Ecology

1.Bengis, R. G. A revue of bovine Brucellosis in free-ranging African wildlife. in Proceedings of the ARC-Onderstepoort, OIE International Congress with WHO-Cosponsorship on anthrax, brucellosis, CBPP, clostridial and mycobacterial diseases : Berg-en-Dal, Kruger National Park, South Africa 178–183 (Onderstepoort Veterinary Inst, 1998).2.Kaliner, G., Staak, C., Kalinerj, G. & Staaklu, C. A case of orchitis caused by Brucella abortus in the African buffalo. J. Wildl. Dis. 9, 251–253 (1973).Article 

Google Scholar 
3.Schiemann, B. & Staak, C. Brucella melitensis in impala (Aepyceros melampus). Vet. Rec. 88, 344–344 (1971).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
4.Ndengu, M. et al. Seroprevalence of brucellosis in cattle and selected wildlife species at selected livestock/wildlife interface areas of the Gonarezhou National Park Zimbabwe. Prev. Vet. Med. 146, 158–165 (2017).PubMed 
Article 
PubMed Central 

Google Scholar 
5.Rollinson, D. H. L. Brucella agglutinins in East African game animals. Vet. Rec. 74, 904 (1962).
Google Scholar 
6.De Vos, V. & Van Niekerk, C. A. W. Brucellosis in the Kruger National Park. J. S. Afr. Vet. Assoc. 40, 331–334 (1969).
Google Scholar 
7.Sachs, R. & Staak, C. Evidence of brucellosis in antelope in the Serengeti. Vet. Record 79, 857–856 (1966).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
8.El-Tras, W. F., Tayel, A. A., Eltholth, M. M. & Guitian, J. Brucella infection in fresh water fish : Evidence for natural infection of Nile catfish, Clarias gariepinus, with Brucella melitensis. Vet. Microbiol. 141, 321–325 (2010).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
9.Lane, E. P. et al. A systematic health assessment of Indian ocean bottlenose (Tursiops aduncus) and indo-pacific humpback (Sousa plumbea) dolphins incidentally caught in shark nets off the KwaZulu-Natal coast South Africa. PLoS ONE 9, e107038 (2014).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
10.Salem, A. A., Hamed, O. M. & Abd-Elkarim, A. M. Studies on some Brucella carriers in Egypt. Assiut Vet Med J 1, 181–187 (1974).
Google Scholar 
11.Condy, J. B. The status of disease in Rhodesian wildlife. Rhod. Sci. News 2, 96–99 (1968).
Google Scholar 
12.Condy, J. B. & Vickers, D. B. The isolation of Brucella abortus from a waterbuck (Kobus ellipsiprymnus). Vet. Rec. 85, 200 (1969).Article 

Google Scholar 
13.Bell, L. M., Hayles, L. B. & Chanda, A. B. Evidence of reservoir hosts of Brucella melitensis. Med. J. Zambia 10, 152–153 (1976).CAS 
PubMed 
PubMed Central 

Google Scholar 
14.Gradwell, D. V., Schutte, A. P., van Niekerk, C. A. & Roux, D. J. The isolation of Brucella abortus biotype I from African buffalo in the Kruger National Park. J. S. Afr. Vet. Assoc. 48, 41–43 (1977).CAS 
PubMed 
PubMed Central 

Google Scholar 
15.Karesh, W. B. et al. Health evaluation of five sympatric duiker species (Cephalophus spp.). J. Zool. Wildl. Med. 26, 485–502 (1995).
Google Scholar 
16.Fick, S. E. & Hijmans, R. J. WorldClim 2: New 1-km spatial resolution climate surfaces for global land areas. Int. J. Climatol. 37, 4302–4315 (2017).Article 

Google Scholar 
17.Bengis, R. G. & Erasmus, J. M. Wildlife diseases in South Africa: A review. Rev. Sci. Tech. Off. Int. des Epizoot. 7, 807–821 (1988).Article 

Google Scholar 
18.Durrheim, D. N. et al. Safety of travel in South Africa: The Kruger National Park. J. Travel Med. 8, 176–191 (2006).Article 

Google Scholar 
19.Eisenberg, T. et al. Isolation of potentially novel Brucella spp. from frogs. Appl. Environ. Microbiol. 78, 3753–3755 (2012).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
20.Hoogstral, H., Kaiser, M. N., Traylor, M. A., Guindy, E. & Gaber, S. Ticks (Ixodidae) on birds migrating from Europe and Asia to Africa 1959–61. Bull. World Health Organ. 28, 235–262 (1963).
Google Scholar 
21.Michel, A. L. A. L. & Bengis, R. G. R. G. The African buffalo: A villain for inter-species spread of infectious diseases in southern Africa. Onderstepoort. J. Vet. Res. 79, 5 (2012).Article 

Google Scholar 
22.Monroe, B. P. et al. Collection and utilization of animal carcasses associated with zoonotic disease in Tshuapa district, the democratic republic of the Congo, 2012. J. Wildl. Dis. 51, 734–738 (2015).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
23.Wolhuter, J., Bengis, R. G., Reilly, B. K. & Cross, P. C. Clinical demodicosis in African buffalo (Syncerus caffer) in the Kruger National Park. J. Wildl. Dis. 45, 2 (2009).Article 

Google Scholar 
24.Worthington, R. W. & Bigalke, R. D. A review of the infectious diseases of African wild ruminants. Onderstepoort. J. Vet. Res. 68, 291–323 (2001).CAS 
PubMed 
PubMed Central 

Google Scholar 
25.Mühldorfer, K. et al. The role of ‘atypical’ Brucella in amphibians: are we facing novel emerging pathogens?. J. Appl. Microbiol. 122, 40–53 (2017).PubMed 
Article 
CAS 
PubMed Central 

Google Scholar 
26.Ducrotoy, M. et al. Brucellosis in Sub-Saharan Africa: Current challenges for management, diagnosis and control. Acta Trop. 165, 179–193 (2017).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
27.Munagandu, et al. Disease constraints for utilization of the African buffalo (Syncerus caffer) on game ranches in Zambia. Jpn. J. Vet. Res. 54, 3–13 (2006).
Google Scholar 
28.Munyua, P. et al. Prioritization of zoonotic diseases in Kenya, 2015. PLoS ONE 11, e0161576 (2016).PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
29.Conrad, P. A., Meek, L. A. & Dumit, J. Operationalizing a One Health approach to global health challenges. Comp. Immunol. Microbiol. Infect. Dis. 36, 211–216 (2013).PubMed 
Article 
PubMed Central 

Google Scholar 
30.Bekker, J. L., Hoffman, L. C. & Jooste, P. J. Wildlife-associated zoonotic diseases in some southern African countries in relation to game meat safety: A review. Onderstepoort. J. Vet. Res. 79, 12 (2012).Article 

Google Scholar 
31.Muma, J. B. et al. The contribution of veterinary medicine to public health and poverty reduction in developing countries. Vet. Ital. 50, 117–129 (2014).PubMed 
PubMed Central 

Google Scholar 
32.Mugizi, D. R. et al. Isolation and Molecular Characterization of Brucella Isolates in Cattle Milk in Uganda. 2015, (2015).33.Mathew, C. et al. First isolation, identification, phenotypic and genotypic characterization of Brucella abortus biovar 3 from dairy cattle in Tanzania. BMC Vet. Res. 11, 2 (2015).Article 

Google Scholar 
34.Meyer, M. E. & Morgan, W. J. B. Designation of neotype strains and of biotype reference strains for species of the genus Brucella Meyer and Shaw. Int. J. Syst. Bacteriol. 23, 135–141 (1973).Article 

Google Scholar 
35.National Academies of Sciences, Engineering, and M. Revisiting brucellosis in the greater yellowstone area. Revisiting Brucellosis in the Greater Yellowstone Area (National Academies Press, 2017). doi:https://doi.org/10.17226/2475036.Muma, J. B. et al. Brucella seroprevalence of the Kafue lechwe (Kobus leche kafuensis) and Black lechwe (Kobus leche smithemani): Exposure associated to contact with cattle. Prev. Vet. Med. 100, 256–260 (2011).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
37.Gorsich, E. E., Ezenwa, V. O., Cross, P. C., Bengis, R. G. & Jolles, A. E. Context-dependent survival, fecundity and predicted population-level consequences of brucellosis in African buffalo. J. Anim. Ecol. 84, 999–1009 (2015).PubMed 
Article 
PubMed Central 

Google Scholar 
38.Hoogstraal, H., Kaiser, M. N., Traylor, M. A., Gaber, S. & Guindy, E. Ticks (Ixodoidea) on birds migrating from Africa to Europe and Asia. Bull. World Health Organ. 24, 197–212 (1961).CAS 
PubMed 
PubMed Central 

Google Scholar 
39.Alexander, K. A. et al. Buffalo, bush meat, and the zoonotic threat of brucellosis in Botswana. PLoS ONE 7, e32842 (2012).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
40.Munn, Z., Moola, S., Riitano, D. & Lisy, K. The development of a critical appraisal tool for use in systematic reviews addressing questions of prevalence. Int. J. Heal. Policy Manag. 3, 123–128 (2014).Article 

