25, 443–454 (2016).Article 

Google Scholar 
44.Gaussen, H. et al. International Map of the Vegetation at Scale 1:1.000.000 (French Institute of Pondycherry, 1964).
Google Scholar 
45.ESRI Inc. ArcGIS Pro (ESRI Inc., 2019).
Google Scholar 
46.Saha, K. Tropical Circulation Systems and Monsoons (Springer, 2010).Book 

Google Scholar 
47.Goswami, B. N. South Asian monsoon. In Intraseasonal Variability in the Atmosphere–Ocean Climate System (eds. Lau, W. K. M. & Waliser, D. E.) 19–61 (Springer, 2005).48.Dabadghao, P. M. & Shankarnarayan, K. A. The Grass Cover of India (Indian Council of Agricultural Research, 1973).
Google Scholar 
49.Prasad, S. & Enzel, Y. Holocene paleoclimates of India. Quat. Res. 66, 442–453 (2006).Article 

Google Scholar 
50.Fleitmann, D. et al. Palaeoclimatic interpretation of high-resolution oxygen isotope profiles derived from annually laminated speleothems from Southern Oman. Quat. Sci. Rev. 23, 935–945 (2004).ADS 
Article 

Google Scholar 
51.Kale, V. S. Fluvio–sedimentary response of the monsoon-fed Indian rivers to Late Pleistocene–Holocene changes in monsoon strength: reconstruction based on existing 14C dates. Quat. Sci. Rev. 26, 1610–1620 (2007).ADS 
MathSciNet 
Article 

Google Scholar 
52.Prasad, S. et al. Prolonged monsoon droughts and links to Indo-Pacific warm pool: a Holocene record from Lonar Lake, central India. Earth Planet. Sci. Lett. 391, 171–182 (2014).ADS 
CAS 
Article 

Google Scholar 
53.Dixit, Y., Hodell, D. A. & Petrie, C. A. Abrupt weakening of the summer monsoon in northwest India ∼ 4100 yr ago. Geology https://doi.org/10.1130/G35236.1 (2014).Article 

Google Scholar 
54.Laskar, J. et al. A long-term numerical solution for the insolation quantities of the Earth. Astron. Astrophys. 428, 261–285 (2004).ADS 
Article 

Google Scholar 
55.Marzin, C. & Braconnot, P. Variations of Indian and African monsoons induced by insolation changes at 6 and 9.5 kyr BP. Clim. Dyn. 33, 215–231 (2009).Article 

Google Scholar 
56.Bush, R. T. & McInerney, F. A. Leaf wax n-alkane distributions in and across modern plants: implications for paleoecology and chemotaxonomy. Geochim. Cosmochim. Acta 117, 161–179 (2013).ADS 
CAS 
Article 

Google Scholar 
57.Murphy, C. & Fuller, D. Q. The agriculture of early India. In Oxford Research Encyclopedia of Environmental Science (ed. Shugart, H.) (Oxford University Press, 2017).
Google Scholar 
58.Kumaran, N. K. P. et al. Vegetation response and landscape dynamics of Indian Summer Monsoon variations during Holocene: an eco-geomorphological appraisal of tropical evergreen forest subfossil logs. PLoS ONE 9, e93596 (2014).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
59.Singh, G., Wasson, R. J. & Agrawal, D. P. Vegetational and seasonal climatic changes since the last full glacial in the Thar Desert, northwestern India. Rev. Palaeobot. Palynol. 64, 351–358 (1990).Article 

Google Scholar 
60.Cole, M. M. The Savannas, Biogeography and Geobotany (Academic Press, 1986).61.Sankaran, M. et al. Determinants of woody cover in African savannas. Nature 438, 846–849 (2005).ADS 
CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
62.Kodandapani, N., Cochrane, M. A. & Sukumar, R. A comparative analysis of spatial, temporal, and ecological characteristics of forest fires in seasonally dry tropical ecosystems in the Western Ghats, India. For. Ecol. Manag. 256, 607–617 (2008).Article 

Google Scholar 
63.Hegde, V., Chandran, M. D. S. & Gadgil, M. Variation in bark thickness in a tropical forest community of Western Ghats in India. Funct. Ecol. 12, 313–318 (1998).Article 

