Ecology

Animals and dietsThe animal experiments were conducted in accordance with the Guidelines for Animal Experiments and Act on Welfare and Management of Animals, Hokkaido University, and all experimental procedures were approved by the Animal Care and Use Committee of Hokkaido University. All animal experiments were carried out in accordance with ARRIVE guidelines. Twenty newborn female Holstein calves with an average birth weight of 37.1 ± 1.0 kg (mean ± standard error) were randomly allocated to either the control or treatment group at birth. All calves were housed individually in separate calf hutches containing sawdust bedding. Feeding and managing of animals until weaning at 50 d of age was performed as described previously17. After supplementing colostrum at birth, calves in both groups were fed 4 L of pasteurized whole milk (44.2% crude protein [CP] and 29.3% fat on a dry matter [DM] basis) as a transition milk during the first week since birth. From 8 days until weaning at 50 days of age, milk replacer (28.0% CP and 18.0% fat on a DM basis) was fed twice daily at 0830 and 1600 h. Water, calf starter (22.9% CP, 11.0% neutral detergent fiber [NDF], 5.6% acid detergent fiber [ADF], 6.2% crude ash, and 3.0% ether extract on a DM basis), and chopped Timothy hay (3.4% CP, 53.1% NDF, 34.2% ADF, 4.3% crude ash, and 1.7% ether extract on a DM basis) were provided for ad libitum intake from 3 days of age. In addition to voluntary intake of solid diets, the calves in the treatment group were orally administered with a mixture of ground Timothy hay and psyllium (4.4% CP, 78.6% NDF, 5.8% ADF, 3.9% crude ash, and 0.3% ether extract on a DM basis) from 3 days until weaning at 50 days of age. Timothy hay was ground for oral administration using a Wiley grinder (WM-3, Irie Shokai) with a 2-mm screen. To improve the handling of the treatment diet for oral administration, we incorporated psyllium, which is a dietary fiber that primarily improves gastrointestinal conditions in humans and can be incorporated in oral electrolyte solution supplemented to neonatal calves38. As a treatment diet, ground Timothy hay (50 g) and psyllium (6 g) were mixed with 200 mL of water. Owing to the adhesiveness of psyllium, the treatment diet formed a “hay ball” and showed slight stickiness, which facilitates swallowing by calves. At 3–7 days of age, one hay ball (50 g of fibrous diet) was orally administered after morning milk feeding. From 8 days of age to weaning, an additional hay ball was fed immediately after evening milk feeding (100 g fibrous diet per day).After weaning, animals in both dietary groups were merged into the same herd and managed on the same farm under identical conditions. From 9 months of age until calving, heifers were fed a ration containing Timothy hay, alfalfa hay, fescue hay, and concentrate. After calving, the cows were fed a diet for lactating cows, as described in Supplementary Table S8. Diets comprised a total mixed ration and were fed twice daily at 0900 and 1600 h. All animals had ad libitum access to water and mineral blocks throughout the experiment. Daily milk production for each cow was measured for the first 30 days of the lactation period and the average values for each dietary group on a weekly and monthly basis were calculated. Milk yield for four animals in each dietary group were not recorded due to health problems including mastitis and displaced abomasum symptoms after calving.In this study, all animals (n = 20) were maintained until 9 months of age, without severe problems. Owing to health problems, several animals were excluded from the experiment before parturition as follows: three animals (one in the control group and two in the treatment group) at 60 days before the expected calving date and one animal in the control group at 21 days before the expected calving date. One animal in the control group (15 days after calving) and two animals in the treatment group (calving day) were diagnosed with displaced abomasum symptoms and were excluded from further sampling. Owing to technical problems, samples were not collected from three animals aged 7 days in the treatment group and one animal aged 21 days in the control group. All other samples (n = 176) were obtained at the target sampling points.Sampling of rumen contentsRumen contents were collected orally using a stomach tube. The stomach tube and the sample collection flask were thoroughly cleaned using water between sample collections from individual animals; the first fraction of the sample was discarded to avoid contamination from the previous sample and saliva. All samples were collected at 4 h after morning feeding. Rumen contents were collected at 7, 21, 35, 49, and 56 days, and at 9 months of age, 60 and 21 days before the expected calving date, at calving day, and 21 days after calving. The pH was measured using a pH meter (pH meter F-51; Horiba, Kyoto, Japan) immediately after sampling. Samples were collected in a sterile 50 mL tube and immediately placed on ice, followed by storage at − 30 °C until use.Chemical analysisRumen contents (1.0 g) were centrifuged at 16,000×g at 4 °C for 5 min, and the supernatant was collected. The SCFA content was analyzed using a gas chromatograph (GC-14B; Shimadzu, Kyoto, Japan) as described previously39. In brief, the supernatant of the rumen contents was mixed with 25% meta-phosphoric acid at a 5:1 ratio, incubated overnight at 4 °C, and centrifuged at 10,000×g at 4 °C. The supernatant was then mixed with crotonic acid as an internal standard and injected into a gas chromatograph equipped with an ULBON HR-20 M fused silica capillary column (0.53 mm i.d. × 30 m length, 3.0 µm film; Shinwa, Kyoto, Japan) and a flame-ionization detector. d/l-lactic acid levels were measured using a commercial assay kit (Megazyme International Ireland, Wicklow, Ireland) according to the manufacturer’s instructions. NH3-N levels were measured via the phenol-hypochloride reaction method40 using a microplate reader at 660 nm (ARVO MX; Perkin Elmer, Yokohama, Japan).DNA extraction and rumen microbiota profiling via amplicon sequencingTotal DNA was extracted and purified using the repeated bead-beating plus column method41. Rumen contents (0.25 g) were homogenized using sterile glass beads (0.4 g; 0.3 g of 0.1 mm and 0.1 g of 0.5 mm) and cell lysis buffer (1 mL; 500 mM NaCl, 50 mM Tris–HCl [pH 8.0], 50 mM ethylenediaminetetraacetic acid (EDTA), and 4% sodium dodecyl sulfate). The lysates were then incubated at 70 °C for 15 min, and the supernatant was collected for further processing. Bead-beating and incubation steps were repeated once, and all supernatants were combined. Total DNA was precipitated using 10 M ammonium acetate and isopropanol, followed by purification using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany). The DNA concentration was quantified using a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and adjusted with Tris–EDTA buffer to the appropriate concentration.For a comprehensive analysis of rumen bacterial communities, the MiSeq sequencing platform (Illumina, San Diego, CA, USA) was used. Total DNA obtained from the rumen contents was diluted to a final concentration of 5 ng/μL and subjected to PCR amplification of the V3-V4 regions of the 16S rRNA gene using the primer sets S-D-Bact-0341-b-S-17 (5′-CCTACGGGNGGCWGCAG-3′) and S-D-Bact-0785-a-A-21 (5′-GACTACHVGGGTATCTAATCC-3′)42. The PCR mixture consisted of 12.5 μL of 2× KAPA HiFi HotStart Ready Mix (Roche Sequencing, Basel, Switzerland), 0.1 μM of each primer, and 2.5 μL of DNA (5 ng/μL). PCR amplification was performed according to the following program described previously9: initial denaturation at 95 °C for 3 min; 25 cycles at 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 30 s; and a final extension step at 72 °C for 5 min. Amplicons were purified using AMPure XP beads (Beckman-Coulter, Brea, CA, USA) and subjected to sequencing on the Illumina MiSeq platform (Illumina) using the MiSeq Reagent Kit v3 (2 × 300 paired-end). Data obtained from amplicon sequencing using the MiSeq platform were analyzed using QIIME2 version 2019.443. Paired reads were filtered, dereplicated, merged, and chimera-filtered using the q2-dada2 plugin44 to generate ASVs. Taxonomic classification of the ASVs was performed at the phylum, class, order, family, and genus levels using the SILVA 132 99% operational taxonomic units, full length, seven level taxonomy classifier (silva-132-99-nb-classifier.qza). Sequenced data were processed further and analyzed using R software version 3.6.245. ASV and taxonomy tables generated using QIIME2 were imported into R and merged with the sample metadata using the Phyloseq Bioconductor packages46. ASVs identified as Archaea, chloroplasts, and mitochondria were excluded. All samples were rarefied to a sampling depth of 16,805 reads, which was the smallest number of reads observed per sample in the filtered ASV table. Alpha diversity indices including Chao1, ACE, Shannon, and Simpson indices were calculated using the phyloseq function “estimate_richness”. PCoA was performed to determine differences in the microbial community structure based on the Bray–Curtis dissimilarity matrices at the genus level using the Phyloseq package. Venn diagrams were generated using ASVs showing mean relative sequence abundances of  > 0.1% in either the control or the treatment groups at each sampling point. The relative abundance of each bacterial taxon was calculated by dividing the number of reads assigned to each taxon by the total number of reads. Taxa with an average relative abundance  > 0.1% in  > 50% of samples in either the control or treatment group during at least one sampling point were used for the analysis. Hierarchical cluster analysis of bacterial genera determined via amplicon sequencing at 21 days after calving and the weekly and monthly average milk yield for the first 30 days of lactation period was performed using the distances calculated from Spearman’s correlation and average linkage clustering.Quantification of target bacterial species/groups using real-time PCRThe relative abundance of known ruminal bacterial species and groups, including the total bacteria, F. succinogenes, R. flavefaciens, Ruminococcus albus, Butyrivibrio spp., Prevotella spp., Selenomonas ruminantium, Megasphaera elsdenii, Treponema spp., Streptococcus bovis, Anaerovibrio lipolytica, and Ruminobacter amylophilus, was quantified using real-time PCR. Amplification was performed using a Light Cycler 480 system (Roche Applied Science, Mannheim, Germany) with a KAPA SYBR Fast qPCR Kit (Roche Sequencing, Basel, Switzerland) and the respective primer sets (Supplementary Table S9). The standards used for the real-time PCR were prepared as described previously47. Briefly, plasmid DNA containing the respective target bacterial 16S rRNA gene sequence was obtained by PCR cloning using the species/genus-specific or bacterial universal primer sets. The concentration of the plasmid was determined with a spectrometer. Copy number of each standard plasmid was calculated using the molecular weight of nucleic acid and the length (base pair) of the cloned standard plasmid. Ten-fold dilution series ranging from 1 to 108 copies were prepared for each target and run along with the samples. The respective genes were quantified using standard curves obtained from the amplification profile of the dilution series of the plasmid DNA standard (Supplementary Table S9). The PCR cycling conditions and reaction mixture were the same as those reported previously48. The relative abundance of each bacterial target was expressed as the proportion (%) of the abundance of the 16S rRNA genes of each bacterial target relative to that of the total bacteria.Statistical analysisAll data were sorted based on animal age into two sets, from 7 to 56 days of age and from 9 months of age to 21 days after calving, and analyzed separately. Data on fermentation parameters and bacterial abundance quantified via real-time PCR were analyzed using a repeated measures model using GraphPad Prism software version 9.1 (GraphPad Software, San Diego, CA, USA) with the fixed effects of dietary group, age, and diet × age interaction, and the random effect of animals within the groups. The Greenhouse–Geisser correction was used where sphericity was violated. If the P-value for the treatment effect was  More

Database utilizedThe database comprises information obtained with permission from the Catalogue of Endangered Languages that is hosted on the Endangered Languages Project platform (https://www.endangeredlanguages.com/). The Endangered Languages Project was first developed and launched by Google, and is currently overseen by First People’s Cultural Council and the Institute for Language Information and Technology at Eastern Michigan University. Information about the languages in this project is provided by the Catalogue, which is produced by the University of Hawai’i at Mānoa and Eastern Michigan University, with funding provided by the U.S. National Science Foundation (Grants #1058096 and #1057725) and the Luce Foundation. The project is supported by a team of global experts comprising its Governance Council and Advisory Committee.In general, the Catalogue aims to present all languages that communities and scholars have pointed out to be at some level of risk as well as languages that have become dormant. In addition to being the largest database of endangered languages globally, the Catalogue is updated periodically based on feedback gathered from language communities and scholars worldwide. The data therefore represents what was most accurately known about the state of each language’s vitality at its point of utilization. At the time of usage, there were 3423 languages represented in the Catalogue that were determined to be at various levels of risk. Assessment of each language’s risk level is carried out using the Language Endangerment Index, which was developed for the Catalogue’s purposes. The Index is used to assess the level of endangerment of any given language based on whether there is intergenerational transmission of the language (whether the language is being passed on to younger generations), its absolute number of speakers, speaker number trends (whether numbers are stable, increasing, or decreasing), and domains of language use (whether the language is used in a wide number of domains or limited ones). The levels of endangerment that the Index generates include ‘safe’, ‘vulnerable’, ‘threatened’, ‘endangered’, ‘severely endangered’, and ‘critically endangered’. Languages for which it remains unclear if the language has gone extinct or whose last fluent speaker is reported to have died in recent times are referred to as ‘dormant’. Given that the focus of the Catalogue is languages that are at some level of threat, safe languages are excluded in general. Where locality information is available, each language is also accompanied with its latitudinal and longitudinal coordinates.Steps taken to prepare the data for network analysisThe data obtained from the Catalogue was further organized and cleaned up for analysis.