Google Scholar 
41.Madsen, M. et al. Serologic survey of Zimbabwean wildlife for brucellosis. J. Zoo. Wildl. Med. 26, 240–245 (1995).
Google Scholar 
42.Roberts, M. G. & Heesterbeek, J. A. P. Quantifying the dilution effect for models in ecological epidemiology. J. R. Soc. Interface 15, 2 (2018).Article 

Google Scholar 
43.Viana, M. et al. Assembling evidence for identifying reservoirs of infection. Trends Ecol. Evol. 29, 270–279 (2014).PubMed 
PubMed Central 
Article 

Google Scholar 
44.Souley Kouato, B. et al. Spatio-temporal patterns of foot-and-mouth disease transmission in cattle between 2007 and 2015 and quantitative assessment of the economic impact of the disease in Niger. Transbound Emerg. Dis. 65, 1049–1066 (2018).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
45.Godfroid, J., Nielsen, K. & Saegerman, C. Diagnosis of brucellosis in livestock and wildlife. Croat Med. J. 51, 296–305 (2010).PubMed 
PubMed Central 
Article 

Google Scholar 
46.Hartling, L. et al. Grey literature in systematic reviews : a cross-sectional study of the contribution of non-English reports, unpublished studies and dissertations to the results of meta- analyses in child-relevant reviews. 1–11 (2017). doi:https://doi.org/10.1186/s12874-017-0347-z47.Condy, J. B. & Vickers, D. B. Brucellosis in Rhodesian wildlife. J. S. Afr. Vet. Assoc. 43, 175–179 (1972).CAS 
PubMed 
PubMed Central 

Google Scholar 
48.Erume, J. et al. Serological and molecular investigation for brucellosis in swine in selected districts of Uganda. Trop. Anim. Health Prod. 48, 1147–1155 (2016).PubMed 
PubMed Central 
Article 

Google Scholar 
49.Godfroid, J., Beckmen, K. & Helena Nymo, I. Removal of lipid from serum increases coherence between brucellosis rapid agglutination test and enzyme-linked immunosorbent assay in bears in Alaska, USA. J. Wildl. Dis. 52, 912–915 (2016).PubMed 
Article 
PubMed Central 

Google Scholar 
50.Matope, G., Bhebhe, E., Muma, J. B. B., Lund, A. & Skjerve, E. Herd-level factors for Brucella seropositivity in cattle reared in smallholder dairy farms of Zimbabwe. Prev. Vet. Med. 94, 213–221 (2010).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
51.Mwebe, R., Nakavuma, J. & Moriyón, I. Brucellosis seroprevalence in livestock in Uganda from 1998 to 2008: a retrospective study. Trop. Anim. Health Prod. 43, 603–608 (2011).PubMed 
Article 
PubMed Central 

Google Scholar 
52.Aune, K., Rhyan, J. C., Russell, R., Roffe, T. J. & Corso, B. Environmental persistence of Brucella abortus in the Greater Yellowstone Area. J. Wildl. Manag. 76, 253–261 (2012).Article 

Google Scholar 
53.Enström, S. et al. Brucella seroprevalence in cattle near a wildlife reserve in Kenya. BMC Res. Notes 10, 2 (2017).Article 

Google Scholar 
54.Godfroid, J. Brucellosis in wildlife. Rev. Sci. Tech. 21, 277–286 (2002).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
55.Martin, C., Pastoret, P. P., Brochier, B., Humblet, M. F. & Saegerman, C. A survey of the transmission of infectious diseases/infections between wild and domestic ungulates in Europe. Vet. Res. 42, 70 (2011).PubMed 
PubMed Central 
Article 

Google Scholar 
56.Godfroid, J. et al. A ‘One Health’ surveillance and control of brucellosis in developing countries: Moving away from improvisation. Comp. Immunol. Microbiol. Infect. Dis. 36, 241–248 (2013).PubMed 
Article 
PubMed Central 

Google Scholar 
57.Kamath, P. L. et al. Genomics reveals historic and contemporary transmission dynamics of a bacterial disease among wildlife and livestock. Nat. Commun. 7, 11448 (2016).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
58.Michel, A. L. et al. Wildlife tuberculosis in South African conservation areas: Implications and challenges. Vet. Microbiol. 112, 91–100 (2006).ADS 
CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
59.Pandey, G. S. et al. Serosurvey of brucella spp. infection in the Kafue Lechwe (Kobus leche kafuensis) of the Kafue flats in Zambia. Indian Vet. J. 76, 275–278 (1999).
Google Scholar 
60.Olsen, S. & Tatum, F. Swine brucellosis: Current perspectives. Vet. Med. Res. Rep. 8, 1–12 (2016).
Google Scholar 
61.Menshawy, A. M. S. et al. Assessment of Genetic Diversity of Zoonotic Brucella spp. Recovered from Livestock in Egypt Using Multiple Locus VNTR Analysis. (2014). doi:https://doi.org/10.1155/2014/35387662.Ibrahim, S. Studies on swine brucellosis in Egypt. J. Egypt Vet. Med. Assoc. 56, 1–12 (1996).
Google Scholar 
63.Ledwaba, B., Mafofo, J. & Van Heerden, H. Genome sequences of Brucella abortus and Brucella suis strains isolated from Bovine in Zimbabwe. Genome Announc. 2, 1063–1077 (2014).Article 

Google Scholar 
64.Fretin, D. et al. Unexpected Brucella suis biovar 2 infection in a dairy cow, Belgium. Emerging Infectious Diseases 19, 2053–2054 (Centers for Disease Control and Prevention, 2013).65.Maurin, M. Brucellosis at the dawn of the 21st century. Médecine Mal. Infect. 35, 6–16 (2005).CAS 
Article 

Google Scholar 
66.Whatmore, A. M. et al. Brucella papionis sp. nov., isolated from baboons (Papio spp.). Int. J. Syst. Evol. Microbiol. 64, 4120–4128 (2014).PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
67.Godfroid, J., Garin-Bastuji, B., Saegerman, C. & Blasco, J. M. Brucellosis in terrestrial wildlife. Rev. Sci. Tech. Off. Int. Epiz. 32, 27–42 (2013).CAS 
Article 

Google Scholar 
68.Barendregt, J. J., Doi, S. A., Lee, Y. Y., Norman, R. E. & Vos, T. Meta-analysis of prevalence. J. Epidemiol. Community Health 67, 974–978 (2013).PubMed 
Article 
PubMed Central 

Google Scholar 
69.EpiGear. EpiGear International. Available at: http://www.epigear.com/. (Accessed: 8th February 2018)70.Doi, S. A. R. R., Barendregt, J. J., Khan, S., Thalib, L. & Williams, G. M. Advances in the meta-analysis of heterogeneous clinical trials I: The inverse variance heterogeneity model. Contemp. Clin. Trials 45, 130–138 (2015).PubMed 
Article 
PubMed Central 

Google Scholar 
71.Higgins, J. P. T., Thompson, S. G., Deeks, J. J. & Altman, D. G. Measuring inconsistency in meta-analyses. BMJ 327, 557–560 (2003).PubMed 
PubMed Central 
Article 

Google Scholar 
72.Heisch, R. B., Cooke, E. R., Harvey, A. E. & De Souz, F. The isolation of Brucella suis from rodents in Kenya. East Afr. Med. J. 40, 132–133 (1963).CAS 
PubMed 
PubMed Central 

Google Scholar 
73.Motsi, T. R., Tichiwangana, S. C., Matope, G., Mukarati, N. L. & Studies, V. A serological survey of brucellosis in wild ungulate species from five game parks in Zimbabwe. Onderstepoort. J. Vet. Res. 80, 586 (2013).PubMed 
Article 
PubMed Central 

Google Scholar 
74.Roth, H. H. A survey of brucellosis in game animals in Rhodesia. Bull. Epizoot. Dis. Afr. Bull. des Epizoot en Afrique 15, 133–142 (1967).CAS 

Google Scholar 
75.Condy, J. B. & Vickers, D. B. Brucellosis in buffalo in Wankie National Park. Rhod. Vet. J. 8, 58–60 (1976).
Google Scholar 
76.Caron, A. et al. Relationship between burden of infection in ungulate populations and wildlife/livestock interfaces. Epidemiol. Infect. 141, 1522–1535 (2013).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
77.Gomo, C. et al. Detection of Brucella abortus in Chiredzi district in Zimbabwe. Onderstepoort. J. Vet. Res. 79, 1–5 (2012).Article 

Google Scholar 
78.Chaparro, F., Lawrence, J. V., Bengis, R. & Myburgh, J. G. A serological survey for brucellosis in buffalo (Syncerus caffer) in the Kruger National Park. J. S. Afr. Vet. Assoc. 61, 110–111 (1990).CAS 
PubMed 
PubMed Central 

Google Scholar 
79.Fischer-Tenhagen, C., Hamblin, C., Quandt, S., Frö;lich, K. & Frö Lich, K. Serosurvey for selected infectious disease agents in free-ranging black and white rhinoceros in Africa. Journal of Wildlife Diseases 36, 316–323 (2000).80.Caron, A. et al. African buffalo movement and zoonotic disease risk across transfrontier conservation areas Southern Africa. Emerg. Infect. Dis. 22, 277–280 (2016).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
81.Herr, S. & Marshall, C. Brucellosis in free-living African buffalo (Syncerus caffer): A serological survey. Onderstepoort. J. Vet. Res. 48, 133–134 (1981).CAS 
PubMed 
PubMed Central 