Google Scholar 
64.Stott, P. A., Goldammer, J. G. & Werner, W. L. The role of fire in the tropical lowland deciduous forests of Asia. In Fire in the Tropical Biota. Ecosystem Processes and Global Challenges (ed. Goldammer, J. G.) 32–44 (Springer, 1990).65.Murphy, C. & Fuller, D. Q. Seed coat thinning during horsegram (Macrotyloma uniflorum) domestication documented through synchrotron tomography of archaeological seeds. Sci. Rep. 7, 5369 (2017).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
66.Kingwell-Banham, E. & Fuller, D. Q. Shifting cultivators in South Asia: expansion, marginalisation and specialisation over the long term. Quat. Int. 249, 84–95 (2012).Article 

Google Scholar 
67.Kajale, M. Excavation at Inamgaon (Deccan College Postgraduate and Research Institute, 1988).
Google Scholar 
68.Shirvalkar, P. & Prasad, E. The archaeology of the Late Holocene on the Deccan Plateau (The Deccan Chalcolithic). In A Companion to South Asia in the Past (eds. Schug, G. R. & Walimbe, S. R.) 240-254 (John Wiley & Sons, 2016).69.Roberts, P. et al. Local diversity in settlement, demography and subsistence across the southern Indian Neolithic-Iron Age transition: site growth and abandonment at Sanganakallu-Kupgal. Archaeol. Anthropol. Sci. 8, 575–599 (2016).Article 

Google Scholar 
70.Nayar, T. S. Pollen Flora of Maharashtra State, India (Today & Tomorrow Printers and Publishers, 1990).
Google Scholar 
71.APSA Members. The Australasian Pollen and Spore Atlas V1.0 (Australian National University, 2007).
Google Scholar 
72.Tinner, W. & Hu, F. S. Size parameters, size-class distribution and area-number relationship of microscopic charcoal: relevance for fire reconstruction. Holocene 13, 499–505 (2003).ADS 
Article 

Google Scholar 
73.Conedera, M. et al. Reconstructing past fire regimes: methods, applications, and relevance to fire management and conservation. Quat. Sci. Rev. 28, 555–576 (2009).ADS 
Article 

Google Scholar 
74.Higuera, P., Peters, M., Brubaker, L. & Gavin, D. Understanding the origin and analysis of sediment-charcoal records with a simulation model. Quat. Sci. Rev. 26, 1790–1809 (2007).ADS 
Article 

Google Scholar 
75.McDermott, F. Palaeo-climate reconstruction from stable isotope variations in speleothems: a review. Quat. Sci. Rev. 23, 901–918 (2004).ADS 
Article 

Google Scholar 
76.Baldini, J., McDermott, F. & Fairchild, I. Spatial variability in cave drip water hydrochemistry: implications for stalagmite paleoclimate records. Chem. Geol. 235, 390–404 (2006).ADS 
CAS 
Article 

Google Scholar 
77.Allchin, B. & Allchin, F. R. The Rise of Civilization in India and Pakistan (Cambridge University Press, 1982).
Google Scholar 
78.Shinde, V. S. New light on the origin, settlement system and decline of the Jorwe culture in the Deccan India. South Asian Stud. 5, 59–72 (1989).Article 

Google Scholar 
79.Shinde, V. S. Settlement pattern of the Savalda culture—the first farming community of Maharashtra. Bull. Deccan Coll. Res. Inst. 49–50, 417–426 (1990).
Google Scholar 
80.Paddayya, K. Investigations Into the Neolithic Culture of the Shorapur Doab, South India Vol. 3 (Brill, 1973).
Google Scholar  More