1.

Identifier code
Where available, the ISO 639-3 code for each language was utilized as its unique identifier. Otherwise, its LINGUIST List local use code was utilized. These are temporary codes that are not in the current version of the ISO 639-3 Standard for languages. For languages with neither, unique 3-letter codes were constructed.

2.

Endangerment level
Each language’s endangerment level appeared together with a level of certainty score in the same cell in the original data file. Both pieces of information were split into separate columns and only endangerment levels were utilized.
For languages where different data were available in the Catalogue depending on resource utilized, the data was listed in additional columns. The endangerment level data points utilized in these cases were the ones with the most complete and updated information. If there was no data available regarding endangerment level, this information was also reflected.

3.

Coordinates
Where exact coordinates were not available, coordinates were approximated using Google maps based on the location description provided in the Catalogue source (e.g., the Tel Aviv district), attained from other sources such as Glottolog, UNESCO Atlas of the World’s Languages in Danger, or approximated from maps provided in other sources. ‘NA’ was indicated in the field for coordinates if none could be found.
Coordinates found to be inaccurate were rejected, for example in the instance that coordinates provided indicate a different location than the country the language is supposedly found in. The above steps were then taken to populate the coordinates field.
In instances where a language appears in more than one country, these are listed in separate rows as separate entries. Where there are two sets of coordinates for a country, the set that best corresponds with the written description in the Catalogue source, has greater detail, or is more recent is chosen. Where there are more than two sets of coordinates, a middle point is chosen as being representative of the language’s location, by plotting all coordinates on MapCustomizer (www.mapcustomizer.com).

4.

Language family
On the Catalogue, the information regarding language family may be multi-tiered. For example, Laghuu falls under the Lolo-Burmese branch of the Sino-Tibetan family. For this study, the broader family is utilized—in the case of Laghuu the label ‘Sino-Tibetan’ is used.
Mixed languages, pidgins, and creoles have all been categorized as ‘contact languages’.
Language isolates are listed as ‘isolates’.

5.

Region

The Catalogue groups ‘Mexico, Central America, Caribbean’ together under region. Central America and Caribbean are listed as separate regions in this study, with Mexico falling under Central America.Network constructionA spatial network of endangered languages was constructed from the database. Each node represented an endangered language, and edges or links depicted the distance between the locations of the languages as specified in the database. A distance matrix containing the distances between all endangered languages was computed by using functions from the ‘geosphere’ R package. Specifically, Haversine distances were computed for each pair of longitude and latitude points in the dataset. The radius of the earth used in the Haversine distance calculation is 6,378,137 m (for more details see: https://www.rdocumentation.org/packages/geosphere/versions/1.5-14/topics/distHaversine). Haversine distance refers to the shortest distance between two points on a spherical earth, also referred to as the “great-circle-distance”29.Sensitivity analyses of edge thresholdsThe distance matrix is a fully connected network with weighted, undirected links. We set out to capture the strongest or “closest” spatial relationships among the endangered languages, therefore an edge threshold was applied to the distance matrix such that only the edges in the xth lowest percentile were retained in the spatial network. Such an approach allows for the analysis of the most meaningful (i.e., the physically closest) spatial relations in the dataset and how they relate to language endangerment status. The edges were then transformed into unweighted connections to create a simple unweighted, undirected graph for analysis. In order to determine the value of x (i.e., the percentile at which the edge threshold is to be applied), we constructed 10 spatial networks that retained edges with distances below the 1st, 2nd, 3rd… 10th percentile (in increments of 1%) of all distances in the matrix. Additional information of the distances depicted by the edges in each of the 10 networks is provided in Supplementary Information.These 10 networks were then analyzed for their macro- and meso-scale network properties. A summary of macro and meso-scale network measures used in this analysis and their definitions is provided in Table 1, which depicts the 10 networks showing similar patterns in their network structures.Table 1 An overview of macro- and meso-level network measures of spatial networks with different thresholds.Full size tableResultsAs expected, network density and average degree of the networks, which serve as indicators of the number of edges relative to the number of nodes in the network, increased as the edge threshold used to connect nodes became more liberal. The relatively high values of C (i.e., high levels of local clustering among nodes) and low values of ASPL (i.e., relatively short paths despite large size of network) suggested the presence of small world structure30. The community detection analysis using the Louvain method31 indicated strong evidence of community structure in the networks—suggesting the presence of clusters of endangered languages.The point at which the vast majority of nodes was located within the largest connected component of the network occurred at the 5% edge threshold. Because the 5% network was not too fragmented, we report the analyses conducted on the largest connected component of the 5% network in the following subsections. Please see Supplementary Information for additional details behind the rationale for selecting the 5% network for further analyses. The smaller connected components were excluded. Note however that our results are robust across spatial networks of various edge thresholds (due to lack of space, please see Supplementary Information for a complete summary of all reported analyses conducted on all 10 spatial networks).Macro-level analysis: assortative mixing of endangerment statusesMethodTo investigate the macro-level structure of the spatial network of endangered languages, we computed the assortativity coefficient of the spatial network. Specifically, we wanted to know if the endangerment statuses of the languages tended to cluster at the global level of the entire network. If the assortativity coefficient is positive, the languages in the network would tend to be connected to languages of similar levels of endangerment. If the assortativity coefficient is negative, the languages in the network would tend to be connected to languages of dissimilar levels of endangerment.ResultsThere is a significant positive correlation (Spearman’s rank correlation) between the endangerment status of connected pairs of endangered languages in the network, r = 0.20, p  More

Experimental designThe comparative field microcosm experiments were conducted on Laojun Mountain in China (26.6959 N; 99.7759 E) in September–October 2013, and on Balggesvarri Mountain in Norway (69.3809 N; 20.3483 E) in July 2013, designed to be broadly representative of subtropical and subarctic climatic zones, respectively, as first reported in Wang et al.29. In the Laojun Mountain region, mean annual temperatures ranged from 4.2 to 12.9 °C, with July mean temperatures of 17–25 °C. In the Balggesvarri Mountain region, mean annual temperatures ranged from −2.9 to 0.7 °C, with July mean temperatures of 8–16 °C. The experiments were characterised by an aquatic ecosystem with consistent initial DOM composition but different locally colonised microbial communities and newly produced endogenous DOM. While allowing us to minimise the complexity of natural ecosystems, the experiment provided a means for investigating DOM-microbe associations at large spatial scales by controlling the initial DOM supply. Briefly, we selected locations with five different elevations on each mountainside. The elevations were 3822, 3505, 2915, 2580 and 2286 m a.s.l. on Laojun Mountain in China, and 750, 550, 350, 170 and 20 m a.s.l. on Balggesvarri Mountain in Norway. At each elevation, we established 30 aquatic microcosms (1.5 L bottle) composed of 15 g of sterilised lake sediment and 1.2 L of sterilised artificial lake water at one of ten nutrient levels of 0, 0.45, 1.80, 4.05, 7.65, 11.25, 15.75, 21.60, 28.80 and 36.00 mg N L−1 of KNO3 in the overlying water. To compensate for nitrate additions shifting stoichiometric ratios, KH2PO4 was added to the bottles so that the N/P ratio of the initial overlying water was 14.93, which was similar to the annual average ratio in Taihu Lake during 2007 (that is, 14.49). Thus, we use “nutrient enrichment” to indicate a series of targeted nutrient levels of both nitrate and phosphate, the former of which was used to represent nutrient enrichment in the statistical analyses. Each nutrient level was replicated three times. The lake sediments were obtained from the centre of Taihu Lake, China, and were aseptically canned per bottle after autoclaving at 121 °C for 30 min. Nutrient levels for the experiments were selected based on conditions of the eutrophic Taihu Lake, and the highest nitrate concentration was based on the maximum total nitrogen in 2007 (20.79 mg L−1; Fig. S19). We chose the nutrient level of this year because a massive cyanobacteria bloom in Taihu Lake happened in May 2007 and initiated an odorous drinking water crisis in the nearby city of Wuxi.The microcosms were left in the field for one month allowing airborne bacteria to freely colonise the sediments and water. To keep the microbial dispersal events as natural as possible, we did not cover the experimental microcosms in case of rainfall. To avoid or minimize potential influence of extreme nature events, we (i) left the top 20% of each microcosm empty to prevent water from overflowing during heavy rains, and (ii) checked the experimental sites twice during each experimental period, and added sterilized water to obtain a final volume of approximately 1.2 L. The bottom of our microcosm was buried into the local soils by 10% of the bottle height, partly to reduce UV exposure to sediments. More considerations of the experimental design were detailed in the Supplementary Methods. To avoid the effects of daily temperature variation, we measured the water temperature and pH within 2 h before noon at all elevations in the day before the final sample collection. At the end of the experimental period, we aseptically sampled the water and sediments of the 300 bottles (that is, 2 mountains × 5 elevations × 10 nutrient levels × 3 replicates) for the following analyses of physiochemical variables, bacterial community and DOM composition.Physiochemical variables and bacterial communityWe measured environmental variables, namely, the total nitrogen (TN), total phosphorus (TP), dissolved nutrients (that is, NOx−, NO2−, NH4+ and PO43−), total organic carbon (TOC), dissolved organic carbon (DOC) and chlorophyll a (Chl a) in the sediments, and the NO3−, NO2−, NH4+, PO43− and pH in the overlying water (Table S2, Fig. S20), according to Wang et al.29.The sediment bacteria were examined using high-throughput sequencing of 16S rRNA genes. The sequences were processed in QIIME (v1.9)45 and OTUs were defined at 97% sequence similarity. The bacterial sequences were rarefied to 20,000 per sample. Further details on physicochemical and bacterial community analyses are available in Wang et al.29.ESI FT-ICR MS analysis of DOM samplesHighly accurate mass measurements of DOM within the sediment samples were conducted using a 15 Tesla solariX XR system, a ultrahigh-resolution Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR MS, Bruker Daltonics, Billerica, MA) coupled with an electrospray ionization (ESI) interface, as demonstrated previously46 with some modifications. It should be noted that FT-ICR MS does not identify molecules, but only molecular formulae in terms of elemental composition and there can be many molecular structures sharing the same elemental compositions. DOM was solid-phase extracted (SPE) with Agilent VacElut resins before FT-ICR MS measurement47 with minor modifications. Briefly, an aliquot of 0.7 g freeze-dried sediment was sonicated with 30 ml ultrapure water for 2 h, and centrifuged at 5000 × g for 20 min. The extracted water was filtered through the 0.45 μm Millipore filter and further acidified to pH 2 using 1 M HCl. Cartridges were drained, rinsed with ultrapure water and methanol (ULC-MS grade), and conditioned with pH 2 ultrapure water. Calculated volumes of extracts were slowly passed through cartridges based on DOC concentration. Cartridges were rinsed with pH 2 ultrapure water and dried with N2 gas. Samples were finally eluted with methanol into precombusted amber glass vials, dried with N2 gas and stored at −20 °C until DOM analysis. The extracts were continuously injected into the standard ESI source with a flow rate of 2 μl min−1 and an ESI capillary voltage of 3.5 kV in negative ion mode. One hundred single scans with a transient size of 4 mega word (MW) data points, an ion accumulation time of 0.3 s, and within the mass range of m/z 150–1200, were co-added to a spectrum with absorption mode for phase correction, thereby resulting in a resolving power of 750,000 (FWHM at m/z 400). All FT-ICR mass spectra were internally calibrated using organic matter homologous series separated by 14 Da (-CH2 groups). The mass measurement accuracy was typically within 1 ppm for singly charged ions across a broad m/z range (150–1200 m/z).Data Analysis software (BrukerDaltonik v4.2) was used to convert raw spectra to a list of m/z values using FT-MS peak picker with a signal-to-noise ratio (S/N) threshold set to 7 and absolute intensity threshold to the default value of 100. Putative chemical formulae were assigned using the software Formularity (v1.0)48 following the Compound Identification Algorithm49. In total, 19,538 molecular formulas were putatively assigned for all samples (n = 300) based on the following criteria: S/N  > 7, and mass measurement error  0.80, P ≤ 0.001; Fig. S9). Similar conclusions were also obtained with either OTUs or genera when relating the pairwise distances of molecular traits with SparCC correlation coefficient ρ values among DOM molecules in Fig. 4c. To reduce type I errors in the correlation calculations created by low-occurrence genera or molecules, the majority rule was applied; that is, we retained genera or molecules that were observed in more than half of the total samples (≥75 samples) in China or Norway. The filtered table, including 1340 and 1246 DOM molecules, and 75 and 49 bacterial genera in China and Norway, respectively, was then used for pairwise correlation calculation of DOM and bacteria using SparCC with default parameters35.Finally, bipartite network analysis at a molecular level was performed to quantify the specialization of DOM-bacteria networks (Box 1). The specialization considers interaction abundance and is standardised to account for heterogeneity in the interaction strength and species richness, which describes the levels of “vulnerability” of DOM molecules and “generality” of bacterial taxa27. The threshold correlation for inclusion in bipartite networks was |ρ| = 0.30 to exclude weak interactions and we retained the adjacent matrix with only the interactions between DOM and bacteria. We then constructed two types of interaction networks (i.e., negative and positive networks) based on negative and positive correlation coefficients (SparCC ρ ≤ −0.30 and ρ ≥ 0.30, respectively). According to resource-consumer relationships, negative networks likely indicate the degradation of larger molecules into smaller structures, while positive networks may suggest the production of new molecules via degradation or biosynthetic processes. The SparCC ρ values were multiplied by 10,000 and rounded to integers, and the absolute values were taken for negative networks to enable the calculations of specialization indices. A separate negative and positive sub-network was obtained for each microcosm by selecting the DOM molecules and bacterial taxa in each sample based on its bacterial and DOM compositions. For the network level analysis, we calculated H2′, a measure of specialization27, for each network:$${H}_{2}=-mathop{sum }limits_{i{{mbox{=}}}1}^{i}mathop{sum }limits_{j{{mbox{=}}}1}^{j}{{mbox{(}}}{{{mbox{p}}}}_{{ij}}{{{{{{rm{ln}}}}}}}{{{mbox{p}}}}_{{ij}}{{mbox{)}}}$$
(2)
$${H}_{2}{prime} =frac{{H}_{2{max }}{-}{H}_{2}}{{H}_{2{max }}{-}{H}_{2{min }}}$$
(3)
where ({{{mbox{p}}}}_{{ij}}{{mbox{=}}}{{{mbox{a}}}}_{{ij}}{{mbox{/}}}m), represents the proportion of interactions in a i × j matrix. ({{{mbox{a}}}}_{{ij}}) is the number of interactions between DOM molecule i and bacterial genus j, which is also referred as “link weight”. m is the total number of interactions between all DOM molecules and bacterial genera. H2′ is the standardised H2 against the minimum (H2min) and maximum (H2max) possible for the same distribution of interaction totals. For the molecular level analysis, we calculated the specialization index Kullback–Leibler distance (d′) for DOM molecules (di′) and bacterial genera (dj′), which describes the levels of “vulnerability” of DOM molecules and “generality” of bacterial genera, respectively:$${d}_{i}=mathop{sum }limits_{j=1}^{j}left(frac{{{{mbox{a}}}}_{{ij}}}{{{{mbox{A}}}}_{i}}{{{mbox{ln}}}}frac{{{{mbox{a}}}}_{{ij}}m}{{{{mbox{A}}}}_{i}{{{mbox{A}}}}_{j}}right)$$
(4)
$${d}_{i}{prime} =frac{{d}_{i}-{d}_{{min }}}{{d}_{{max }}-{d}_{{min }}}$$
(5)
where ({A}_{i}) = (mathop{sum }limits_{j{{mbox{=}}}1}^{j}{{{mbox{a}}}}_{{ij}}) and ({A}_{j}) = (mathop{sum }limits_{i{{mbox{=}}}1}^{i}{{{mbox{a}}}}_{{ij}}), are the total number of interactions of DOM molecule i and bacterial genus j, respectively. di′ is the standardised di against the minimum (dmin) and maximum (dmax) possible for the same distribution of interaction totals. The equations of dj′ are analogous to di′, replacing j by i. Weighted means of d′ for DOM were calculated for each network as the sum of the product of d′ for each individual molecule i (di′) and relative intensity Ii divided by the sum of all intensities d′  = Ʃ(di′ × Ii)/Ʃ(Ii). Weighted means of d′ for bacteria were calculated as the sum of the d′ of each individual bacterial genus j (dj′) and relative abundance of bacterial genus Ij divided by the sum of all abundance. All calculations were performed using the R package FD V1.0.12. The observed H2′ and d′ values ranged from 0 (complete generalization) to 1 (complete specialization)28 (Fig. S21). Specifically, elevated H2′ or d′ values indicate a high degree of specialization, while lower values suggest increased generalization, that is, higher vulnerability of DOM and/or higher generality of microbes. To directly compare the network indices across the elevations or nutrient enrichment levels, we used a null modelling approach. We standardised the three observed specialization indices (Sobserved; that is, H2′, d′ of DOM, and d′ of bacteria) by calculating their z-scores63 using the equation:$${z}_{S}=({S}_{{{{{{rm{observed}}}}}}}-overline{{{S}}_{{{{{{rm{null}}}}}}}})/({sigma }_{S_{{{{{rm{null}}}}}}})$$
(6)
where (overline{{{S}}_{{{{{{rm{null}}}}}}}}) and ({sigma }_{S_{{{{{rm{null}}}}}}}) were, respectively, the mean and standard deviation of the null distribution of S (Snull). One hundred randomised null networks were generated for each bipartite network to derive Snull using the swap.web algorithm, which keeps species richness and the number of interactions per species constant along with network connectance. This null model analysis indicates that interactions between DOM and bacteria were non-random as the observed network specialization indices were generally significantly lower than expected by chance (P  0.05), which tests whether the model structure differs from the observed data, high comparative fit index (CFI  > 0.95) and low standardised root mean squared residual (SRMR  More