Google Scholar 
82.De Vos, V., Van Niekerk, G. A. W. J. & McConell, E. E. A survey of selected bacteriological infections of the Chacma Baboon Papio Ursinus from the Kruger National Park. Koedoe 16, 1–10 (1973).
Google Scholar 
83.Hamblin, C., Anderson, C. E., Jago, M., Mlengeya, T. & Hirji, K. Antibodies to some pathogenic agents in free-living wild species in Tanzania. Epidemiol. Infect. 105, 585–594 (1990).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
84.Assenga, J. A., Matemba, L. E., Muller, S. K., Malakalinga, J. J. & Kazwala, R. R. Epidemiology of Brucella infection in the human, livestock and wildlife interface in the Katavi-Rukwa ecosystem, Tanzania. BMC Vet. Res. 11, 8 (2015).Article 

Google Scholar 
85.Sachs, R., Staak, C. & Groocock, C. M. Serological investigation of brucellosis in game animals in Tanzania. Bull. Epizoo. Dis. Afr. 16, 93–100 (1968).CAS 

Google Scholar 
86.Fyumagwa, R. D., Wambura, P. N., Mellau, L. S. B. & Hoare, R. Seroprevalence of Brucella abortus in buffaloes and wildebeests in the Serengeti ecosystem: A threat to humans and domestic ruminants. Tanzania Vet. J. 26, 2 (2010).
Google Scholar 
87.Matope, G. et al. Evaluation of sensitivity and specificity of RBT, c-ELISA and fluorescence polarisation assay for diagnosis of brucellosis in cattle using latent class analysis. Vet. Immunol. Immunopathol. 141, 58–63 (2011).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
88.Muma, J. B. et al. Serosurvey of Brucella Spp Infection in the Kafue Lechwe (Kobus Leche Kafuensis) of the Kafue Flats in Zambia. J. Wildl. Dis. 46, 1063–1069 (2010).PubMed 
Article 
PubMed Central 

Google Scholar 
89.Waghela, S. Animal brucellosis in Kenya: A review. Bull. Anim. Heal. Prod. Afr. 24, 53–59 (1976).CAS 

Google Scholar 
90.Waghela, S., Karstad, L., Waghela, A. S. & Karstad, L. Antibodies to Brucella Spp among blue wildebeest and African Buffalo in Kenya. J. Wildl. Dis. 22, 189–192 (1986).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
91.Magwedere, K. et al. Brucellae through the food chain: the role of sheep, goats and springbok (Antidorcus marsupialis) as sources of human infections in Namibia. J. South Afr. Vet. Assoc. Van Die Suid-Afrikaanse Veterinere Ver 82, 205–212 (2011).CAS 

Google Scholar 
92.Karesh, W. B. et al. Health evaluation of black-faced impala (Aepyceros melampus petersi) using blood chemistry and serology. J. Zoo. Wildl. Med. 28, 361–367 (1997).CAS 
PubMed 
PubMed Central 

Google Scholar 
93.Cooper, A. C. D. & Carmichael, I. H. The incidence of brucellosis in game in Botswana. Bull. Epizoot. Dis. Afr. 22, 119–124 (1974).CAS 
PubMed 
PubMed Central 

Google Scholar 
94.Thimm, B. Brucellosis in Uganda.pdf. Bull Epizoot Dis Africa 20, 43–56 (1972).95.Tanner, M. et al. Bovine tuberculosis and brucellosis in cattle and african buffalo in the limpopo national park mozambique. Transbound Emerg. Dis. 62, 632–638 (2015).CAS 
PubMed 
Article 
PubMed Central 

Google Scholar 
96.Gomo, C., de Garine-Wichatitsky, M., Caron, A. & Pfukenyi, D. M. Survey of brucellosis at the wildlife-livestock interface on the Zimbabwean side of the Great Limpopo Transfrontier Conservation Area. Trop. Anim. Health Prod. 44, 77–85 (2012).PubMed 
Article 
PubMed Central 

Google Scholar 
97.Herr, S. & Marshall, C. Brucellosis in Free-Living African Buffalo (Syncerus-Caffer)—a Serological Survey. Onderstepoort. J. Vet. Res. 48, 133–134 (1981).CAS 
PubMed 
PubMed Central 

Google Scholar  More

Soil microbial diversity and community structureIn total, 1,334,381 reads were obtained for the bacterial 16S rRNA genes by high-throughput sequencing. After screening these gene sequences with strict criteria (described in “Materials and methods”), 1,061,916 valid sequences were obtained, accounting for 79.6% of the raw reads. Figure 1A shows that the observed richness, Chao1, and Shannon index in the SS (sweet sugarcane) group supported significantly higher richness (P  More

Rural areas of low-to-middle-income countries host most biodiversity hotspots, where interactions between people and wildlife are frequent. These regions have less access to vaccines than do urban centres (Local Burden of Disease Vaccine Coverage Collaborators Nature 589, 415–419; 2021).Given the broad potential range of hosts for SARS-CoV-2, we suggest that vaccinating often-neglected populations around protected areas will reduce the risk of people infecting wildlife and creating secondary reservoirs of disease, and thence risking potential reinfection of humans with new variants. This should be considered after vaccination of priority groups, such as older people and health workers.Vaccinating people who live near felids, non-human primates, bats and other animals protects wildlife and limits ‘reverse spillovers’. Such events have been documented for various human respiratory viruses, for instance in wild great apes in west Africa (S. Köndgen et al. Curr. Biol. 18, 260–264; 2008).Non-standard actors, such as national park authorities or conservation organizations, could help vaccination to reach remote regions. This is called a One Health approach: it protects the health of people, animals and the environment. More

Sample collection and handlingSource of samples to be screenedSouth Africa currently has 72 official points of entry—8 seaports, 10 airports and 54 land border posts10. The DALRRD has border inspectors at most of these points (although staffing levels have varied considerably). DALRRD border inspectors inspect goods and travellers entering the country for plant contaminants. As part of DALRRD’s biosecurity protocol, three types of samples are collected and sent to DALRRD laboratories in Stellenbosch or Pretoria for further investigation (Fig. 1).

1.

Intervention samples. If the border inspector finds or suspects a pest or pathogen in a consignment, he/she will take a sample and send it to one of DALRRD’s diagnostic laboratories. A suspicion of contamination is often the result of quarantine organisms being detected on previous consignments of the same commodity. The imported consignment is detained at the border until laboratory results are completed. Due to the time-sensitivity of such imports, the samples are usually only inspected or tested for the taxa of concern.

2.

Audit samples. As above, these samples are drawn from consignments of plant products for immediate use. However, they are drawn on an ad hoc (haphazard) basis from consignments that show no signs of contamination during border inspections. In the laboratory, these samples are often inspected or tested for multiple taxa.

3.

Post-entry quarantine (PEQ) samples. Plant products for propagation purposes or nursery material (e.g. in vitro plantlets, seedlings, budwood) are shipped in sealed packages and transported directly to DALRRD’s agricultural quarantine facilities. For small consignments (under 50 units), all units in the consignment are tested and inspected by laboratory officials. For larger consignments, random samples are drawn and inspected following a hypergeometric sampling protocol11. Inspection for arthropods and initial examination for micro-organisms takes place in a biosecurity containment facility (see Saccaggi & Pieterse12 for further details). The material is then grown in a dedicated quarantine facility and further testing for pathogens takes place when the plants are in active growth.