For each case study area, a search query was executed (Table 1). Query terms were based on the hashtags of the geographical name of the study areas; therefore, the post download was related to the name of the study area (e.g., Galapagos), with all downloaded posts including this name as query. Query search was limited to English, the most common language amongst tourists. This might have overlooked posts where the name of the place was in a different language. For most marine areas, this was considered irrelevant as the name of the place is not translated to other languages (e.g., Tayrona, Vamizi, Skomer). In some of the cases, the name of the place could appear in a variety of languages (e.g., Great Barrier Reef), however, the use of non-English place hashtags as queries generally retrieved a significantly lower number of posts (e.g., Gran Barrera de Coral in Spanish with 1900 posts, or Grand Barrière de Corail in French with 14 posts, while Great Barrier Reef had over 10,000 posts). In the specific case of Easter Island, we observed that the use of three particular queries was linked to a high number of posts: Easter Island and the local name Rapanui had over 10,000 posts each, and Isla de Pascua in Spanish had 8700 posts. In this case, three separate posts’ downloads were performed, and data were merged for subsequent analysis. The above, rather than a limitation of the methodological approach, demonstrates its flexibility to adapt to different data acquisition requirements.To illustrate the most relevant information contained as part of the posts downloaded for each of the 14 areas, we selected the 150 most frequent hashtags from each dataset in order to create the network graph and represent the dominant discourse in relation to the area in question. Network graphs were delineated using eigenvector, betweenness and edge betweenness as centrality measures. Eigenvector centrality measure (hereafter Eigenvector) allows identifying those hashtags that are frequently posted with other hashtags also frequently posted, and it can be interpreted as the pairs or groups of features more frequently related to the case study by the users. Betweenness centrality (hereafter betweenness) and edge betweenness centrality (hereafter edge betweenness) provide information about clusters of hashtags that describe users’ experiences or perceptions and that connect (by means of a hashtag) to other clusters representing other types of experiences or perceptions. These high betweenness hashtags structure the general discourse about an area and their removal would fragment the network and disconnect distant concepts. Therefore, hashtags and links with high betweenness can show the discourse parallel or additional to the main discourse and their relations, allowing to identify less frequent activities or perceptions but that are equally important to understand the network as a whole.Network centrality measuresResults indicated that network graphs captured information on distinct types of ecosystem services, for example, those based on wildlife and nature, heritage, or beach tourism. In areas such as Galapagos, central hashtags were nature, wildlife, photography, travel and adventure, evidencing a preference for wildlife and nature-based tourism. In this area, betweenness evidenced the connections between the most frequent hashtags group with other peripheric hashtags and provided a complete picture on the discourse of Galapagos’ visitors (Fig. 2). As such, nature and wildlife-based travel and photography is related with natural science concepts like evolution and endemism, and specific biotic and abiotic components like crabs and waves, altogether related with positive feelings (i.e., happy). Other areas emerging for their wildlife and nature were Skomer nature reserve, characterised by the hashtags birds (including the species Puffin), nature and wildlife photography; and Península Valdés, characterized by many locality names and by fauna, with the frequently posted hashtags’ wildlife, whales and nature funnelling most connections to other less frequent hashtags (e.g., wind, hiking, relax) and providing a full picture of the social perception on nature recreation activities, iconic fauna and positive feelings. Three networks, Sandwich Harbour, Glacier Bay and Macquarie Island also included popular hashtags related with nature, wildlife and photography; however, most hashtags had low betweenness and edge betweenness limiting the diversity of the posts (all network graphs are available at the Figshare repository, https://doi.org/10.6084/m9.figshare.13325627.v2).Figure 2Example of network graphs in Galapagos case study. In plot (A) node size represents the Eigenvector centrality and edges represent normalized strength (weighted degree). In plot (B) node size represents normalized Betweenness centrality and edges represent normalized Edge betweenness.Full size imageRegarding cultural heritage, Easter Island was characterised by popular hashtags related with Easter Island stone statues (moais) and with travel; and edge betweenness evidenced a diversity of peripherical nodes that describe other cultural elements, like design, music and food, and evidence social preferences for different cultural elements of the island, beyond the moais. Other areas reflected cultural identity by the frequent post of local names (e.g., Ytrehvaler), words related with the country’s identity (e.g., Isole Egadi) and positive feelings about this identity (e.g., Tawharanui). In Tayrona National Park network, the full discourse identified cultural identity like Kogui (indigenous culture) linked with the popular posts related with nature and summer holidays. Similarly, in Tawharanui and Isole Egadi, beach, nature and summer where the most frequent posts that, in some cases, where connected with places and activities. In these cases, and particularly in Isole Egadi and Ytrehvaler, edge betweenness allows to identify connections between places and activities, wildlife or natural