Wellnhofer, P. The Illustrated Encyclopedia of Pterosaurs (Crescent Books, 1991).Unwin, D. M. The pterosaurs from deep time (Pi Press, 2005).Witton, M. P. Pterosaurs: Natural History (Anatomy (Princeton University Press, 2013).Book 

Google Scholar 
Williams, C. J. et al. Helically arranged cross struts in azhdarchid pterosaur cervical vertebrae and their biomechanical implications. iScience 24, 102338 (2021).PubMed 
PubMed Central 
Article 

Google Scholar 
Bestwick, J., Unwin, D. M., Butler, R. J. & Purnell, M. A. Dietary diversity and evolution of the earliest flying vertebrates revealed by dental microwear texture analysis. Nat. Commun. 11, 5293 (2020).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Ryang, W. H. Characteristics of strike-slip basin formation and sedimentary fills and the Cretaceous small basins of the Korean Peninsula. J. Geo. Soc. Korea 49, 31–45 (2013).CAS 

Google Scholar 
Kim, B. G. & Park, B. G. Geological report of the Dongbok sheet (1:50,000) (Geological Survey of Korea, Seoul, 1966).Lee, H., Sim, M. S. & Choi, T. Stratigraphic evolution of the northern part of the Cretaceous Neungju basin South Korea. Geosci. J. 23, 849–865 (2019).CAS 
Article 

Google Scholar 
Paik, I. S., Huh, M., So, Y. H., Lee, J. E. & Kim, H. J. Traces of evaporites in Upper Cretaceous lacustrine deposits of Korea: Origin and paleoenvironmental implications. J. Asian Earth Sci. 30, 93–107 (2007).Article 

Google Scholar 
Cohen, K. M., Finney, S. M., Gibbard, P. L. & Fan, J.-X. The ICS international Chronostratigraphic chart. Episodes 36, 199–204 (2013).Article 

Google Scholar 
Calvo, J. O. & Lockley, M. G. The first pterosaur tracks from Gondwana. Cretac. Res. 22, 585–590 (2001).Article 

Google Scholar 
Kukihara, R. & Lockley, M. G. Fossil footprints from the dakota group (Cretaceous) john martin reservoir, bent county, Colorado: New insights into the paleoecology of the Dinosaur freeway. Cretac. Res. 33, 165–182 (2012).Article 

Google Scholar 
Lockley, M. & Schumacher, B. A new pterosaur swim tracks locality from the Cretaceous Dakota Group of eastern Colorado: implications for pterosaur swim track behavior. Fossil Footprints of Western North America. Bull. NM Mus. Nat. Hist. Sci, 365–371 (2014).Smith, R. E., Martill, D. M., Unwin, D. M. & Steel, L. Edentulous pterosaurs from the Cambridge Greensand (Cretaceous) of eastern England with a review of Ornithostoma Seeley, 1871. Proc. Geol. Assoc. (2020).Ibrahim, N., Unwin, D. M., Martill, D. M., Baidder, L. & Zouhri, S. A new pterosaur (Pterodactyloidea: Azhdarchidae) from the Upper Cretaceous of Morocco. PLoS ONE 5, e10875 (2010).PubMed 
PubMed Central 
Article 

Google Scholar 
Martill, D. M. & Ibrahim, N. An unusual modification of the jaws in cf. Alanqa, a mid-Cretaceous azhdarchid pterosaur from the Kem Kem beds of Morocco. Cretac. Res. 53, 59–67 (2015).Article 

Google Scholar 
Jacobs, M. L., Martill, D. M., Ibrahim, N. & Longrich, N. A new species of Coloborhynchus (Pterosauria, Ornithocheiridae) from the mid-Cretaceous of North Africa. Cretac. Res. 95, 77–88 (2019).Article 

Google Scholar 
Jacobs, M. L. et al. New toothed pterosaurs (Pterosauria: Ornithocheiridae) from the middle Cretaceous Kem Kem beds of Morocco and implications for pterosaur palaeobiogeography and diversity. Cretac. Res. 110, 104413 (2020).Article 

Google Scholar 
McPhee, J. et al. A new ? Chaoyangopterid (Pterosauria: Pterodactyloidea) from the Cretaceous Kem Kem beds of southern Morocco. Cretac. Res. 110, 104410 (2020).Article 

Google Scholar 
Martill, D. M. et al. A new tapejarid (Pterosauria, Azhdarchoidea) from the mid-Cretaceous Kem Kem beds of Takmout, southern Morocco. Cretac. Res. 112, 104424 (2020).Article 

Google Scholar 
Martill, D. M., Unwin, D. M., Ibrahim, N. & Longrich, N. A new edentulous pterosaur from the Cretaceous Kem Kem beds of south eastern Morocco. Cretac. Res. 84, 1–12 (2018).Article 

Google Scholar 
Smith, R. E. et al. Small, immature pterosaurs from the Cretaceous of Africa: implications for taphonomic bias and palaeocommunity structure in flying reptiles. Cretac. Res. 130, 105061 (2022).Article 

Google Scholar 
Smith, R. E., Martill, D. M., Kao, A., Zouhri, S. & Longrich, N. A long-billed, possible probe-feeding pterosaur (Pterodactyloidea: ?Azhdarchoidea) from the mid-Cretaceous of Morocco North Africa. Cretac. Res. 118, 104643 (2021).Article 

Google Scholar 
Kellner, A. W. A. et al. First complete pterosaur from the Afro-Arabian continent: insight into pterodactyloid diversity. Sci. Rep. 9, 17875 (2019).PubMed 
PubMed Central 
Article 

Google Scholar 
Elgin, R. A. & Frey, E. A new azhdarchoid pterosaur from the Cenomanian (Late Cretaceous) of Lebanon. Swiss J. Geosci. 104, 21–33 (2011).Article 

Google Scholar 
Averianov, A. O., Kurochkin, E. N., Pervushov, E. M. & Ivanov, A. V. Two bone fragments of ornithocheiroid pterosaurs from the Cenomanian of Volgograd Region, southern Russia. Acta Palaeontol. Pol. 50 (2005).Averianov, A. & Kurochkin, E. A new pterosaurian record from the Cenomanian of the Volga region. Paleontol. J. 44, 695–697 (2010).Article 

Google Scholar 
Nessov, L. Flying reptiles from the Jurassic and cretaceous of the USSR and significance of their remains for the reconstruction of paleogeographical conditions. Vestn. Leningr. Gos. Univ. Ser. 7, 28 (1990).
Google Scholar 
Bakhurina, N. N. & Unwin, D. M. A survey of pterosaurs from the Jurassic and Cretaceous of the former Soviet Union and Mongolia. (1995).Averianov, A. O. New records of azhdarchids (Pterosauria, Azhdarchidae) from the Late Cretaceous of Russia, Kazakhstan, and Central Asia. Paleontol. J. 41, 189–197 (2007).Article 

Google Scholar 
Averianov, A. Mid-Cretaceous ornithocheirids (Pterosauria, Ornithocheiridae) from Russia and Uzbekistan. Paleontol. J. 41, 79–86 (2007).Article 