Fig. 1Summary of border and laboratory processes associated with each of the three import sample sources included in this dataset, namely post-entry quarantine (PEQ), intervention and audit samples. Solid lines indicate that these processes are always followed, while dashed lines indicate that the process is sometimes followed. PEQ samples are received from plant propagation or nursery material that needs to be quarantined upon arrival. Intervention samples are received from consignments in which the border inspector finds or suspects a pest or pathogen. Audit samples are ad hoc samples drawn from consignments that show no sign of contamination. These sample sources are explained in more detail in the text.Full size imageTaxa inspection, testing and identification methodsAll inspections, testing and identifications are carried out by DALRRD laboratory officials specialised in each taxonomic group. Taxonomic identifications are routinely done by DALRRD officials, taxonomists at the Biosystematics Division of the South African Agricultural Research Council (ARC) or higher education institutions, depending on the expertise available at the time. All recorded identifications in the dataset were retained, regardless of level of identification or biosecurity status of the organism. It should, however, be noted that all organisms found were not always recorded (see below for further explanation).Arthropods (mostly insects and mites) and Molluscs are detected via visual inspection using a stereo-microscope. For these taxa, all organisms detected are recorded. Organisms are most commonly identified morphologically, with molecular identification being performed for certain groups. Identification is performed to the point at which a reasonable phytosanitary decision can be made (i.e. sometimes taxonomic precision is sacrificed for time and/or resource efficiency and logistic reasons). Thus specimens from predatory or saprophytic groups are often only identified to family or genus, while specimens within plant-feeding groups are identified to species where possible.Nematodes are detected by extraction from samples using relevant extraction methods. Saprophytic and predatory nematodes are sometimes noted, but often ignored as they are not considered to be of phytosanitary concern. Plant-feeding nematodes are identified morphologically to species where possible.Fungi and Bacteria are detected visually in the growing plant, as well as by conventional isolation and plating techniques, followed by biochemical tests and/or morphological identification. Some targeted pathogens are detected and identified by molecular techniques such as PCR and DNA sequencing. Saprophytic or secondary fungi or bacteria are sometimes noted, but often not recorded as part of the sample record.Viruses are screened for by immunological techniques, notably ELISA and hardwood and herbaceous indexing. ELISA techniques detect a target virus of concern and give no information as to the presence or absence of other viruses in the sample. Hardwood and herbaceous indexing are used to determine if any graft- or mechanically-transmissible viruses are present in the sample, although these methods cannot be used to determine the viruses’ identity.Phytoplasma screening is done by nested PCR designed to detect any phytoplasma. On specific crops, phytoplasma groups are detected by using targeted PCR methods. If necessary, sequencing of PCR products is used for more specific identification.Data collection and handlingMetadata for samples were recorded by the border inspector before submission to DALRRD’s laboratories. Ideally, he/she recorded geographic origin of the commodity, crop and sample type, date of collection, details of importer and exporter, organisms to test for and any additional observations. However, in practice, this information was not always recorded in full. See Tables 1, 2 and 3 for more details on information included in the dataset. Due to the sensitivity of this kind of trade data, some of the data in the current dataset are grouped or anonymised to protect confidentiality. In particular, import date is only listed as month and year and the names of importers and exporters are removed.Table 1 A summary of information fields and descriptions for each imported sample recorded in the South African plant import dataset used in the datasheet “List of contaminants on SA plant imports 1994–2019.csv”23.Full size tableTable 2 Information fields and descriptions for taxa information associated with contaminant organisms detected on import samples received by South Africa used in the datasheet “Metadata of contaminants on SA plant imports 1994–2019.csv”23.Full size tableTable 3 List of import commodity types used in the datasheet “List of contamiants on SA plant imports 1994–2019.csv”23. The original categories listed by the inspectors were expanded to 30 commodity types based on additional laboratory information and expert experience.Full size tableElectronic databases of samples received by the DALRRD laboratories were maintained by the laboratory staff. These databases were not official departmental databases and therefore did not need to include information relevant to other sections involved in biosecurity. For instance, total number of imports, total size of each consignment, observations of the inspector, details of phytosanitary certificates and release or detention of the consignment were never recorded. The databases also included samples processed by the laboratory for export or for national pest surveys. Partly due to their unofficial status, the databases were transient, with new databases started once software became outdated, the old one became too big or when new categories or information were to be included. For this study, we collated, curated and cross-checked information from nine of these databases, spanning 26 years from 1994 to 2019.Recorded laboratory data varied between taxa and over time and as priorities and understanding of biosecurity changed. In the initial years considered here (ca. 1994–2000), the focus was on pests or pathogens of quarantine importance, i.e. those on the prohibited list. Other organisms found on samples were not consistently recorded and, when they were, they were often recorded in broad groupings (e.g. “saprophytic nematodes”). More recently, there has been a shift towards recording all organisms detected, but this has still not been done consistently [although from ~2005 onwards the officials responsible for arthropods and molluscs have tried to record everything found (DS, MA personal observations)]. Thus prohibited (i.e. quarantine organisms) were always recorded, but the recording of other contaminants was inconsistent.Data clean-up started with collation of all data from the nine databases. Initially, these contained 99,023 records, with 50,655 recorded as imports, 31,163 as exports, 11,004 as surveys with the remaining 6,201 falling into other categories or uncategorised. Only imports were retained, as this was the only category of interest for this study. For some imports, sample information was recorded in one database, while results of inspections/tests for different taxa were recorded in other databases. Thus a single sample could have up to four duplicate records. Each of these was checked individually and collated into one record for the sample. Spelling mistakes, incorrectly recorded information (e.g. information recorded in the wrong field) and missing information were traced back through paper records and corrected wherever possible. If the original data could not be found, these ambiguous records were excluded. After this data clean-up, the dataset comprised a list of 26,291 import records, of which 2,572 resulted from intervention samples (sample source 1 above, Fig. 1), 10,629 were audit samples (sample source 2 above, Fig. 1) and 13,090 were PEQ samples (sample source 3 above, Fig. 1). Data clean-up then continued for the organisms found on the imported samples.Taxon names were extracted and spelling and classification were corrected and/or added by hand. The list of taxa was checked against the Global Biodiversity Information Facility (GBIF)13 using the software package ‘rgbif’14 in Rstudio version 1.3.95915 running R version 4.0.216. This highlighted additional spelling mistakes and provided a taxonomic backbone to work from. The classification of a number of taxa had changed over the years and thus using a common taxonomic backbone was needed for consistency. Some taxa, most notably some mite species, could not be found on GBIF. In these cases, the taxonomy provided by the taxonomist who initially identified the organism was retained. Virus taxonomic information was also not available on GBIF and the database of the International Committee on Taxonomy of Viruses (ICTV) was used17.Species occurrence in South Africa was determined by consulting published species distribution lists. The following data sources were consulted: GBIF13 (accessed 29 July and 03 Aug 2020); CABI Crop Pest Compendiums and Invasive Species Compendium18,19,20; the Catalogue of Life21; animal species checklists published by the South African Biodiversity Institute (SANBI)22; and for any remaining species internet searches were conducted for literature citing distributions (listed in Table 2).In South Africa, lists of organisms prohibited from entering the country have been compiled by DALRRD and the Department of Forestry, Fisheries and the Environment (DFFtE). DFFtE’s list of prohibited species focussed mostly on organisms of environmental concern, although some prohibited organisms were also of agricultural concern, while DALRRD is only concerned with agricultural pests. DALRRD issues import permits for each unique crop, commodity and country combination from which plant products originate. Thus there is no single consolidated quarantine list for South Africa. Furthermore, any quarantine list is not static, but needs to change as species’ distributions, taxonomic revisions or pest status changes. Thus it is very difficult to provide a list of which detected organisms are of quarantine status to South Africa at any given time and particularly in a dataset spanning 26 years. As far as possible, we have indicated the regulatory status of the species in the datasheet “Metadata of contaminants on SA plant imports 1994–2019.csv”23. This regulatory status would have been of critical importance to inform contemporary phytosanitary decisions. However, given that such lists are dynamic and a core aim of presenting these data is to facilitate analyses of future invaders9, it is important to present information on all organisms detected. Moreover, this allows a more comprehensive assessment of the role of different pathways and will facilitate comparisons with other countries. More

This study describes a unique point-based data set for coral reef environments, collected using a photoquadrat survey method published for seagrass environments1. The data set describes the spatial and temporal distribution of benthic community abundance and composition for Heron Reef, a 28 km2 shallow platform reef located in the Capricorn Bunker Group, Southern Great Barrier Reef (GBR), Australia. On average, 3,600 coral reef data points were collected annually over the period 2002 to 2018. Annual data sets were acquired for independent research projects, but the collection methods were consistent. The initial field data collection design was planned to acquire detailed field data to describe the spatial distribution and variability of benthic composition across the study site to assist with calibration and validation of earth observation-based mapping products.To create a map based on earth observation imagery, it is common to use training or calibration data to transform the imagery into a map of surface properties using a supervised algorithm (e.g. multivariate statistical clustering, random forest)2. To report on the accuracy measures of the maps, reference or validation data are contrasted with the output maps3. Hence for calibration and validation purposes, georeferenced field data must be representative of all the features to be mapped and collection should ideally coincide with satellite image acquisition. Many earth observation approaches have been implemented for mapping the benthic communities of Heron Reef4,5,6,7,8,9,10,11,12 and several of these maps are now accessible online6,13,14.Several studies have utilised time series benthic data to analyse changes in benthic community and coral type trends, supporting broad ecological knowledge of coral reef ecosystems such as the Caribbean reef degradation15 and coral cover decline on the GBR16. Similarly, benthic community and coral cover data sets have been identified as important indicators of coral reef health providing the backbone for monitoring and management initiatives around the world17,18.Articles and data sets have been published that describe the benthic community properties of Heron Reef, however, their spatial coverage, number of georeferenced data points, and revisit times are limited19. The time series photoquadrat data sets presented in this paper could be used for further understanding of benthic community distribution, including statistical analysis of trends in coral cover, analysis of changes in benthic community and coral type, or used for testing of other earth observation-based mapping and modelling approaches. Additionally, as our methodology describes machine annotation of the field photoquadrats, it would be possible to reanalyse the photoquadrats with new categories not previously considered important from a biological perspective (e.g. unknown disease or impact, or a specific benthic community type), or for other features (e.g. the counting of sea cucumbers (Holothuroidea sp.)).Detailed analyses of our complete data set may permit a greater understanding of the persistence and/or dynamics of the benthic community at Heron Reef. As such, our ongoing analyses include evaluation of changes in community composition following major impacts such as cyclones, coral bleaching, crown of thorns predation, etc., and additionally, statistical analyses of coral recovery after such impacts. To this degree, these benthic community data sets are invaluable. More