structures, providing relevant information for area management and conservation.A group of areas were appreciated by their underwater ecosystems. For Great Barrier Reef, popular hashtags were related with the coral reef: ocean, diving, underwater photography, travel, nature, coral and reef; whereas betweenness highlighted a set of hashtags related with conservation: science, sustainability, save the reef, 4 ocean (Fig. 3) and evidenced the presence of a conservationist discourse in the social media. In Toguean Island network, the frequent hashtags beach, wonderful and charming are connected to peripherical hashtags related with the sea (e.g., sea life, diving), while in Vamizi, popular hashtags were related with high-income tourism, private island, travel, luxury travel, and were connected to less frequent hashtags linked to the sea, including recreational fisheries. These last two examples illustrate differences in the benefits, and beneficiaries, provided by two popular touristic destinations.Figure 3Example of network graphs in Great Barrier Reef case study. In plot (A) node size represents the Eigenvector centrality and edges represent normalized strength (weighted degree). In plot (B) node size represents normalized Betweenness centrality and edges represent normalized Edge betweenness.Full size imageNetwork communitiesThe division of hashtags in communities allows for a more detailed exploration of the words included in the 150 most frequent hashtags selection, independently of their centrality measures, and allowed a categorisation of hashtags within cultural ecosystem services classes in each area (Table 2). Hashtags were grouped in 3 to 5 communities, with some communities relatively constant across case studies, e.g., aesthetics, wildlife and nature appreciation (Fig. 4) (all other network graphs are available at the Figshare repository, https://doi.org/10.6084/m9.figshare.13325627.v2).Table 2 Cultural Ecosystem Services’ types (CES) depicted from the community analysis (Fast Greedy algorithm). The order of the CES class does not imply a priority rank.Full size tableFigure 4Communities assessed through Fast-Greedy algorithm for the case studies Glacier Bay (A) and Tayrona (C). The node size represents the normalized Eigenvector and the colour represents the community. The colour and width of the edges represents the normalized edge strength (weighted degree).Full size imageIn some of the areas, the communities were diverse in hashtag composition, for example, in Galapagos, wildlife (and related words) was distinctive of several communities, but other communities were characterised by different concepts: beach, holidays, happiness, snorkelling and diving. In Easter Island, the hashtags related with the stone statues and cultural heritage characterise one community, while the other communities include a diversity of hashtags classified under adventure, nature, underwater recreational activities; therefore, it widens the information provided by the centrality metrics. Tayrona (Fig. 4) is also a diverse network with one community characterised by hashtags like beach, summer, happiness (wellbeing), but other communities contain a diversity of hashtags like forest, hiking, indigenous and wildlife (classified in recreational, cultural heritage, nature and aesthetics; Table 2).In some areas, the communities were not so diverse, but provided additional information on the posts. For example, in MacQuarie Island the communities highlighted iconic fauna, including several penguin species, and biodiversity conservation. In several areas, network communities informed of the iconic fauna and specific places: puffins and other bird species in Skomer; southern right whale, sealions and penguins in Península Valdés; glaciers and mountains in Glacier bay (Fig. 4); desert and dunes in Sandwich harbour. Finally, Ytrehvaler is a network characterised by many local names (in Norwegian), evidencing a national tourism, and hashtags related with scenery.Merged network of the 14 case studiesThe merged network highlighted several hashtags that act as bridges between communities of hashtags (Fig. 5). Nature, travel, photo and travel photography are key to structure the global network. However, several low eigenvector hashtags connect smaller groups: sunset and island connect the subgroups from Easter Island, Isole Egadi and Vamizi.Figure 5Global network graph including the fourteen case studies where the node size represents the Eigenvector centrality. The coloured clusters arrange the case studies to facilitate the visual identification of areas connected in the network.Full size imageFrom the hashtag travel photography diverges a branch that connects 7 areas through adventure; a small group of hashtags deriving from this node represent Sandwich harbour and Vamizi, connected through Africa. The hashtag ocean, connected to adventure, relates Great Barrier Reef with Tawharanui, and to wanderlust (a German expression for the desire to explore the world) that connects Península Valdés, Skomer and Macquairie Island. These three areas and Tayrona are also connected through the central hashtag travel photography, and Skomer and Macquairie Island through wildlife photography. The hashtag adventure is also connected to a group of hashtags from Galapagos that also derive to the high eigenvector hashtag nature.The hashtag nature is key to include the fragile sub-network Ytrehvaler, and also derives to other high eigenvector hashtag, travel, that in turn, connects to the small sub-network from Glacier bay. Photo, a central hashtag related with travel, connects to paradise, that is key to integrate Toguean Island, a few hashtags from Tayrona related with the Caribbean and beach, and a group of hashtags from Peninsula Valdez related with whale watching. Some other small hashtags, that are connected to high eigenvector hashtags but are not included in any particular area are shared by many of the areas, e.g., sun, relax, landscape photography, nature lovers, sunset, sky. More