Google Scholar 
Huh, M., Paik, I. S., Chung, C. H., Hwang, K. G. & Kim, B. S. Theropod tracks from Seoyuri in Hwasun, Jeollanamdo, Korea: occurrence and paleontological significance. J. Geo. Soc. Korea 39, 461–478 (2003).CAS 

Google Scholar 
Huh, M. et al. Well-preserved theropod tracks from the Upper Cretaceous of Hwasun County, southwestern South Korea, and their paleobiological implications. Cretac. Res. 27, 123–138 (2006).Article 

Google Scholar 
Lockley, M. G., Huh, M. & Kim, B. S. Ornithopodichnus and pes-only sauropod Trackways from the Hwasun tracksite Cretaceous of Korea. Ichnos 19, 93–100 (2012).Article 

Google Scholar 
Hwang, K. G., Huh, M. & Paik, I. S. A unique trackway of small theropod from Seoyu-ri, Hwasun-gun Jeollanam province. J. Geo. Soc. Korea 42, 69–78 (2006).CAS 

Google Scholar 
Kim, B. S. & Huh, M. Analysis of the acceleration phase of a theropod dinosaur based on a Cretaceous trackway from Korea. Palaeogeogr. Palaeoclimatol. Palaeoecol. 293, 1–8 (2010).Article 

Google Scholar 
Marchetti, L. et al. Defining the morphological quality of fossil footprints. Problems and principles of preservation in tetrapod ichnology with examples from the Palaeozoic to the present. Earth-Sci. Rev. 193, 109–145 (2019).Article 

Google Scholar 
Rodríguez-de La Rosa, R. A. Pterosaur tracks from the latest Campanian Cerro del Pueblo formation of southeastern Coahuila. Mexico. Geol. Soc. Spec. Publ. 271, 275–282 (2003).Article 

Google Scholar 
Lockley, M. G. & Meyer, C. Crocodylomorph trackways from the Jurassic to early cretaceous of North America and Europe: Implications for Ichnotaxonomy. Ichnos 11, 167–178 (2004).Article 

Google Scholar 
Ambroggi, R. & De Lapparent, A. Les empreintes de pas fossiles du Maestrichtien d’Agadir. Notes du Service Géologique du Maroc 10, 43–57 (1954).
Google Scholar 
Stokes, W. L. Pterodactyl tracks from the Morrison Formation. J. Paleontol. 31, 952–954 (1957).
Google Scholar 
Delair, J. Note on Purbeck fossil footprints, with descriptions of two hitherto unknown forms from Dorset. Proceedings of the Dorset Natural History and Archaeological Society. 92–100 (1963).Hwang, K.-G., Huh, M. I. N., Lockley, M. G., Unwin, D. M. & Wright, J. L. New pterosaur tracks (Pteraichnidae) from the Late Cretaceous Uhangri Formation, southwestern Korea. Geol. Mag. 139, 421–435 (2002).Article 

Google Scholar 
Mazin, J.-M. & Pouech, J. The first non-pterodactyloid pterosaurian trackways and the terrestrial ability of non-pterodactyloid pterosaurs. Geobios 58, 39–53 (2020).Article 

Google Scholar 
Masrour, M., de Ducla, M., Billon-Bruyat, J.-P. & Mazin, J.-M. Rediscovery of the Tagragra tracksite (Maastrichtian, Agadir, Morocco): Agadirichnus elegans Ambroggi and Lapparent 1954 is Pterosaurian Ichnotaxon. Ichnos 25, 285–294 (2018).Article 

Google Scholar 
Wright, J. L., Unwin, D. M., Lockley, M. G. & Rainforth, E. C. Pterosaur tracks from the Purbeck limestone formation of Dorset England. Proc. Geol. Assoc. 108, 39–48 (1997).Article 

Google Scholar 
Lockley, M. G. et al. The fossil trackway Pteraichnusis pterosaurian, not crocodilian: Implications for the global distribution of pterosaur tracks. Ichnos 4, 7–20 (1995).Article 

Google Scholar 
Billon-Bruyat, J.-P. & Mazin, J.-M. The systematic problem of tetrapod ichnotaxa: the case study of Pteraichnus Stokes, 1957 (Pterosauria, Pterodactyloidae). Geol. Soc. Spec. Publ. 217, 315–324 (2003).Article 

Google Scholar 
Pascual Arribas, C. & Sanz Pérez, E. Huellas de Pterosaurios en el grupo Oncala (Soria, España). Pteraichnus palaciei-saenzi, nov. icnosp. Estudios Geol. 56, 73–100 (2000).
Google Scholar 
Calvo, M. M., Vidarte, C. F., Fuentes, F. M. & Fuentes, M. M. Huellas de Pterosaurios en la Sierra de Oncala (Soria, España). Nuevas icnoespecies: pteraichnus vetustior, Pteraichnus parvus. Pteraichnus manueli. Celtiberia 54, 471–490 (2004).
Google Scholar 
Fuentes Vidarte, C., Meijide Calvo, M., Meijide Fuentes, F. & Meijide Fuentes, M. Pteraichnus longipodus nov. icnosp. en la Sierra de Oncala (Soria, España). Studia Geologica Salmanticensia, 103–114 (2004).Peng, B.-X., Du, Y.-S., Li, D.-Q. & Bai, Z.-C. The first discovery of the early Cretaceous Pterosaur track and its significance in Yanguoxia, Yongjing County, Gansu Province. Earth Sci.-J. China Univ. Geosci. 29, 21–24 (2004).
Google Scholar 
Lee, Y.-N., Lee, H.-J., Lü, J. & Kobayashi, Y. New pterosaur tracks from the Hasandong formation (Lower Cretaceous) of Hadong County South Korea. Cretac. Res. 29, 345–353 (2008).Article 

Google Scholar 
Lee, Y.-N., Azuma, Y., Lee, H.-J., Shibata, M. & Lü, J. The first pterosaur trackways from Japan. Cretac. Res. 31, 263–273 (2010).Article 

Google Scholar 
Chen, R. et al. Pterosaur tracks from the early late cretaceous of Dongyang City, Zhejiang Province China. Geol. Bull. China. 32, 693–698 (2013).CAS 

Google Scholar 
Li, Y., Wang, X. & Jiang, S. A new pterosaur tracksite from the Lower Cretaceous of Wuerho, Junggar Basin, China: inferring the first putative pterosaur trackmaker. PeerJ 9, e11361 (2021).PubMed 
PubMed Central 
Article 

Google Scholar 
Ha, S. et al. Diminutive pterosaur tracks and trackways (Pteraichnus gracilis ichnosp. Nov.) from the lower Cretaceous Jinju formation, Gyeongsang basin. Korea. Cretac. Res. 131, 105080 (2021).Article 

Google Scholar 
Sánchez-Hernández, B., Przewieslik, A. G. & Benton, M. J. A reassessment of the Pteraichnus ichnospecies from the early Cretaceous of Soria Province Spain. J. Vertebr. Paleontol. 29, 487–497 (2009).Article 

Google Scholar 
Zhou, X. et al. A new darwinopteran pterosaur reveals arborealism and an opposed thumb. Curr. Biol. 31, 2429-2436.e2427 (2021).CAS 
PubMed 
Article 

Google Scholar 
Lü, J. et al. Dragons of the Skies (recent advances on the study of pterosaurs from China) (Zhejiang Science and Technology Press, 2013).
Google Scholar 
Beccari, V. et al. Osteology of an exceptionally well-preserved tapejarid skeleton from Brazil: Revealing the anatomy of a curious pterodactyloid clade. PLoS ONE 16, e0254789 (2021).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Lü, J. A new boreopterid pterodactyloid pterosaur from the Early Cretaceous Yixian Formation of Liaoning Province, northeastern China. Acta Geologica Sinica-English Edition 84, 241–246 (2010).Article 

Google Scholar 
Bennett, S. C. Terrestrial locomotion of pterosaurs: A reconstruction based on Pteraichnus trackways. J. Vertebr. Paleontol. 17, 104–113 (2010).Article 

Google Scholar 
Wang, X. & Lü, J. Discovery of a pterodactylid pterosaur from the Yixian Formation of western Liaoning China. Chin. Sci. Bull. 46, A3–A8 (2001).Article 

Google Scholar 
Frey, E. et al. A new specimen of nyctosaurid pterosaur, cf. Muzquizopteryx sp. from the Late Cretaceous of northeast Mexico. Revista mexicana de ciencias geológicas 29, 131–139 (2012).
Google Scholar 
Wu, W.-H., Zhou, C.-F. & Andres, B. The toothless pterosaur Jidapterus edentus (Pterodactyloidea: Azhdarchoidea) from the Early Cretaceous Jehol Biota and its paleoecological implications. PLoS ONE 12, e0185486 (2017).PubMed 
PubMed Central 
Article 

Google Scholar 
Lü, J. et al. The toothless pterosaurs from China. Acta Geol. Sin. 90, 2513–2525 (2016).
Google Scholar 
Zhang, X., Jiang, S., Cheng, X. & Wang, X. New Material of Sinopterus (Pterosauria, Tapejaridae) from the Early Cretaceous Jehol Biota of China. An. Acad. Bras. Cienc. 91 (2019).Bestwick, J., Unwin, D. M., Butler, R. J., Henderson, D. M. & Purnell, M. A. Pterosaur dietary hypotheses: A review of ideas and approaches. Biol. Rev. 93, 2021–2048 (2018).PubMed 
Article 

Google Scholar 
Chen, H. et al. New anatomical information on Dsungaripterus weii Young, 1964 with focus on the palatal region. PeerJ 8, e8741 (2020).PubMed 
PubMed Central 
Article 

Google Scholar 
Li, D. et al. A manus dominated pterosaur track assemblage from Gansu, China: Implications for behavior. Sci. Bull. 60, 264–272 (2015).Article 

Google Scholar 
Masrour, M., Pascual-Arribas, C., de Ducla, M., Hernández-Medrano, N. & Pérez-Lorente, F. Anza palaeoichnological site. Late Cretaceous. Morocco. Part I. The first African pterosaur trackway (manus only). J. African Earth Sci. 134, 766–775 (2017).Article 

Google Scholar 
Bramwell, C. D. & Whitfield, G. R. Biomechanics of Pteranodon. Phil. Trans. R. Soc. Lond. B. 267, 503–581 (1974).Article 

Google Scholar 
Bennett, S. C. Terrestrial locomotion of pterosaurs: a reconstruction based on Pteraichnus trackways. J. Vertebr. Paleontol. 17, 104–113 (1997).Article 

Google Scholar 
Mazin, J.-M., Billon-Bruyat, J.-P., Hantzpergue, P. & Lafaurie, G. Ichnological evidence for quadrupedal locomotion in pterodactyloid pterosaurs: Trackways from the Late Jurassic of Crayssac (southwestern France). Geol. Soc. Spec. Publ. 217, 283–296 (2003).Article 

Google Scholar 
Henderson, D. M. Pterosaur body mass estimates from three-dimensional mathematical slicing. J. Vertebr. Paleontol. 30, 768–785 (2010).Article 

Google Scholar 
Lockley, M. G. & Wright, J. L. Pterosaur swim tracks and other ichnological evidnce of behaviour and ecology. Geol. Soc. Spec. Publ. 217, 297–313 (2003).Article 

Google Scholar 
Lockley, M., Mitchell, L. & Odier, G. P. Small Theropod track assemblages from middle Jurassic Eolianites of eastern Utah: Paleoecological insights from dune Ichnofacies in a transgressive sequence. Ichnos 14, 131–142 (2007).Article 

Google Scholar 
Fiorillo, A. R., Hasiotis, S. T., Kobayashi, Y. & Tomsich, C. S. A pterosaur manus track from Denali National park, Alaska Range, Alaska United States. Palaios 24, 466–472 (2009).Article 

Google Scholar 
Bell, P. R., Fanti, F. & Sissons, R. A possible pterosaur manus track from the late Cretaceous of Alberta. Lethaia 46, 274–279 (2013).Article 

Google Scholar 
Stinnesbeck, W. et al. Theropod, avian, pterosaur, and arthropod tracks from the uppermost Cretaceous Las Encinas Formation, Coahuila, northeastern Mexico, and their significance for the end-Cretaceous mass extinction. Geol. Soc. Am. Bull. 129, 331–348 (2017).Article 

Google Scholar 
Xing, L. et al. Late Cretaceous ornithopod-dominated, theropod, and pterosaur track assemblages from the Nanxiong Basin, China: New discoveries, ichnotaxonomy, and paleoecology. Palaeogeogr. Palaeoclimatol. Palaeoecol. 466, 303–313 (2017).Article 