Species occurrence data
We compiled species’ geographic ranges from a combination of datasets. We employed the IUCN Red List of Threatened Species database, which provides geographic range polygons for 8,564 freshwater fish species (~56% of freshwater fish species39), compiled from literature and expert knowledge40. We complemented these ranges with data from Barbarossa et al23, who compiled geographic range polygons for 6,213 freshwater fish species not yet represented in the IUCN dataset, and the Amazonfish dataset41, which provides range maps for 2,406 species occurring in the Amazon basin. We harmonized the species names based on Fishbase (www.fishbase.org)42 and merged the ranges (i.e., union of polygons) from the different datasets to obtain one geographic range per species. We then resampled the range polygons of each species to the 5 arcminutes (~10 km) hydrography of the global hydrological model (see below), with a given species marked as occurring in a cell if ≥ 50% of the cell area overlapped with the species’ polygon. In total, we obtained geographic ranges for 12,934 freshwater fish species, covering ~90% of the known freshwater fish species43. We excluded 1,160 exclusively lentic species because our hydrological model is less adequate for lakes than for rivers, i.e., it does not account for water temperature stratification (see section “Phylogenetic regression on species traits” for an explanation of how habitat information was extracted). Out of the 11,774 (partially or entirely) lotic fish species, we excluded 349 species (~3%) because their occurrence range was smaller than ~1,000 km2 (i.e., ten grid cells), which we considered too small relative to the spatial resolution of the hydrological model (see below). Hence, the analysis was based on 11,425 species in total (Supplementary Figs. 9, 10; a raster layer providing the number of species at each five arcminutes grid cell is available as Supplementary Data 6).
Hydrological data
We employed the Global Hydrological Model (GHM) PCR-GLOBWB20 with a full dynamical two-way coupling to the Dynamical Water temperate model (DynWAT)21 at 5 arcminutes spatial resolution (~10 km at the Equator), to retrieve weekly streamflow and water temperature worldwide20,21. PCR-GLOBWB simulates the vertical water balance between two soil layers and a groundwater layer, with up to four land cover types considered per grid cell. Surface runoff, interflow, and groundwater discharge are routed along the river network using the kinematic wave approximation of the Saint–Venant Equations21 and includes floodplain inundation. Apart from the larger lakes, PCR-GLOBWB includes over 6,000 man-made reservoirs44 as well as the effects of water use for irrigation, livestock, domestic, and industrial sectors. PCR-GLOBWB computes river discharge, river and lake water levels, surface water levels and runoff fluxes (surface runoff, interflow and groundwater discharge). These fluxes are dynamically coupled to DynWAT along with the meteorological forcing, such as air temperature and radiation from the GCMs to compute water temperature. DynWAT thus includes temperature advection, radiation and sensible heating but also ice formation and breakup, thermal mixing and stratification in large water bodies, effects of water abstraction and reservoir operations. We selected this model combination because it allows a full representation of the hydrological cycle (considering also anthropogenic stressors, e.g., water use), it fully integrates water temperature and calculates the hydrological variables on a high-resolution hydrography. The choice of one hydrological model over an ensemble was motivated by the fact that very few GHMs or Land Surface Models calculate water temperature at the spatial resolution desired for this study20,21. The PCR-GLOBWB model setup was similar to Wanders et al.21, with the exception that flow and water temperature were aggregated at the weekly scale to capture the fish species’ tolerance levels to extreme events45.
Species-specific thresholds for extreme flow and water temperature
To assess climate change threats to freshwater fishes, we focused on climate extremes rather than hydrothermal niche characteristics in general, because extremes are more decisive for local extinctions and potential geographic range contractions16,17. We quantified climate extremes using long-term average maximum and minimum water temperature (Tmax, Tmin), maximum and minimum flow (Qmax, Qmin), and the number of zero flow weeks (Qzf), based on the weekly hydrograph and thermograph of the hydrological model. Water temperature is considered the most important physiological threshold for fish species, as mortality of ectothermic species occurs above and below lethal thresholds8,46. Decreases in minimum flow directly affect riffle-pool systems and connectivity between viable habitat patches, leading to a rapid loss of biodiversity47. We included the number of zero-flow weeks because increases in the frequency of dry-spells directly correlates with reduction in diversity and biomass due to the loss of suitable aquatic habitat47. We considered maximum flow because increases in high flow might reduce abundance of young-of-the-year fish by washing away eggs and displacing juveniles and larvae, impeding them from reaching nursery and shelter habitats47,48.
We quantified species-specific thresholds for minimum and maximum weekly flow, maximum number of zero flow weeks and maximum and minimum weekly water temperature based on the present-day distribution of these characteristics within the geographic range of each species, similarly to previous studies45,49,50. To this end, we overlaid the species’ range maps with the weekly flow and water temperature metrics from the output of the hydrological model, calculated for each year and averaged over a 30-years historical period to conform to the standard for climate analyses51,52 (1976–2005, for each GCM employed in the study). We calculated for each 5 arcminutes grid cell the long-term average minimum and maximum weekly flow (Qmin, Qmax, Eqs. (1) and (2)), the long-term average frequency of zero-flow weeks (Qzf, Eq. (3)) and the long-term average minimum and maximum weekly temperature (Tmin, Tmax, Eqs. (4) and (5)), as follows:

$$Q_{mathrm{min}} = frac{{mathop {sum }nolimits_{i = 1}^N {mathrm{min}}({mathrm{Q}}7_i)}}{N}$$
(1)

$$Q_{mathrm{max}} = frac{{mathop {sum }nolimits_{i = 1}^N {mathrm{max}}({mathrm{Q}}7_i)}}{N}$$
(2)

$$Q_{zf} = frac{{mathop {sum }nolimits_{i = 1}^N left{ {j in left{ {1, ldots ,M} right}:{mathrm{q}}7_j = 0} right}_i}}{N}$$
(3)

$$T_{{mathrm{min}}} = frac{{mathop {sum }nolimits_{i = 1}^N {mathrm{min}}({mathrm{T}}7_i)}}{N}$$
(4)

$$T_{{mathrm{max}}} = frac{{mathop {sum }nolimits_{i = 1}^N {mathrm{max}}({mathrm{T}}7_i)}}{N}$$
(5)

where Q7 and T7 are the vectors of weekly streamflow and water temperature values for a given year i, respectively; q7 is the streamflow value for the week j; N is the number of years considered (30 in this case) and M is the number of weeks in a year (~52). We then used the spatial distributions of these values within the range of each species to determine species-specific ‘thresholds’ for each of the variables, defined as the 2.5 percentile of the minimum flow and minimum temperature and the 97.5 percentile of the maximum water temperature and zero flow weeks values. We preferred these to using the absolute minimum and maximum values to reduce the influence of uncertainties and outliers in the threshold definition. Only for maximum flow we used the maximum value across the range, because of the highly right-skewed distribution of flow values within the range of the species. An overview of the thresholds’ distribution is available in Supplementary Fig. 8.
Climate forcing and warming targets
We considered four main future scenarios based on increases of global mean air temperature equal to 1.5, 2.0, 3.2, and 4.5 °C. The global mean temperature increase refers to a 30-years average, in accordance with guidelines for climate analyses51, and with pre-industrial reference set at 1850–190031. To obtain estimates of weekly water temperature and flow for each warming level, we forced the hydrological model with the output from an ensemble of five Global Climate Models (GCMs), each run for four Representative Concentration Pathway (RCP) scenarios, namely RCP 2.6, 4.5, 6.0, and 8.5 (see “Supplementary Methods” for details). Hence, each RCP–GCM combination would reach each warming level at a different point in time, with some of the RCP–GCM combinations not reaching certain warming levels. Consequently, the number of scenarios available differed among warming levels (an overview is provided in Supplementary Table 1). In total we modeled 42 scenarios (one scenario = one GCM–RCP combination at a certain point in the future), including 17 scenarios for 1.5 °C, 15 for 2.0 °C, 7 for 3.2 °C and 3 for 4.5 °C.
Projecting species-specific future climate threats
For each species and each of the 42 scenarios as described in the previous section, we quantified the proportion of the range where projected extremes exceed the present-day values within the species’ range for at least one of the variables. Thus, for each species x we quantified the percentage of geographic range threatened (RT [%]) at each GCM-RCP scenario combination c and for a variable (or group of variables) v as,

$${mathrm{RT}}_{x,c,v} = frac{{{mathrm{AT}}_{x,c,v}}}{{A_x}} cdot 100$$
(6)

where AT is the portion of area threatened [km2] and A is the current geographic range size [km2]. That is, we assessed for all grid cells within the species’ range if a projected minimum or maximum weekly flow would fall below the minimum or above the maximum flow threshold, if there would be a higher number of zero flow weeks than the threshold would allow, or if the minimum or maximum weekly water temperature would be lower than the minimum or higher than the maximum water temperature threshold. The variable-by-variable evaluation allowed us to identify which (groups of) variable(s) contributed to the threat. For simplicity, we grouped the number of zero flow weeks, minimum and maximum weekly flow variables to assess threat imposed by altered flow regimes. Similarly, we grouped threats imposed by amplified minimum and maximum weekly water temperature to assess temperature-related threats. In the aggregated results, a grid-cell is thus flagged as threatened if any of the underlying thresholds is exceeded.
Accounting for dispersal
In general, organisms may adapt to climate change (or escape from future extremes) by moving to more suitable locations53. Accounting for this possibility is challenging due to the uncertainties and data gaps associated with current and future barriers in freshwater systems (e.g., dams, weirs, culverts, sluices)54. In addition, data needed to reliably estimate dispersal ability is still lacking for the majority of the species55. We therefore employed two relatively simple dispersal assumptions in our calculations. Under the “no dispersal” assumption, fishes are restricted to their current geographic range, whereas under the “maximal dispersal” assumption, fishes are assumed to be able to reach any cell within the sub-basin units encompassing their current geograhic range. We defined the sub-basin units by intersecting the physical boundaries of main basins (defined as having an outlet to the sea/internal sink) with the boundaries defined by the freshwater ecoregions of the world, which provide intra-basins divisions based on evolutionary history and additional ecological factors relevant to freshwater fishes22 (Supplementary Fig. 11). Basins smaller than 1,000 km2 were combined with adjacent larger units. In total, we delineated 6,525 sub-basin units (area: µ = 20,376 km2, σ = 90,717 km2) from 10,884 main hydrologic basins and 449 freshwater ecoregions. To model future climate threats under the maximal dispersal assumption, we first expanded the geographic range for the current situation, allowing the species to occupy grid cells within the encompassing sub-basin boundaries if suitable according to the species-specific thresholds. Then we assessed future climate threats for the 42 different scenarios relative to the present-day range plus all cells potentially available to the species within the encompassing sub-basins (excluding cells that would become threatened in the future), as