Google Scholar 
Lockley, M. G., Gierlinski, G. D., Adach, L., Schumacher, B. & Cart, K. Newly discovered tetrapod ichnotaxa from the Upper Blackhawk Formation Utah. Bull. N. M. M. Nat. Hist. Sci. 79, 469–480 (2018).
Google Scholar 
Lockley, M. G. & Gillette, D. Pterosaur and bird tracks from a new Late Cretaceous locality in Utah. Verteb. Paleontol. Utah 99, 355–359 (1999).
Google Scholar 
Bennett, S. C. The ontogeny of Pteranodon and other pterosaurs. Paleobiology 19, 92–106 (1993).Article 

Google Scholar 
Bennett, S. C. Year-classes of pterosaurs from the Solnhofen Limestone of Germany: taxonomic and systematic implications. J. Vertebr. Paleontol. 16, 432–444 (1996).Article 

Google Scholar 
Chiappe, L. M., Codorniú, L., Grellet-Tinner, G. & Rivarola, D. Argentinian unhatched pterosaur fossil. Nature 432, 571–572 (2004).CAS 
PubMed 
Article 

Google Scholar 
Codorniú, L., Chiappe, L. & Rivarola, D. Neonate morphology and development in pterosaurs: evidence from a Ctenochasmatid embryo from the Early Cretaceous of Argentina. Geol. Soc. Spec. Publ. 455, 83–94 (2018).Article 

Google Scholar 
Mickelson, D. L., Lockley, M. G., Bishop, J. & Kirkland, J. A New Pterosaur Tracksite from the Jurassic Summerville Formation, near Ferron Utah. Ichnos 11, 125–142 (2004).Article 

Google Scholar  More

Model 1: Evolution of multiple mating and mate monopolisation under ancestral monogamyIn all models, I assume a large population with a 1:1 sex ratio. I begin with what is possibly the simplest model set-up for deriving Bateman functions in a scenario that is completely symmetrical aside from gamete number. Assume a monogamous, externally fertilising population where parents pair up and release their gametes into a nest. That is, every individual in the initial population participates in exactly one fertilisation event (the equivalent of a mating). Now consider a mutant individual that can attract multiple mates of the opposite type to release gametes into its nest, with no competition from other individuals of its own type. This simple set-up avoids asymmetries arising from internal fertilisation, and the complication of direct gamete competition for the multiply mating mutant individual (which is examined in Models 2–3), placing focus directly on the core of the problem: the asymmetry arising in fertilisation from imbalanced gamete numbers. All gametes are released in one burst by all individuals, but the focal individual may achieve ‘multiple matings’ simply by monopolising multiple mates at its nest. The reproductive success of the focal individual is then equivalent to the number of fertilisations that take place in that nest. Our aim is to understand how the reproductive success of an individual deviating from the monogamous population strategy and instead mating with (hat{m}) individuals of the opposite type is altered. A strong positive relationship between (hat{m}) and reproductive success then indicates a steep Bateman gradient. If Bateman’s assertion is correct, the resulting gradient should be steeper for the type that produces the larger number of gametes. Note that there is a game-theoretical25 flavour to this setting, where the focus is on the fitness of a rare mutant in a population with a fixed resident strategy.The two types are labelled with x and y, which could correspond to the two sexes, depending on what gamete numbers are assigned to them. The number of gametes produced by a single individual is labelled nx and ny, and the total number of gametes in a nest (or more generally, a fertilisation arena which could be internal or external) is labelled with Nx and Ny. To compute the number of fertilisations in a nest with a total of Nx and Ny gametes, I use a fertilisation function first derived by Togashi et al.24 purely from biophysical principles, treating the two gamete types symmetrically, with no pre-existing assumptions about differences between females and males or their gametes (for a broader context and comparison to other functions, see Table 1 and function F7 in19). Any sex-specific differences arise only retrospectively after different gamete numbers are assigned to x and y of which either one could be male or female. The fertilisation function is (fleft({N}_{x},{N}_{y}right)={N}_{x}{N}_{y}frac{{e}^{a{N}_{x}}-{e}^{a{N}_{y}}}{{{N}_{x}e}^{a{N}_{x}}-{N}_{y}{e}^{a{N}_{y}}}), where a is a parameter controlling fertilisation efficiency (for the special case Nx = Ny the function is defined as (fleft({N}_{x},{N}_{y}right)=frac{a{N}_{x}^{2}}{1+a{N}_{x}})19,24, which is also the limit of f when Ny → Nx).In a monogamous resident pair, we have simply Nx = nx and Ny = ny. But if a mutant individual of type x is able to attract (hat{m}) fertilisation partners of type y, then for that individual ({N}_{y}=hat{m}{n}_{y}), and the corresponding Bateman function is$${b}_{x}left(hat{m}right)=fleft({N}_{x},{N}_{y}right)=fleft({n}_{x},hat{m}{n}_{y}right)$$
(1)
where the fertilisation function f is as described above. Because of symmetry, the corresponding function for y is found simply by swapping x and y. This function can reproduce the characteristic Bateman gradient asymmetry as gamete numbers diverge (progressing from isogamy to anisogamy in Fig. 1), showing how Bateman’s assertion follows from biophysical effects that arise from unequal numbers of fusing particles: the fertilisation function f is derived solely from such biophysical effects, not from any sex-specific assumptions. Equation (1) makes no reference to sexes, and they only become specified when values are assigned to nx and ny. For example, if nx = 10 and ny = 10,000, the female Bateman function is ({b}_{x}left(hat{m}right)) and the male Bateman function ({b}_{y}left(hat{m}right)), where for the latter all xs in Eq. (1) are replaced with ys and vice versa. The labels x and y are truly just labels. While there are inevitably assumptions built into the equations, crucially we can be certain there are no sex-specific assumptions. Yet the typical shapes reminiscent of Bateman gradients arise from the model when different values are specified for nx and ny (Fig. 1).Fig. 1: The Bateman function of Eq. (1).This figure shows how the basic Bateman gradient asymmetry arises from simple biophysics and mathematics of fertilisation. The population is monogamous aside from a mutant individual, whose number of fertilisation partners (‘matings’) varies on the horizontal axes within panels. a–d show the effect of variation in sex-specific gamete numbers under efficient fertilisation, while e–h show the effect of variation in sex-specific gamete numbers under inefficient fertilisation. Parameter values used are shown in the figure. Females (gamete number nx) are indicated by blue crosses and connecting lines, while males (gamete number ny) are indicated by black dots and connecting lines. Under isogamy, females and males are undefined, and the two colours overlap. The typical sex-specific shapes of Bateman gradients arise from a single equation (which itself is not sex-specific) when a difference in gamete numbers is assigned to nx and ny, confirming Bateman’s intuition that the primary cause of the difference in selection is that females produce fewer gametes than males. The entire range of gamete number ratios presented in the figure is observed in nature, from equal gamete size in many unicellular organisms39 to vertebrates, where sperm count per ejaculate can commonly exceed 109 (see ref. 40 and Supplementary Information therein).Full size imageGamete limitation changes the results quantitatively so that under conditions of poor fertilisation efficiency a larger imbalance in gamete numbers is needed for Bateman gradients to diverge to a similar extent. However, even under inefficient fertilisation, the Bateman gradients do not reverse.Model 2: An external fertiliser model with population-level polygamy and gamete competitionModel 1 presented the simplest possible scenario, where all individuals except a rare mutant mate only once, and gamete competition (sperm competition26, but without assigning either gamete type to be sperm) was thus excluded for the focal mutant individual. Now I generalise from this to a situation that remains entirely symmetrical, but where the resident number of matings can take on any value, and then derive the Bateman function for a rare mutant that deviates from this population-level value. This set-up allows for gamete competition for the focal mutant individual, a crucial addition because of the empirical and theoretical importance of sperm competition26, as well as earlier theory suggesting that polyandry decreases the sex difference in Bateman gradients2.The biological set-up is such that there is a large population and a large number of patches (fertilisation arenas) where multiple individuals of both sexes can release their gametes for fertilisation. After all individuals have released their gametes, those in each patch mix freely and fertilisations take place randomly. Set up in this way, the model is again identical from the perspective of both sexes, and gamete number can be isolated as the sole possible causal factor in any subsequent differences that may arise, extending from the initially monogamous and gamete competition-free scenario of Model 1. All individuals of both sexes are assumed to initially have the same strategy: to divide their nx or ny gametes equally between m patches, and distribute themselves in such a way that gametes from m individuals of each type release gametes into each patch (the number of individuals of each sex per patch need not necessarily be strictly equal to m, but this is the simplest assumption to account for the fact that gamete competition tends to increase with multiple ‘matings’). Now, if a rare x mutant divides its gametes evenly into (hat{m}) randomly selected patches, its gamete number per patch and consequently competitiveness in each patch is altered. Therefore, gametes of a mutant of type x will gain, on average, a fraction ({c}_{x}=left({n}_{x}/hat{m}right)/{N}_{x}) of the fertilisations in that patch, where ({N}_{x}={n}_{x}/hat{m}+(m-1){n}_{x}/m). To compute the number of realised fertilisations in a patch, I use the same fertilisation function as in Model 1, where the mutant number of gametes in a patch is Nx as above and the number of gametes of the opposite type is ({N}_{y}=mfrac{{n}_{y}}{m}={n}_{y}). All the components are now in place to write down the Bateman function corresponding to this scenario, for a mutant of type x:$${b}_{x}left(hat{m},mright)=hat{m}{c}_{x}fleft({N}_{x},{N}_{y}right)$$
(2)
where cx, Nx and Ny are as defined above, and the fertilisation function f is as in Model 1. For completeness, define bx(0, m) = 0, which is necessarily true, but useful to define separately because division by 0 renders Eq. (2) formally undefined when (hat{m}=0).As in Model 1, Eq. (2) makes no reference to sexes, and they only become specified when values are assigned to nx and ny (Fig. 2).Fig. 2: The Bateman function of Eq. (2) for an externally fertilising population with potential for population-wide polygamy and gamete competition.Results are shown for two values of resident matings (m = 1 and m = 2). a–h show the effect of variation in sex-specific gamete numbers and in fertilisation efficiency with m = 1, while i–p show the same with m = 2. Parameter values used are shown in the figure. The value m = 2 is used here because it is comparable to the mean number of matings in Bateman’s1 work (see Fig. 3 for corresponding results with internal fertilisation, but note that the aim of the models is not to quantitatively reproduce Bateman’s results). Females (gamete number nx) are indicated by blue crosses and connecting lines, while males (gamete number ny) are indicated by black dots and connecting lines. Under isogamy, females and males are undefined, and the two colours overlap. Further variation in m is examined in Fig. 4.Full size imageModel 3: An internal fertiliser modelModels 1–2 were set up with the central aim of full symmetry and exclusion of any sex-specific assumptions. Internal fertilisation breaks this symmetry by introducing a sex-specific assumption other than gamete number. Bateman gradients are, however, most commonly applied to situations with internal fertilisation where females are gamete recipients and males are gamete donors27. I therefore construct a model accounting for internal fertilisation. Where Eqs. (1) and (2) allowed no sex differences aside from gamete number, here I additionally consider the fact that females receive gametes while males donate them.As in model 2, there is a very large population, and I assume that in the resident population, all females and males mate exactly m times. It is then considered how a rare mutant individual’s (of either sex) fitness depends on its number of matings (hat{m}).I use the same fertilisation function as in Models 1-2. Consider first the female perspective (labelled with x). A female produces nx gametes and retains them internally. Each female mates with m males, who also mate with m females, dividing their gametes evenly over these matings. Therefore a mutant female receives (hat{m}frac{{n}_{y}}{m}) male gametes, and her reproductive success is$${b}_{x}left(hat{m},mright)=fleft({n}_{x},hat{m}frac{{n}_{y}}{m}right)$$
(3)
A mutant male, on the other hand, mates with (hat{m}) females, each of which mates with m−1 additional males. Therefore, the mutant male’s mating partners will receive a total of ({{N}_{y}=n}_{y}/hat{m}+(m-1){n}_{y}/{m}) male gametes. Thus, the mutant male gains a fraction ({c}_{y}=left({n}_{y}/hat{m}right)/{N}_{y}) of the fertilisations with each female, while the total reproductive success per female is f(nx,Ny). The mutant male’s reproductive success is therefore$${b}_{y}left(hat{m},mright)=hat{m}{c}_{y}fleft({n}_{x},{N}_{y}right)$$
(4)
To avoid division by 0, we can again define by (0, m) = 0, analogous to Model 2. In contrast to Models 1–2, there are now separate equations for each sex because of the additional sex-specific assumption of internal fertilisation, but no further sex-specific assumptions are used in their derivation. Visually the Bateman functions (Fig. 3) are nevertheless very similar to Model 2, and again reproduce the sex-specific shapes first proposed by Bateman1 when fertilisation is efficient. However, an interesting exception arises when relatively weak asymmetry in gamete numbers is combined with inefficient fertilisation and gamete limitation. When these conditions are combined with internal fertilisation, Bateman gradients can theoretically be reversed.Fig. 3: The Bateman functions of Eqs. (3) and (4) for internal fertilisation.Where Figs. 1 and 2 show that the sex-specific shapes of Bateman functions are ultimately caused by differences in gamete number, Fig. 3 shows that internal fertilisation does not invalidate this outcome when fertilisation is efficient. As in Fig. 2, results are shown for two values of resident matings (1 and 2), and the value m = 2 is used because it is comparable to the mean number of matings in Bateman’s1 work. a–h show the effect of variation in sex-specific gamete numbers and in fertilisation efficiency with m = 1, while i–p show the same with m = 2. Parameter values used are shown in the figure. Inefficient fertilisation combined with relatively low asymmetry in gamete numbers and the added asymmetry of internal fertilisation can in principle reverse the Bateman gradients (second and fourth row). Females (gamete number nx) are indicated by blue crosses and connecting lines, while males (gamete number ny) are indicated by black dots and connecting lines.Full size image More

Evans, J. D. & Spivak, M. Socialized medicine: individual and communal disease barriers in honey bees. J. Invertebr. Pathol. 103, S62–S72 (2010).PubMed 
Article 

Google Scholar 
Hughes, D. P., Pierce, N. E. & Boomsma, J. J. Social insect symbionts: evolution in homeostatic fortresses. Trends Ecol. Evol. 23, 672–677 (2008).PubMed 
Article 

Google Scholar 
Simone, M., Evans, J. D. & Spivak, M. Resin collection and social immunity in honey bees. Evolution 63, 3016–3022 (2009).CAS 
PubMed 
Article 

Google Scholar 
Dalenberg, H., Maes, P., Mott, B., Anderson, K. E. & Spivak, M. Propolis envelope promotes beneficial bacteria in the honey bee (Apis mellifera) mouthpart microbiome. Insects 11, 1–12 (2020).Article 

Google Scholar 
Poulsen, M., Bot, A. N. M., Nielsen, M. G. & Boomsma, J. J. Experimental evidence for the costs and hygienic significance of the antibiotic metapleural gland secretion in leaf-cutting ants. Behav. Ecol. Sociobiol. 52, 151–157 (2002).Article 

Google Scholar 
Rosengaus, R. B., Traniello, J. F. A., Lefebvre, M. L. & Maxmen, A. B. Fungistatic activity of the sternal gland secretion of the dampwood termite Zootermopsis angusticollis. Insect. Soc. 51, 259–264 (2004).Article 

Google Scholar 
Kwong, W. K. & Moran, N. A. Gut microbial communities of social bees. Nat. Rev. Microbiol. 14, 374–384 (2016).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Maes, P. W., Floyd, A. S., Mott, B. M. & Anderson, K. E. Overwintering honey bee colonies: effect of worker age and climate on the hindgut microbiota. Insects 12, 1–16 (2021).Article 

Google Scholar 
Brown, B. P. & Wernegreen, J. J. Deep divergence and rapid evolutionary rates in gut-associated Acetobacteraceae of ants. BMC Microbiol. 16, 140 (2016).PubMed 
PubMed Central 
Article 

Google Scholar 
Douglas, A. E. The microbial dimension in insect nutritional ecology. Funct. Ecol. 23, 38–47 (2009).Article 

Google Scholar 
Kešnerová, L. et al. Gut microbiota structure differs between honeybees in winter and summer. ISME J. 14, 801–814 (2020).PubMed 
Article 

Google Scholar 
Raymann, K., Shaffer, Z. & Moran, N. A. Antibiotic exposure perturbs the gut microbiota and elevates mortality in honeybees. PLoS Biol. 15, 1–22 (2017).Article 
CAS 

Google Scholar 
Anderson, K. E. & Ricigliano, V. A. Honey bee gut dysbiosis: a novel context of disease ecology. Curr. Opin. Insect Sci. 22, 125–132 (2017).PubMed 
Article 

Google Scholar 
Maes, P. W., Rodrigues, P. A. P., Oliver, R., Mott, B. M. & Anderson, K. E. Diet-related gut bacterial dysbiosis correlates with impaired development, increased mortality and Nosema disease in the honeybee (Apis mellifera). Mol. Ecol. 25, 5439–5450 (2016).CAS 
PubMed 
Article 

Google Scholar 
Miller, D. L., Smith, E. A. & Newton, I. L. G. A bacterial symbiont protects honey bees from fungal disease. bioRxiv https://doi.org/10.1101/2020.01.21.914325 (2020).Article 
PubMed 
PubMed Central 

Google Scholar 
Motta, E. V. S., Raymann, K. & Moran, N. A. Glyphosate perturbs the gut microbiota of honey bees. Proc. Natl. Acad. Sci. USA 115, 10305–10310 (2018).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Grice, E. A. & Segre, J. A. The skin microbiome. Nat. Rev. Microbiol. 9, 244–253 (2011).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Corby-Harris, V. et al. Origin and effect of Alpha 2.2 Acetobacteraceae in honey bee larvae and description of Parasaccharibacter apium gen. nov., sp. nov.. Appl. Environ. Microbiol. 80, 7460–7472 (2014).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
Floyd, A. S. et al. Microbial ecology of european foul brood disease in the honey bee (Apis mellifera): towards a microbiome understanding of disease susceptibility. Insects 11, 1–16 (2020).MathSciNet 
Article 

Google Scholar 
Babendreier, D., Joller, D., Romeis, J., Bigler, F. & Widmer, F. Bacterial community structures in honeybee intestines and their response to two insecticidal proteins. FEMS Microbiol. Ecol. 59, 600–610 (2007).CAS 
PubMed 
Article 

Google Scholar 
Sabree, Z. L., Hansen, A. K. & Moran, N. A. Independent studies using deep sequencing resolve the same set of core bacterial species dominating gut communities of honey bees. PLoS ONE 7, e41250 (2012).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Anderson, K. E. et al. Microbial ecology of the hive and pollination landscape: bacterial associates from floral nectar, the alimentary tract and stored food of honey bees (Apis mellifera). PLoS ONE 8, e83125 (2013).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
Rokop, Z. P., Horton, M. A. & Newton, I. L. G. Interactions between cooccurring lactic acid bacteria in honey bee hives. Appl. Environ. Microbiol. 81, 7261–7270 (2015).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Cox-foster, D. L. et al. A metagenomic survey of microbes in honey bee colony collapse disorder. Science 318, 283–287 (2007).ADS 
CAS 
PubMed 
Article 

Google Scholar 
Anderson, K. E., Rodrigues, P. A. P., Mott, B. M., Maes, P. & Corby-Harris, V. Ecological succession in the honey bee gut: shift in lactobacillus strain dominance during early adult development. Microb. Ecol. 71, 1008–1019 (2016).CAS 
PubMed 
Article 

Google Scholar 
Powell, J. E., Martinson, V. G., Urban-Mead, K. & Moran, N. A. Routes of acquisition of the gut microbiota of the honey bee Apis mellifera. Appl. Environ. Microbiol. 80, 7378–7387 (2014).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
Zheng, H., Powell, J. E., Steele, M. I., Dietrich, C. & Moran, N. A. Honeybee gut microbiota promotes host weight gain via bacterial metabolism and hormonal signaling. Proc. Natl. Acad. Sci. USA 114, 4775–4780 (2017).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Anderson, K. E. et al. Hive-stored pollen of honey bees: many lines of evidence are consistent with pollen preservation, not nutrient conversion. Mol. Ecol. https://doi.org/10.1111/mec.12966 (2014).Article 
PubMed 
PubMed Central 

Google Scholar 
Ludvigsen, J. et al. Shifts in the midgut/pyloric microbiota composition within a honey bee apiary throughout a season. Microb. Environ. 30, 235–244 (2015).Article 

Google Scholar 
Corby-Harris, V., Maes, P. & Anderson, K. E. The bacterial communities associated with honey bee (Apis mellifera) foragers. PLoS ONE 9, e95056 (2014).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
Münch, D., Kreibich, C. D. & Amdam, G. V. Aging and its modulation in a long-lived worker caste of the honey bee. J. Exp. Biol. 216, 1638–1649 (2013).PubMed 
PubMed Central 
Article 

Google Scholar 
Amdam, G. V. Social context, stress, and plasticity of aging. Aging Cell 10, 18–27 (2011).CAS 
PubMed 
Article 

Google Scholar 
Haddad, L. S., Kelbert, L. & Hulbert, A. J. Extended longevity of queen honey bees compared to workers is associated with peroxidation-resistant membranes. Exp. Gerontol. 42, 601–609 (2007).CAS 
PubMed 
Article 

Google Scholar 
Robinson, G. E. Hormonal and genetic control of honeybee division of labour. Behav. Physiol. Bees 14–27 (1991).Anderson, K. E. et al. The queen gut refines with age: longevity phenotypes in a social insect model. bioRxiv https://doi.org/10.1101/297507 (2018).Article 

Google Scholar 
Amdam, G. V., Norberg, K., Hagen, A. & Omholt, S. W. Social exploitation of vitellogenin. Proc. Natl. Acad. Sci. 100, 1799–1802 (2003).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Jones, B., Shipley, E. & Arnold, K. E. Social immunity in honeybees—density dependence, diet, and body mass trade-offs. Ecol. Evol. 8, 4852–4859 (2018).PubMed 
PubMed Central 
Article 

Google Scholar 
Alaux, C., Ducloz, F., Crauser, D. & Le Conte, Y. Diet effects on honeybee immunocompetence. Biol. Lett. 6, 562–565 (2010).PubMed 
PubMed Central 
Article 

Google Scholar 
Ohashi, K., Natori, S. & Kubo, T. Expression of amylase and glucose oxidase in the hypopharyngeal gland with an age-dependent role change of the worker honeybee (Apis mellifera L.). Eur. J. Biochem. 265, 127–133 (1999).CAS 
PubMed 
Article 

Google Scholar 
Vannette, R. L., Mohamed, A. & Johnson, B. R. Forager bees (Apis mellifera) highly express immune and detoxification genes in tissues associated with nectar processing. Sci. Rep. 5, (2015).Ohashi, K., Natori, S. & Kubo, T. Change in the mode of gene expression of the hypopharyngeal gland cells with an age-dependent role change of the worker honeybee Apis mellifera L.. Eur. J. Biochem. 249, 797–802 (1997).CAS 
PubMed 
Article 

Google Scholar 
Huang, Z. Y. & Robinson, G. E. Regulation of honey bee division of labor by colony age demography. Behav. Ecol. Sociobiol. 39, 147–158 (1996).Article 

Google Scholar 
Vojvodic, S. et al. The transcriptomic and evolutionary signature of social interactions regulating honey bee caste development. Ecol. Evol. 5, 4795–4807 (2015).PubMed 
PubMed Central 
Article 

Google Scholar 
Ohashi, K. et al. Functional flexibility of the honey bee hypopharyngeal gland in a dequeened colony. Zool. Sci. 17, 1089–1094 (2000).CAS 
Article 

Google Scholar 
Harwood, G., Salmela, H., Freitak, D. & Amdam, G. Social immunity in honey bees: royal jelly as a vehicle in transferring bacterial pathogen fragments between nestmates. J. Exp. Biol. 224 (2021).Santos, K. S. et al. Profiling the proteome complement of the secretion from hypopharyngeal gland of Africanized nurse-honeybees (Apis mellifera L.). Insect. Biochem. Mol. Biol. 35, 85–91 (2005).CAS 
PubMed 
Article 

Google Scholar 
Cremer, S., Armitage, S. A. O. & Schmid-Hempel, P. Social immunity. Curr. Biol. 17, 693–702 (2007).Article 
CAS 

Google Scholar 
Mattila, H. R. & Otis, G. W. Dwindling pollen resources trigger the transition to broodless populations of long-lived honeybees each autumn. Ecol. Entomol. 32, 496–505 (2007).Article 

Google Scholar 
Crailsheim, K., Riessberger, U., Blaschon, B., Nowogrodzki, R. & Hrassnigg, N. Short-term effects of simulated bad weather conditions upon the behaviour of food-storer honeybees during day and night (Apis mellifera carnica Pollmann). Apidologie 30, 299–310 (1999).Article 

Google Scholar 
Ricigliano, V. A. et al. Honey bees overwintering in a southern climate: Longitudinal effects of nutrition and queen age on colony-level molecular physiology and performance. Sci. Rep. 8, 1–11 (2018).CAS 
Article 

Google Scholar 
Ricigliano, V. A. et al. Honey bee colony performance and health are enhanced by apiary proximity to US Conservation Reserve Program (CRP) lands. Sci. Rep. 9, 1–11 (2019).CAS 
Article 