$${mathrm{RT}}_{x,c,v} = frac{{{mathrm{AT}}_{x,c,v}}}{{A_x + ({mathrm{AE}}_x – {mathrm{AET}}_{x,c,v})}} cdot 100$$
(7)

where AE is the expanded part of the geographic range [km2] and AET is the area threatened within the expanded part of the geographic range [km2].
Aggregation of results
To summarize our results, we first assessed the proportion of species having more than half of their (expanded) geographic range threatened (i.e., exposed to climate extremes beyond current levels within their range) at each warming level. We did this for each GCM-RCP scenario combination and then calculated the mean and standard deviation across the GCM-RCP combinations at each warming level. We further calculated the proportion of species threatened by future climate extremes in each 5 arcminutes (~10 km at the Equator) grid cell for each warming level, as follows:

$${mathrm{PAF}}_{i,w} = median_cleft( {1 – frac{{S_{i,w}}}{{S_{mathrm{i,present}}}}} right)$$
(8)

where PAF represents the potentially affected fraction of species in grid cell i for warming level w, c represents the scenario (i.e., GCM–RCP combination), Si,w represents the number of species for which extremes in water temperature and flow in grid-cell i according to warming level w do not exceed present-day levels within their range, and Si,present represents the number of species in grid cell i. For both numerator and denominator, the species pool for cell i was determined based on the overlap with the (expanded) geographic range maps (see “Species occurrence data” and “Accounting for dispersal”). We used the median across the GCM–RCP combinations rather than the mean because the data showed skewed distributions. Finally, we averaged the grid-specific proportions of species affected over main basins with an outlet to the ocean/sea or internal sink (e.g., lake), as follows:

$$overline {{mathrm{PAF}}} _{x,w} = frac{{mathop {sum }nolimits_{i = 1}^I {mathrm{PAF}}_{i,w}}}{{I_x}}$$
(9)

where Ix represents the number of grid cells within the watershed x.
Phylogenetic regression on species traits
We performed phylogenetic regression to relate the threat level of each species, quantified as the proportion of the geographic range exposed to future climate extremes beyond current levels within the range (see Eqs. (6) and (7)), to a number of potentially relevant species characteristics, while accounting for the non-independence of observations due to phylogenetic relatedness among species56. We established a phylogenetic regression model per warming level and dispersal scenario (i.e., eight models in total, based on four warming levels times two dispersal assumptions). As species characteristics, we included initial range size (in km2), body length (in cm), climate zone, trophic group and habitat type, as these traits may influence species’ responses to (anthropogenic) environmental change8,30,57,58. We further included IUCN Red List category to evaluate the extent to which current threat status is indicative of potential impacts of future climate change, and commercial importance to evaluate implications of potential extirpations for fisheries. We overlaid each species’ geographic range with the historic Köppen–Geiger climate categories to obtain the main climate zone per species (i.e., capital letter of the climate classification)59. Species falling into multiple climate categories were assigned the climate zone with the largest overlap. We retrieved information on threat status from IUCN40 and on taxonomy from Fishbase42. We used the IUCN and Fishbase data also to gather a list of potential habitats for each species. For the species represented within the IUCN dataset, we classified species as lotic if they were associated with habitats containing at least one of the words “river”, “stream”, “creek”, “canal”, “channel”, “delta”, “estuaries”, and as lentic if the habitat descriptions contained at least one of the words “lake”,“pool”,“bog”,“swamp”,“pond”. For the remaining species, we extracted information on habitat from Fishbase, where we used the highest level of aggregation of habitat types to classify species found in lakes as lentic and species found in rivers as lotic. We classified species occurring in both streams and lakes as lotic-lentic and labeled species found in both freshwater and marine environments as lotic-marine. Further, we retrieved data from Fishbase on maximum body length and commercial importance42. From the same database we also retrieved trophic level values and aggregated them into Carnivore (trophic level >2.79), Omnivore (2.19  More

The Amazon rainforest of French Guiana constantly buzzes and hums, but I keep my focus on the trees. In this picture, taken in November 2020 — the most recent time I was there — I’m walking through a dense forest at the Paracou research station near the coastal town of Kourou. I’m looking at drone pictures of the canopy and working out how each trunk fits into the puzzle. It sounds easy, but the forest is extremely complicated. Even with binoculars and close attention to detail, it’s hard to work out which trunk connects to a particular patch of green when you’re looking at it from an aerial view.
My project is part of a bigger effort to understand the forest’s productivity and rhythms. Of the 750 or so woody tree species in the area, many are deciduous. But unlike trees in temperate climes, which shed leaves in autumn, these follow their own schedules. With drones and LIDAR — a mapping system that uses ultraviolet lasers — we can track the trees at a much larger scale than we ever could before. Observations from the ground help to fill out the picture.
The Amazon rainforest, the largest and most biodiverse forest in the world, stores a huge amount of carbon. The great fear is that climate change could transform Amazonia into a drier, savanna-like ecosystem, which could release incredible amounts of carbon into the atmosphere. Understanding the forest’s carbon flows can help us to predict how the whole system will respond to climate change.
As you walk through the forest, there’s a constant chorus of birds, with squawking parrots and hummingbirds that violently zoom around like a golden snitch, a fast ball in quidditch, a sport in the Harry Potter series. I cover myself in the insect repellent DEET to ward off mosquitos, but a few ticks still crawl on me. I sweat constantly in the heat and humidity, and my clothes never fully dry. At night, I sleep in a hammock under a tin roof that pings in the rain.
I can’t wait to go back. More

Morphology and phenotypic plasticity
The morphological features of the investigated colepid strains differed from those described for C. hirtus, C. spetai and even for N. nolandi1 (Fig. 1). Characteristics that matched the descriptions were the ciliate cell length and width, the barrel-shaped cell (except for strain CIL-2017/7, which was pear-shaped and strain CCAP 1613/15 that had a cylindrical shape), a number of six armor tiers, the structure of the armor tiers (hirtus-type or nolandia-type, respectively), and one caudal cilium (Table S2). Variations (CV > 20%) were found (i) in the number of plate windows in the posterior/anterior main plates even within individual cells, and (ii) in the presence/absence of anterior and posterior spines (Tables 1 + S2, Fig. 1). This phenotypic plasticity of the ciliate could also be observed in freshly collected Coleps specimens and was therefore not an artifact resulting from cultivation conditions (Fig. 1A). Wickham and Gugenberger43 hypothesized that the formation of the spines was a response to grazing pressure on C. hirtus; however, this could not be confirmed by respective experiments. Nevertheless, spineless specimens of C. hirtus have obviously been found before44,45,46,47. Luckily, we were able to investigate two strains (CCAP 1613/1 and CCAP 1613/2) that had been kept in the CCAP culture collection since the 1950ies and the 1960ies and which did not bear any spines or symbionts and could be clearly assigned to C. hirtus (Fig. 1T–V). These observations suggest that without predation pressure, colepid ciliates probably do not need to synthesize spines avoiding ingestion by a predator.
The presence/absence of green algal endosymbionts, one of the diagnostic features for the discrimination among C. hirtus subspecies and C. spetai, was also not a stable feature (Table S2). Under culture conditions, some strains lost their endosymbionts completely, other strains consisted of symbiotic and aposymbiotic individuals, and some strains showed only symbiont-bearing individuals (e.g., CCAP 1613/5 and CIL-2017/6). This indicates that the symbiosis is facultative and might be probably influenced by cultivation or environmental conditions (presumably, though not tested, food availability). Consequently, the morphological separation of C. hirtus into the two subspecies may no longer be valid. We clearly demonstrated that the morphological features used for species descriptions can vary and have severe consequences for colepid species identification. Moreover, even the strains belonging to the groups 1 and 2 discovered by the phylogenetic analyses (Fig. 2) cannot discriminate morphotypes because they can neither be assigned to a certain cell morphology nor to the possession of algal endosymbionts. This questions the traditional morphology-based taxonomy. The separation of Coleps hirtus hirtus, C. hirtus viridis and C. spetai, which Foissner et al.1 differentiated by the presence of zoochlorellae in the latter two species and the number of windows in the armor plates, could not be supported by our analyses. C. hirtus viridis was originally described by Ehrenberg48,49 as C. viridis and later transferred as synonym of C. hirtus by Kahl50 based on almost identical morphological features. However, Foissner22 described C. spetai for the green Coleps because of the morphological discrepancies to the Ehrenberg’s C. viridis (presence of only 11 windows per plate row and smaller cell size in C. viridis; see Table 1 for comparison). Our study has clearly demonstrated that most of the morphological features are variable and the limits for species separation were too narrow. Therefore, we propose the re-establishment of C. viridis for group 1 and C. hirtus for group 2, both with emended descriptions as follows. Considering our findings, the morphological descriptions of C. spetai, C. hirtus viridis and C. hirtus hirtus cannot be applied for (sub-) species separation any more. Consequently, we deal with a cryptic species complex, i.e., two genetically different groups that are fused in a highly variable morphotype including features of all three (sub-) species. To solve this taxonomic problem, two possible scenarios can be proposed: (1) We merge the three morphotypes under C. hirtus, the type species of Coleps. As a consequence, two new species needed to be proposed for both groups 1 and 2, which could be done following the suggestion of Sonneborn51 for the P. aurelia-complex. However, Sonneborn based his new descriptions on results of mating experiments, which are not applicable for Coleps here because conjugations have not been reported and the conditions for the induction of sexual reproduction are unknown. (2) To avoid confusion by introducing new species names, we propose keeping the already existing names, i.e., C. viridis for group 1 and C. hirtus for group 2 including the synonyms (see below).
Clonal cultures of both genetically varying Coleps groups have been deposited in the CCAP culture collection. Future studies may therefore be able to investigate, for example, sibling among strains or predator-prey experiments revealing spine- or wing-formation.
Coleps viridis Ehrenberg 1831 (printed 1832), Abh. Königl. Akad. Wiss. Berlin 1832: 101.
Synonym: Coleps spetai Foissner 1984, Stapfia 12: 21-22, Fig. 7, SP: 1984/10 and 1984/11 (lectotype designated here deposited in LI, see Aescht 2008: Denisia 23: 179), Coleps hirtus sensu Kahl 1930, Tierwelt Deutschlands 18: 134.
Diagnosis: Differed from other colepid ciliates by their SSU and ITS rDNA sequences (MT253680).
Lectotype (designated here): Fig. II, Tab. XXXIII, 3 in Ehrenberg 1838, Infusionsthierchen als vollkommene Organismen, p. 314.
Improved Description (specifications in brackets apply to our reference strain CCAP 1613/7): Coleps with conspicuous armor composed of six tiers with plate windows of the hirtus-type. With or without green algal endosymbionts. Cell size 44–63 × 21–35 μm (52–54 × 35–36 μm). Total number of windows in length rows 12–16 (14–16), number of windows of anterior primary plates 3–6 (4–6), number of windows of anterior secondary plates 2–3 (2), number of windows of posterior primary plates 4–5 (4–5), number of windows of posterior secondary plates 2–3 (2–3). One caudal cilium (1). With 0-2 anterior (0–1) and 0–5 posterior (1–4) spines, respectively.
Reference material (designated here for HTS approaches): The reference strain CCAP 1613/7 permanently cryopreserved at CCAP in a metabolically inactive stage.
Locality of reference strain: Plankton of Lake Mondsee, Upper Austria, Austria (47° 50′ N, 13° 23′ E).
Coleps hirtus (O.F. Müller) Nitzsch ex Ersch & Gruber 1827, Allgemeine Encyclopädie der Wissenschaften und Künste 16: 69, NT (proposed by Foissner 1984, Stapfia 12: 22, fig. 8): 1984/12 and 1984/13 (LI, in Aescht 2008: Denisia 23: 159).
Protonym: Cercaria hirta O.F. Müller 1786, Animalcula Infusoria: 128, tab. XIX, fig. 17, 18 (lectotype designated here).
Diagnosis: Differed from other colepid ciliates by their SSU and ITS rDNA sequences (MT253687).
Improved Description: Coleps with spiny armor composed of six tiers with plates of the hirtus-type. Without green algal endosymbionts. Cell size 42–52 × 23–28 μm. Total number of windows in length rows 12-13, number of windows of anterior primary plates 3-5, number of windows of anterior secondary plates 2, number of windows of posterior primary plates 4-5, number of windows of posterior secondary plates 2. One caudal cilium. Without anterior and 1-4 posterior spines, respectively.
Reference strain (designated here for HTS approaches): The strain CCAP 1613/14 permanently cryopreserved at CCAP in a metabolically inactive stage.
Locality of reference strain: Plankton of Lake Piburg, Tyrol, Austria (47° 11′ N, 10° 53′ E).
Molecular phylogeny of the Colepidae (Prostomatea)
The colepids belonging to the Prostomatea form a monophyletic lineage in the phylogenetic analyses of SSU rDNA sequences (Fig. 2). Mixotrophic as well as heterotrophic Coleps strains that resembled C. hirtus and C. spetai clustered in group 1 whereas group 2 included only two specimens which were identified as C. hirtus. These findings confirm the results of Barth et al.29 with one exception. The authors found a clear separation into mixotrophic and heterotrophic species, which were therefore assigned to a C. spetai-(with endosymbionts) and a C. hirtus-group (without endosymbionts), respectively. Despite the difficulties of identifying these species by morphology, both groups clearly differed in their SSU and ITS rDNA sequences (Fig. 3). The ITS-2/CBC approach introduced for green algae (details in Darienko et al.52) clearly demonstrated that both groups represented two separate ciliate species from a molecular point of view, which was also confirmed by analyses of the V9 region of the SSU, a region commonly used for metabarcoding (Figs. 4 and 5).
Our study also confirmed the findings of Chen et al.7, Lu et al.9, and Moon et al.28, showing that the generic concept of colepid ciliates needs to be revised. None of the genera represented by more than one species is monophyletic. For example, the three species of Nolandia belonged to separate lineages. Nolandia nolandi was a sister to our studied strains, whereas both other species were closely related to taxa of Apocoleps, Pinacocoleps, and Tiarina (Fig. 2). The genus Levicoleps and Coleps amphacanthus formed a monophyletic clade representing another example that the generic conception is artificial and needs to be revised. However, to provide a new generic concept of colepid ciliates, it is necessary to study more of the described species by using an integrative approach including experimental approaches on, e.g., the formation of spines. For example, we clearly demonstrated that one key feature, which is the presence/absence of anterior/posterior spines, is highly variable and can therefore not be used to separate colepid genera as indicated by Foissner et al.12 (Fig. 1). There is a need for more experimental studies with colepids belonging to the Cyclidium viridis and C. hirtus morphotype. Therefore, we deposited all clones used in this study in the CCAP culture collection. One option would be to incorporate all species into one genus, i.e., Coleps in revised form.
Endosymbiosis in Coleps
Some strains of Coleps are known to bear green algal endosymbionts1. These green algae have Chlorella-like morphology (Fig. 6) and were identified as Micractinium conductrix (Fig. 7). So far, this alga was only known as endosymbiont of the ciliate Paramecium bursaria34. All green algal endosymbionts of Coleps harbored this Micractinium species. In contrast, Pröschold et al.34 found that one ciliate strain identified as C. hirtus viridis had Chlorella vulgaris as endosymbiont (the algae has been deposited in the Culture Collection of Algae and Protozoa under the number CCAP 211/111). Unfortunately, this ciliate strain is not available anymore53.
Ecology and distribution
For limnological studies, the preservation with Bouin’s solution and QPS is an appropriate method for quantifying and identifying ciliate species in environmental samples54. However, the quality in characterization of ciliates at the species level is sometimes limited as, in case of Coleps, the characteristic armored calcium carbonate plates are dissolved by the acidified fixation solution. Therefore, in our study, we could only distinguish between algal-bearing (mixotrophic) and non-algal-bearing (heterotrophic) Coleps. Despite that limitation, we could clearly see that the heterotrophic ones were only found in the deepest zones of both lakes (Fig. 8A). Not surprisingly, Coleps is often observed in nutrient- and ion-rich and also oxygen-depleted freshwater habitats or areas, e.g., sulfurous and crater lakes1,5,6,27,55,56 or even in the sludge of wastewater treatment plants57. Mixotrophic individuals of Coleps were mainly found in the upper layers of both lakes, whereas in Lake Mondsee we could also detect specimens down to 40 m depth (Fig. 8). In contrast to the mero- and monomictic Lake Zurich4,10,58,59, Lake Mondsee is holo- and dimictic60. During mixis events, algal-bearing Coleps specimens can be transferred passively from the upper layers into the deeper zones and vice versa. Although morphotype countings and HTS analyses reads matched quite well, we found discrepancies that have already been discussed before10,61 (Fig. S2).
Biogeographic aspects (haplotype network)
Our metabarcoding approach showed that C. viridis was found in both lakes as a common ciliate (Fig. S2). In contrast, C. hirtus could not be detected during the sampling period. To obtain more information about the distribution of both species, we used the BLASTn search algorithm62 (100 coverage, >97% identity) for the V4 and the V9 regions of the SSU and the ITS-2 sequences. No records using the V9 and the ITS-2 approaches could be discovered in GenBank, but 25 reference sequences using the V4 (Table S3). Together with the newly sequenced strains, we therefore constructed a V4 haplotype network (Fig. 10). Both groups are obviously widely distributed and subdivided into five (group 1) and four (group 2) haplotypes, respectively. All reference sequences were collected from freshwater habitats except for two marine records63 (EU446361 and EU446396; Mediterranean Sea) and showed no geographical preferences.
Figure 10

TCS haplotype network inferred from V4 sequences of Coleps viridis and C. hirtus. This network was inferred using the algorithm described by Clement et al.64,65. Sequence nodes corresponding to samples collected from different geographical regions and from different habitats.