Google Scholar 
Fukuda, H. S. K. Seasonal change of the honey bee worker longevity in Sapporo, North Japan with notes on some factors affecting life span. Ecol. Soc. Jpn. 16, 206–212 (1966).
Google Scholar 
Mattila, H. R., Harris, J. L. & Otis, G. W. Timing of production of winter bees in honey bee (Apis mellifera) colonies. Insect. Soc. 48, 88–93 (2001).Article 

Google Scholar 
Feliciano-Cardona, S. et al. Honey bees in the tropics show winter bee-like longevity in response to seasonal dearth and brood reduction. Front. Ecol. Evol. 8, 1–8 (2020).Article 

Google Scholar 
Döke, M. A., Frazier, M. & Grozinger, C. M. Overwintering honey bees: biology and management. Curr. Opin. Insect. Sci. 10, 185–193 (2015).PubMed 
Article 

Google Scholar 
Liu, C. M. et al. BactQuant: an enhanced broad-coverage bacterial quantitative real-time PCR assay. BMC Microbiol. 12, 56 (2012).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Liu, C. M. et al. FungiQuant: a broad-coverage fungal quantitative real-time PCR assay. BMC Microbiol. 12, 1 (2012).CAS 
Article 

Google Scholar 
Schloss, P. D. et al. Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities. Appl. Environ. Microbiol. 75, 7537–7541 (2009).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Evans, J. D. Beepath: an ordered quantitative-PCR array for exploring honey bee immunity and disease. J. Invertebr. Pathol. 93, 135–139 (2006).CAS 
PubMed 
Article 

Google Scholar 
Bourgeois, A. L., Rinderer, T. E., Beaman, L. D. & Danka, R. G. Genetic detection and quantification of Nosema apis and N. ceranae in the honey bee. J. Invertebr. Pathol. 103, 53–58 (2010).CAS 
PubMed 
Article 

Google Scholar 
Pearson, K. Mathematical contributions to the theory of evolution. On a form of spurious correlation which may arise when indices are used in the measurement of organs. Proc. R. Soc. Lond. 60, 489–498 (1986).Gloor, G. B. & Reid, G. Compositional analysis: a valid approach to analyze microbiome high throughput sequencing data. Can. J. Microbiol. 703, 0821 (2016).
Google Scholar 
Comas, M. CoDaPack 2.0: a stand-alone, multi-platform compositional software. Options 1–10 (2011).Větrovský, T. & Baldrian, P. The variability of the 16S rRNA gene in bacterial genomes and its consequences for bacterial community analyses. PLoS ONE 8, 1–10 (2013).Article 
CAS 

Google Scholar 
Yek, S. H., Nash, D. R., Jensen, A. B. & Boomsma, J. J. Regulation and specificity of antifungal metapleural gland secretion in leaf-cutting ants. Proc. Biol. Sci. 279, 4215–4222 (2012).PubMed 
PubMed Central 

Google Scholar 
Evans, J. D. et al. Immune pathways and defence mechanisms in honey bees Apis mellifera. Insect. Mol. Biol. 15, 645–656 (2006).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Steinmann, N., Corona, M., Neumann, P. & Dainat, B. Overwintering is associated with reduced expression of immune genes and higher susceptibility to virus infection in honey bees. PLoS ONE 10, 1–18 (2015).Article 
CAS 

Google Scholar 
Seehuus, S.-C.C., Norberg, K., Gimsa, U., Krekling, T. & Amdam, G. V. Reproductive protein protects functionally sterile honey bee workers from oxidative stress. Proc. Natl. Acad. Sci. USA 103, 962–967 (2006).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Liu, J. R., Yang, Y. C., Shi, L. S. & Peng, C. C. Antioxidant properties of royal jelly associated with larval age and time of harvest. J. Agric. Food Chem. 56, 11447–11452 (2008).CAS 
PubMed 
Article 

Google Scholar 
Li-E, M., Jia, L., Yan, J., Xiao-Wen, L. & Xin, L. Isolation, purification and characterization of superoxide dismutase from royal jelly of the Italian worker bee, Apis mellifera. Acta Entomol. Sin. 47, 171–177 (2004).
Google Scholar 
Bottacini, F. et al. Bifidobacterium asteroides PRL2011 genome analysis reveals clues for colonization of the insect gut. 7, 1–14 (2012).Killer, J., Dubná, S., Sedláček, I. & Švec, P. Lactobacillus apis sp. nov., from the stomach of honeybees (Apis mellifera), having an in vitro inhibitory effect on the causative agents of American and European foulbrood. Int. J. Syst. Evol. Microbiol. 64, 152–157 (2014).Casteels, P. et al. Isolation and characterization of abaecin, a major antibacterial response peptide in the honeybee (Apis mellifera). Eur. J. Biochem. 187, 381–386 (1990).CAS 
PubMed 
Article 

Google Scholar 
Casteels, P., Ampe, C., Jacobs, F. & Tempst, P. Functional and chemical characterization of hymenoptaecin, an antibacterial polypeptide that is infection-inducible in the honeybee (Apis mellifera). J. Biol. Chem. 268, 7044–7054 (1993).CAS 
PubMed 
Article 

Google Scholar 
Barke, J. et al. A mixed community of actinomycetes produce multiple antibiotics for the fungus farming ant Acromyrmex octospinosus. BMC Biol. 8, 109 (2010).PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
Lyapunov, Y. E., Kuzyaev, R. Z., Khismatullin, R. G. & Bezgodova, O. A. Intestinal enterobacteria of the hibernating Apis mellifera mellifera L. bees. Microbiology 77, 373–379 (2008).Paiva, C. N. & Bozza, M. T. Are reactive oxygen species always detrimental to pathogens?. Antioxid. Redox Signal. 20, 1000–1034 (2014).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Burritt, N. L. et al. Sepsis and hemocyte loss in honey bees (Apis mellifera) Infected with Serratia marcescens strain sicaria. PLoS ONE 11, 1–26 (2016).Article 
CAS 

Google Scholar 
Bae, Y. S., Choi, M. K. & Lee, W. J. Dual oxidase in mucosal immunity and host-microbe homeostasis. Trends Immunol. 31, 278–287 (2010).CAS 
PubMed 
Article 

Google Scholar 
Ha, E. M., Oh, C. T., Bae, Y. S. & Lee, W. J. A direct role for dual oxidase in Drosophila gut immunity. Science 80(310), 847–850 (2005).ADS 
Article 
CAS 

Google Scholar 
Crailsheim, K., Hrassnigg, N., Gmeinbauer, R., Szolderits, M. J. & Schneider, L. H. W. Pollen utilization in non-breeding honeybees in Winter. J. Insect. Phys. 39, 369–373 (1993).Article 

Google Scholar 
Corona, M. & Robinson, G. E. Genes of the antioxidant system of the honey bee: annotation and phylogeny. 15, 687–701 (2006).Schwarz, R. S., Huang, Q. & Evans, J. D. Hologenome theory and the honey bee pathosphere. Curr. Opin. Insect. Sci. 10, 1–7 (2015).PubMed 
Article 

Google Scholar 
Corona, M., Hughes, K. A., Weaver, D. B. & Robinson, G. E. Gene expression patterns associated with queen honey bee longevity. Mech. Age. Dev. 126, 1230–1238 (2005).CAS 
Article 

Google Scholar 
Santos, D. E., Souza, A. D. O., Tibério, G. J., Alberici, L. C. & Hartfelder, K. Differential expression of antioxidant system genes in honey bee (Apis mellifera L.) caste development mitigates ROS-mediated oxidative damage in queen larvae. 20200173, (2020). More

Geography and climatology of study areaThe area of study, Ghana, is on the coastal edge of tropical West African, bounded in latitude 4.5° N and 11.5° N and longitude 3.5° W and 1.5° E, and characterized by a tropical monsoon climate system23,24. Figure 1 shows map of the study area indicating the selected twenty two (22) sunshine measurement stations distributed across the four main climatological zones and Table 1 summarizes the geographical positions of selected stations.Figure 1Adapted from Asilevi27.Map of the study area showing all twenty two (22) synoptic stations distributed in four main climatological zones countrywide.Full size imageTable 1 Geographical position and elevation for study sites.Full size tableAtmospheric clarity over the area is closely connected to cloud amount distribution and rainfall activities, largely determined by the oscillatory migration of the Inter-Tropical Discontinuity (ITD), accounting for the West African Monsoon (WAM)25,26.Owing to the highly variable spatiotemporal distribution of cloud amount vis-à-vis rainfall activities, resulting in contrasting climatic conditions in different parts of the region, the country is partitioned by the Ghana Meteorological Agency (GMet) into four main agro-ecological zones namely, the Savannah, Transition, Forest and Coastal zones as shown in Fig. 123. As a result, the region experiences an estimated Global solar radiation (GSR) intensity peaks in April–May and then in October–November, with the highest monthly average of 22 MJm−2 day−1 over the savannah climatic zone and the lowest monthly average of 13 MJm−2 day−1 over the forest climatic zone27.Research datasetsGround-based measurement dataDaily sunshine duration measurement datasets (n) spanning 1983–2018 where derived for estimating Global solar radiation (GSR). The measurements were taken by the Campbell-Stokes sunshine recorder, mounted at the 22 stations shown in Fig. 1, under unshaded conditions to ensure optimum sunlight exposure. The device concentrates sunlight onto a thin strip of sunshine card, which causes a burnt line representing the total period in hours during which sunshine intensity exceeds 120.0 Wm−2 according to World Meteorological Organization (WMO) recommendations27. The as-received daily records were quality control checked by ensuring 0 ≤ n ≤ N, where N is the astronomical day length representing the possible maximum duration of sunshine in hours determined by Eq. 1 from the latitude (ϕ) of the site of interest and the solar declination (δ) computed by Eq. 227:$$ {text{N}} = frac{2}{15}cos^{ – 1} left[ { – tan phi tan {updelta }} right] $$
(1)
$$ {updelta } = 23.45sin left[ {360^{{text{o}}} times frac{{284 + {text{J}}}}{365}} right] $$
(2)
where J represents the number for the Julian day of the year (first January is 1 and second January is 2).NASA-POWER Global solar radiation (GSR) reanalysis dataThe satellite-based Global solar radiation (GSR) dataset for specific longitudes and latitudes of all 22 stations, assessed in the study, were retrieved from the National Aeronautics and Space Administration-Prediction of Worldwide Energy Resources (NASA-POWER) reanalysis repository based on the Modern Era Retrospective-Analysis for Research and Applications (MERRA-2) assimilation model products, developed from Surface Radiation Budget, and spanning equal study period (1983–2018). The datasets are accessible on a daily and monthly temporal resolution scales at 0.5° × 0.5° spatial coverage via a user friendly web-based mapping portal: https://power.larc.nasa.gov/data-access-viewer/17. The advantage of the NASA-POWER reanalysis GSR, is the wide spatial coverage, and thus can be used to develop a high spatial resolution of solar radiation across the study area.The POWER Project analyzes, synthesizes and makes available surface radiation related parameters on a global scale, primarily from the World Climate Research Programme (WCRP), Global Energy and Water cycle Experiment (GEWEX), Surface Radiation Budget (SRB) project (Version 2.9), the Clouds and the Earth’s Radiant Energy System (CERES), FLASHFlux (Fast Longwave and Shortwave Radiative Fluxes from CERES and MODIS), and the Global Modeling and Assimilation Office (GMAO)17. Table 2 shows the source satellites and the corresponding temporal coverage used in the development of NASA-POWER GSR products.Table 2 Satellites providing the NASA-POWER GSR datasets20.Full size tableThe monthly average NASA-POWER all-sky shortwave surface radiation reanalysis products are statistically validated, showing reasonable biases of − 6.6–13%, against a global network of surface radiation measurement metadata in an integrated database from the Baseline Surface Radiation Network (BSRN) of the World Radiation Monitoring Center (WRMC)20,22. The datasets are widely used in renewable energy application16,22, agricultural modelling of crop yields28, crop simulation exercises29, and plant disease modelling30.Furthermore, in order to assess the suitability of the NASA-POWER surface solar radiation products for the study area, a synthetic sunshine duration based Global solar radiation (GSR) is developed from the Angstrom-Prescott sunshine duration model by Eq. 3 for comparisons27.$$ {text{GSR}} = left[ {{text{a}} + {text{b}}frac{{text{n}}}{{text{N}}}} right]{text{H}}_{{text{o}}} $$
(3)
were Ho (kWhm−2 day−1) is the daily extraterrestrial solar radiation on an horizontal surface, n is the daily sunshine duration measurements obtained from the Ghana Meteorological Agency (GMet), and N is the maximum possible daily sunshine duration or the day length in hours determined by Eq. 1. Generalized regression constants a = 0.25 and b = 0.5 for the study area were determined by Asilevi27 from experimental radiometric data based on correlation regression analysis between atmospheric clarity index (GSR/Ho) and atmospheric cloudlessness index (n/N), for estimating solar radiation over the study area, and compared with other satellite data retrieved from the National Renewable Energy Laboratory (NREL) and the German Aerospace Centre (DLR)27. Ho was calculated from astronomical parameters by Eq. 4:$$ {text{H}}_{0} = frac{{24{ } cdot { }60}}{pi } cdot {text{G}}_{{{text{sc}}}} cdot {text{d}}_{{text{r}}} left[ {omega_{{text{s}}} sin varphi sin delta + cos varphi cos delta sin omega_{{text{s}}} } right] $$
(4)
where Gsc is the Solar constant in MJm−2 min−1, dr is the relative Earth–Sun distance in meters (m), (omega_{s}) is the sunset hour angle (angular distance between the meridian of the observer and the meridian whose plane contains the sun), (delta) is the angle of declination in degrees (°) and (varphi) is the local latitude. A detailed presentation of the calculation was published in a previous work27.Statistical assessment analysisFor the purpose of assessing the NASA-POWER derived monthly mean GSR (GSRn) datasets in comparison with the estimated Global Solar Radiation (GSRe) datasets used in this paper, the following deviation and correlation methods in Eqs. 5–11, each showing a complimentary result were used: Standard deviation (({upsigma })), residual error (RE), Root mean square error (RMSE), Mean bias error (MBE), Mean percentage error (MPE), Pearson’s correlation coefficient (r), and Willmott index of agreement (d) for n observations31,32,33,34,35. GSRe, GSRn, and RE represent the estimated GSR, NASA-POWER GSR, and the residual error between GSRe and GSRn respectively. A positive RE indicates that sunshine-based estimated GSR is larger than the NASA-POWER reanalysis dataset, while a negative RE indicates that sunshine-based estimated GSR is smaller than the NASA-POWER reanalysis dataset. The arithmetic mean of any dataset is µ.The standard deviation (({upsigma })) was used to check the upper and lower limits of distribution around the mean deviations between GSRe and GSRn in order to ascertain violations between both datasets33. The RMSE is a standard statistical metric to quantify error margins in meteorology and climate research studies, and by definition is always positive, representing zero in the ideal case, plus a smaller value signifying a good marginal deviation31. The MBE is a good indicator for under-or overestimation in observations, with MBE values closest to zero being desirable. The MPE further indicates the percentage deviation between the GSRe and GSRn individual datasets35.$$ {upsigma } = sqrt {frac{1}{{{text{n}} – 1}}mathop sum limits_{{{text{i}} = 1}}^{{text{n}}} left( {{text{GSR}} – {upmu }} right)^{2} } $$
(5)
$$ {text{RE}} = {text{GSR}}_{{text{e}}} – {text{GSR}}_{{text{n}}} $$
(6)
$$ {text{RMSE}} = sqrt {frac{1}{{text{n}}}mathop sum limits_{{{text{i}} = 1}}^{{text{n}}} left( {{text{RE}}} right)^{2} } $$
(7)
$$ {text{MBE}} = frac{1}{{text{n}}}mathop sum limits_{{{text{i}} = 1}}^{{text{n}}} left( {{text{RE}}} right) $$
(8)
$$ {text{MPE}} = frac{1}{{text{n}}}mathop sum limits_{{{text{i}} = 1}}^{{text{n}}} left( {frac{{{text{RE}}}}{{{text{GSR}}_{{text{e}}} }} times 100{text{% }}} right) $$
(9)
$$ {text{r}} = frac{{mathop sum nolimits_{{{text{i}} = 1}}^{{text{n}}} left( {{text{GSR}}_{{text{e}}} – {upsigma }_{{text{e}}} } right)left( {{text{GSR}}_{{text{n}}} – {upsigma }_{{text{n}}} } right)}}{{left( {{text{n}} – 1} right){upsigma }_{{text{e}}} {upsigma }_{{text{n}}} }} $$
(10)
$$ {text{d}} = 1 – left[ {frac{{mathop sum nolimits_{{{text{i}} = 1}}^{{text{n}}} left( {{text{GSR}}_{{text{e}}} – {text{GSR}}_{{text{n}}} } right)^{2} }}{{mathop sum nolimits_{{{text{i}} = 1}}^{{text{n}}} left( {left| {{text{GSR}}_{{text{e}}} – {text{GSR}}_{{{text{nave}}}} left| + right|{text{GSR}}_{{text{n}}} – {text{GSR}}_{{{text{nave}}}} } right|} right)^{2} }}} right] $$
(11)
Further, as with other statistical studies in meteorology36, the Pearson’s correlation coefficient (r) was used to quantify the strength of correlation between GSRe and GSRn. Finally, the Willmott index of agreement (d) commonly used in meteorological literature computed from Eq. 7 is used to assess the degree of GSRe/GSRn agreement34. More