Full size image

Co-occurrence networks
In the sub-networks of C. viridis in both lakes, we found several significant correlations that pointed to either potential prey items, e.g., diverse flagellated autotrophic or heterotrophic protists or co-occurring ciliates (Fig. 9). Also, the smaller ciliates such as Cinetochilum margaritaceum or Cyclidium glaucoma may as well be considered as food for the omnivorous C. viridis (for a compilation of the food spectrum; see Foissner et al.1). However, we identified the endosymbiont M. conductrix and its host C. viridis from both sub-networks of Lake Mondsee but not of Lake Zurich (Fig. 9). Despite this result, we want to point out that we may probably not find M. conductrix free-living in a water body because outside their ciliate host the algae were immediately attacked and killed by so-called Chlorella-viruses66. Therefore, the HTS-detection of M. conductrix was probably only together with a host ciliate. This might further explain why the green algae were detected in Lake Mondsee even in the aphotic 40 m zone where photosynthesis was impossible and individuals probably passively transferred into the deeper area by lake mixis.
Outlook
As demonstrated in our study, the combination of traditional morphological investigations, which includes the phenotypic plasticity of the cloned strains, and modern molecular analyses using both SSU and ITS sequencing as well as HTS approaches advise a taxonomic revision of the genus Coleps. This comprehensive and integrative approach is also applicable for other ciliate species and genera and will provide new insights into the ecology and evolution of this important group of protists.
Experimental procedures
Study sites, lake sampling and origin of the Coleps strains
Our main study sites were Lake Mondsee (Austria) and Lake Zurich (Switzerland), two pre-alpine oligo-mesotrophic lakes that were sampled at the deepest point of each lake (Table S4). Water samples were taken monthly from June 2016 through May 2017 over the whole water column and additionally biweekly at two main depths, i.e., 5 m in both lakes, 40 m in Lake Mondsee, and 120 m in Lake Zurich, respectively. A 5-L-Ruttner water sampler was used for Lake Zurich and a 10-L-Schindler-Patalas sampler (both from Uwitec, Austria) for Lake Mondsee. Twelve Coleps strains were isolated from Lake Mondsee and one from Lake Zurich. Another six clones could be obtained either from already successfully cultivated own strains, fresh isolates or from culture collections. Detailed information about sampling sites, dates and strain numbers is given in Table S2.
Seasonal and spatial distribution and abundance
For quantification, subsamples (200-300 mL) were preserved with Bouin’s solution (5% f.c.) containing 15 parts of picric acid, 5 parts of formaldehyde (37%) and 1 part of glacial acetic acid54. The samples were filtered through 0.8 μm cellulose nitrate filters (Sartorius, Germany) equipped with counting grids. The ciliates were stained following the protocol of the quantitative protargol staining (QPS) method after Skibbe54 with slight modifications after Pfister et al.67. The permanent slides were analyzed by light microscopy up to 1600x magnification with a Zeiss Axio Imager.M1 and an Olympus BX51 microscope. For identification of Coleps and Nolandia cells, the identification key of Foissner et al.1 was used. Microphotographs were taken with a ProgRes C14 plus camera using the ProgRes Capture Pro imaging system (version 2.9.0.1, Jenoptik, Jena, Germany).
Cloning, identification and cultivation of ciliates and endosymbionts
Single cells of Coleps were isolated and washed using the Pasteur pipette method68. The isolated strains were cultivated in 400 μl modified Woods Hole medium69 (MWC; modified) and Volvic mineral water in a mixture of 5:1 and with the addition of 10 μl of an algal culture (Cryptomonas sp., strain SAG 26.80) as food in microtiter plates. These clonal cultures were transferred into larger volumes after successful enrichment. All cultures were maintained at 15–21 °C under a light: dark cycle of 12:12 h (photon flux rate up 50 mol m−2 s−1).
For the isolation of their green algal endosymbionts, single ciliates were washed again and transferred into fresh MWC medium. After starvation and digestion of any food, after approx. 24 hrs, cells were washed again and the ciliates transferred onto agar plates containing Basal Medium with Beef Extract (ESFl; medium 1a in Schlösser70). Before placement of the ciliates onto agar plates, 50 μm of an antibiotic mix (mixture of 1% penicillin G, 0.25% streptomycin, and 0.25% chloramphenicol) were added to prevent bacterial growth. The agar plates were kept under the same conditions as described. After growth (6–8 weeks), the algal colonies were transferred onto agar slopes (1.5%) containing ESFl medium and kept under the described culture conditions.
For light microscopic investigations of the algae, Olympus BX51 and BX60 microscopes (equipped with Nomarski DIC optics) were used. Microphotographs were taken with a ProgRes C14 plus camera using the ProgRes Capture Pro imaging system (version 2.9.0.1, Jenoptik, Jena, Germany).
PCR, sequencing and phylogenetic methods
Single-cell PCR was used to obtain the sequences of the Coleps strains. Before PCR amplification, single cells of Coleps were washed as described above. After starvation followed by additional washing steps, cells were transferred into 5 μm sterile water in PCR tubes and the prepared PCR mastermix containing the primers EAF3 and ITS055R71 was added. After this primary PCR amplification and subsequent PCR purification, a nested PCR was conducted using the primer combinations EAF3/N1400R and N920F/ITS055R71.
The sequences of the Coleps strains were aligned according to their secondary structures of the SSU and ITS rDNA (see detailed folding protocol described in Darienko et al.52) and included into two data sets: (i) 34 SSU rDNA sequences (1,750 bp) of representatives of all members of the Prostomatea and (ii) 19 ITS rDNA sequences (538 bp) of the investigated strains. Genomic DNA of the green algae was extracted using the DNeasy Plant Mini Kit (Qiagen GmbH, Hilden, Germany). The SSU and ITS rDNA were amplified using the Taq PCR Mastermix Kit (Qiagen GmbH, Hilden, Germany) with the primers EAF3 and ITS055R. The SSU and ITS rDNA sequences of the isolated green algae (aligned according to the secondary structures) were included into a data set of 31 sequences (2,604 bp) of representatives of the Chlorellaceae (Trebouxiophyceae).
GenBank accession numbers of all newly deposited sequences can be found in Table S2 and in Fig. 7, respectively. For the phylogenetic analyses, the datasets with unambiguously aligned base positions were used. To test which evolutionary model fit best for both data sets, we calculated the log-likelihood values of 56 models using Modeltest 3.772 and the best models according to the Akaike criterion by Modeltest were chosen for the analyses. The settings of the best models are given in the figure legends. The following methods were used for the phylogenetic analyses: distance, maximum parsimony, maximum likelihood, and Bayesian inference. Programs used included PAUP version 4.0b16473, and MrBayes version 3.2.374.
The secondary structures were folded using the software mfold42, which uses the thermodynamic model (minimal energy) for RNA folding.
Haplotype networks
The haplotypes of the V4 region were identified among the groups of Coleps (see Fig. S1). The present haplotypes and the metadata (geographical origin and habitat) of each strain belonging to the different haplotypes are given in Table S3. To establish an overview on the distribution of the Coleps groups, the V4 haplotypes were used for a BLASTn search62 (100% coverage, >97% identity). To construct the haplotype networks, we used the TCS network tool64,65 implemented in PopART75.
High-throughput sequencing of the V9 18S rDNA region and subsequent bioinformatic analyses
On each sampling date, water samples for a high-throughput sequencing approach (HTS) were taken in depths of 5 m and 40 m at Lake Mondsee and 5 m and 120 m depths in Lake Zurich. DNA extraction, amplification of the V9 SSU rDNA, HTS and quality filtering of the obtained raw reads was conducted as described in Pitsch et al.10. After quality filtering, all remaining reads were subjected to a two-level clustering strategy76. In the first level, replicated reads were clustered in SWARM version 2.2.2 using d=177. In the second level, the representative sequences of all SWARM OTUs were subjected to pairwise sequence alignments in VSEARCH version 2.11.078 to construct sequence similarity networks at 97% sequence similarity. The network sequence clusters (NSCs) resulting from the second level of clustering were then taxonomically assigned by running BLASTn analyses against NCBI’s GenBank flat-file release version 230.0 and the Coleps SSU sequences obtained from single-cell sequencing. Network sequence clusters were assigned to Coleps, if the closest BLAST hit of the NSC representative sequence was a Coleps reference sequence. Furthermore, the NSC representative sequence had to share a fragment of at least 48 consecutive nucleotides and at least 90% sequence similarity to a reference sequence in order to be assigned to Coleps.
Co-occurrence networks
With the protist community data matrix resulting from HTS, we further conducted co-occurrence network analyses to assess biotic and abiotic interactions of Coleps. For each lake and depth, we ran network analyses with NetworkNullHPC (https://github.com/lentendu/NetworkNullHPC) following the null model strategy developed by Connor et al.79. This strategy was especially designed for dealing with HTS datasets and allows for inferring statistically significant correlations between NSCs while minimizing false positive correlation signals. We screened the resulting networks for Coleps nodes and extracted their subnetworks including all directly neighbouring co-occurrence partners as well as all edges between Coleps and its neighbours and the neighbours themselves. More