Soil sample collectionSoil samples were collected from three different crude oil reservoirs et al. Faisaliyyah, Al Sina’iyah, and Ghubairah located in Riyadh, Saudi Arabia. Briefly, 400 g of soil samples were collected at 0–10 cm depth, under aseptic conditions. Samples were sieved by 2.5 mm pore size sieves, homogenized, and stored at 4ºC until use.Sources of different hydrocarbonsDifferent samples of crude oil, kerosene, diesel, and used oil were collected in sterile flasks from the tankers of Saudi Aramco Company (Dammam, Saudi Arabia). Additionally, another flask was prepared by mixing 1% of each oil in MSM liquid media to make up the mixed oil. The oil samples were sterilized by Millex® Syringe Filters (Merck Millipore co., Burlington, MA, United States) and stored at 4 °C for further usage.Isolation and identification of fungal speciesThe fungal species in the soil contaminated by crude oil were identified using the dilution method. Briefly, 10% of each soil sample was dissolved in distilled water and vortexed thoroughly. Then, 0.2 ml of each sample was cultured on a sterile PDA plate incubated at 28 °C for three days until the growth of different fungal colonies. Carefully, each colony was isolated, re-cultured on new PDA McCartney bottles of PDA slant, and incubated at 28 °C for three days. The fungi were identified microscopically using standard taxonomic keys based on typical mycelia growth and morphological characteristics provided in the mycological keys54. Besides, the taxonomy of the isolated yeast strains was confirmed by the API 20 C AUX kit (Biomerieux Corp., Marcy-l’Étoile, France) (data not shown). The morphology of pure cultures was tested and identified under a light microscope as described before55.The incidence of each strain was calculated as follows:$$ Incidence ;(% ) = frac{{{text{Number }};{text{of }};{text{samples }};{text{showed }};{text{microbial }};{text{growth}}}}{{{text{Total }};{text{samples}}}} times 100 $$Hydrocarbon tolerance testThe growth rate of isolated strains was tested in a liquid medium of MSM mixed with 1% of either crude oil, used oil, diesel, kerosene, or mixed oil. Furthermore, a control sample of MSM liquid medium without any of the oils tested and all culture media were autoclaved at 121 °C for 30 min. After cooling, 1 ml of each isolate was inoculated with one of the above mixtures and incubated at 25 °C on an orbital shaker. The growth rate was measured every three days for a month for each treatment versus the control. All experiments were performed in triplicates.Scanning electron microscopy (SEM)The morphology of different strains of the isolated fungi was tested by SEM, as previously described56, with some modifications. Briefly, 1 ml of each growing strain, in the liquid media, was centrifuged at the maximum speed (14,000 rpm) for 1 min, followed by fixation with 2.5% glutaraldehyde, and overnight incubation at 5 °C. Later, the sample was pelleted, washed with distilled water, then dehydrated with different ascending concentrations of ethanol (30, 50, 70, 90, 100 (v/v)) for 15 min at room temperature. Finally, samples were examined in the Prince Naif Research Centre (King Saud University, Riyadh, Saudi Arabia) by the JEOL JEM-2100 microscope (JEOL, Peabody, MA, United States), according to the manufacturer instructions.Crude oil degradation assayA modified version of the DCPIP assay57 was employed to assess the oil-degrading ability of the fungal isolates. For each strain, 100 ml of the autoclaved MSM was mixed with 1% (V/V) of one of the hydrocarbons (crude oil, used oil, diesel, kerosene, or mixed oil), 0.1% (v/v) of Tween 80, and 0.6 mg/mL of the redox indicator (DCPIP). Then, 1–2 ml of different fungi growing in liquid media (24–48 h) add to the Crude Oil Degradation media, prepared previously, and incubated for two weeks in a shaking incubator at 25 °C. All flasks were covered and protected from light, aeration, or temperature exchanges to reduce the effects of oil weathering (evaporation, photooxidation). The surfactant Tween 80 was used for bio-stimulation and acceleration of the biosurfactant production by increasing metabolism58. A non-inoculated Crude Oil Degradation media was used as the negative control. Afterward, the colorimetric analysis for the change in DCPIP color was estimated, spectrophotometrically, at 420 nm. All experiments were performed in triplicates.Preparation of cell-free supernatant (CFS)To prepare the Cell-Free Supernatant (CFS), all isolates were grown in MSM broth medium with 1% of either crude oil, used oil, diesel, kerosene, or mixed oil for 30 days in a shaking incubator at 25 °C. After incubation, the cells were removed by centrifugation at 10,000 rpm for 30 min at 4 °C. The supernatant (CFS) was collected and filter-sterilized with a 0.45 μm pore size sterile membrane. CFS was screened for the production of different biosurfactants. All the experiments were carried out in triplicates, and the average values were calculated.Drop-Collapse assayThe Drop-Collapse assay was performed as previously described9, with some modifications. 100 µl of crude oil was applied on glass slides, then 10 µl of each CFS was added to the center of the slide surface and incubated for a minute at room temperature. The slides were imaged by a light microscope using the 10X objective lenses. The spreading on the soil surface was scored by either « + » to indicate the level of positive spreading, biosurfactant production, or «—» for negative spreading. Biosurfactant production was considered positive at the drop diameter ≥ 0.5 mm, compared to the negative control (treated with distilled water).Oil spreading assayAn amount of 20 ml of water was added to the Petri plate (size of 100 mm) and mixed with 20 µl of crude oil or mixed oil, which created a thin layer on the water surface. Then, 10 µl of CFS was delivered onto the surface of the oil, and the clear zone surrounding the CFS drop was observed. The results were compared to the negative control (without CFS) and positive control of 1% SDS41. We have measured the clear zones diameter from images and calculate the actual values in regards to the diameter of the Petri dish (10 cm). The assay was performed in triplicates.Emulsification activity assayThe emulsification activity of each isolate was assessed by mixing equal volumes of MSM broth medium of each isolate with different oils in separate tubes. The samples were homogenized by vortex at high speed for two minutes at room temperature (25 °C) and allowed to settle for 24 h. The tests were performed in duplicate. Then, the emulsification index was calculated as follows59:$$ Emulsification; activity; left( % right) = frac{{{text{Height }};{text{of }};{text{emulsion }};{text{layer}}}}{{{text{Total }};{text{height}}}} times 100 $$Recovery of biosurfactantsThe recovery of biosurfactants from CFS was tested through different assays:Acid precipitation assay3 ml of each CFS was adjusted by 6 N HCl to pH 2 and incubated for 24 h at 4 °C. Later, equal volumes of chloroform/methanol mixture (2:1 v/v) were added to each tube, vortexed, and incubated overnight at room temperature. Afterward, the samples were centrifuged for 30 min at 10,000 rpm (4 °C), the precipitate (Light brown colored paste) was air-dried in a fume hood, and weighed53.Solvent extraction assayThe CFS containing biosurfactant was treated with a mixture of extraction solvents (equal volumes of methanol, chloroform, and acetone). Then, the new mixture was incubated in a shaking incubator at 200 rpm, 30 °C for 5 h. The precipitate was separated into two layers, in which the lower layer (White) was isolated, dried, weighed, and stored60.Ammonium sulfate precipitation assayThe CFS containing biosurfactant was precipitated with 40% (w/v) ammonium sulfate and incubated overnight at 4 °C. The samples were centrifuged at 10,000 rpm for 30 min (4 °C). The precipitate was collected and extracted with an amount of acetone equal to the volume of the supernatant. After centrifugation, the precipitate (Creamy-white) was isolated, air-dried in a fume hood, and weighed53.Zinc sulfate precipitation methodSimilarly, 40% (w/v) zinc sulfate was mixed with the CFS containing biosurfactant. Then, the mixture was incubated at 4 °C, overnight. The precipitate (Light Brown) was collected by centrifugation at 10,000 rpm for 30 min (4 °C), air-dried in a fume hood, and weighed53.Statistical analysisAll experiments were performed in triplicate, and the results were expressed as the mean values ± standard deviation (SD). One-way ANOVA and Dunnett’s tests were used to estimate the significance levels at P  More