Upscaling biodiversity monitoring: Metabarcoding estimates 31,846 insect species from Malaise traps across Germany

Mol Ecol Resour. 2025;25:e14023.   | 1 of 15 https://doi.org/10.1111/1755-0998.14023 wileyonlinelibrary.com/journal/men Received:28December2023 | Revised:5September2024 | Accepted:12September2024 DOI: 10.1111/1755-0998.14023 RESOURCE ARTICLE Upscaling biodiversity monitoring: Metabarcoding estimates 31,846 insect species from Malaise traps across Germany Dominik Buchner1 | James S. Sinclair2 | Manfred Ayasse3 | Arne J. Beermann1,4 | Jörn Buse5 | Frank Dziock6 | Julian Enss4,7,8 | Mark Frenzel9 | Thomas Hörren7 | Yuanheng Li1 | Michael T. Monaghan10,11 | Carsten Morkel12 | Jörg Müller13,14 | Steffen U. Pauls15,16,17 | Ronny Richter18,19 | Tobias Scharnweber20 | Martin Sorg7 | Stefan Stoll8,21 | Sönke Twietmeyer22 | Wolfgang W. Weisser23 | Benedikt Wiggering24 | Martin Wilmking20 | Gerhard Zotz25 | Mark O. Gessner26,27 | Peter Haase2,4,8 | Florian Leese1,4 This is an open access article under the terms of the CreativeCommonsAttribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. ©2024TheAuthor(s).Molecular Ecology ResourcespublishedbyJohnWiley&SonsLtd. Peter Haase and Florian Leese equally contributing senior authors. For affiliations refer to page 12. Correspondence DominikBuchner,AquaticEcosystem Research,UniversityofDuisburgEssen, Universitätsstraße5,45141Essen, Germany. Email:[email protected] Funding information Hessisches Landesamt für Naturschutz, UmweltundGeologie;EUHorizon 2020projecteLTERPLUS,Grant/Award Number: 871128; LandesOffensive zurEntwicklungWissenschaftlich- ökonomischerExzellenzoftheGerman federalStateofHesse Handling Editor: Carla Martins Lopes Abstract Mitigatingongoinglossesofinsectsandtheirkeyfunctions(e.g.pollination)requires tracking largescale and longterm community changes. However, doing so has been hindered by the high diversity of insect species that requires prohibitively high investments of time, funding and taxonomic expertise when addressed with conventional tools. Here, we show that these concerns can be addressed through a comprehensive,scalableandcost-efficientDNAmetabarcodingworkflow.Weuse1815 samples from 75 Malaise traps across Germany from 2019 and 2020 to demonstrate how metabarcoding can be incorporated into largescale insect monitoring networks for less than 50 € per sample, including supplies, labour and maintenance. We validated thedetectedspeciesusingtwopubliclyavailabledatabases(GBOLandGBIF)andthe judgementoftaxonomicexperts.Withanaverageof1.4 Msequencereadspersample we uncovered 10,803 validated insect species, of which 83.9% were represented by asingleOperationalTaxonomicUnit(OTU).Weestimatedanother21,043plausible species, which we argue either lack a reference barcode or are undescribed. The total of 31,846speciesissimilartothenumberofinsectspeciesknownforGermany(~35,500). Because Malaise traps capture only a subset of insects, our approach identified many species likely unknown from Germany or new to science. Our reproducible workflow (~80%OTU-similarityamongyears)providesablueprintforlarge-scalebiodiversity monitoring of insects and other biodiversity components in near real time. 2 of 15 | BUCHNER et al. 1 | INTRODUCTION InsectsarethemostdiverseanimaltaxonomicgrouponEarthand contribute to many essential ecosystem processes and services, such as pollination, nutrient cycling and organic matter decomposition(Cardosoetal.,2020).However,insectpopulationsare declining (Wagner, 2020) and our ability to mitigate these declines is hindered by poor understanding of spatial distributions, habitat requirements, biotic interactions, dynamics and even the overallnumberofextantspecies.About1millioninsectspecies have been described to date, but recent estimates of total species numbers stand at about 5.5 million (Stork, 2018) or even more (IPBES, 2019). Surprisingly, many new species are being reported even for very wellstudied and generally speciespoor areas with a long history of entomological research. For example,basedon4000speciescaughtwithMalaisetrapsinSweden, Karlssonetal.(2020)reportedalmost700insectspeciesnewto scienceand ~ 1300newtoSweden.Similarly,inaDNAbarcoding analysis of about 62,000 specimens collected with Malaise traps, Chimenoetal.(2022)estimatedover2000dipteranspeciesnew to Germany, raising the total number of 33,341 known insect speciesinthecountry(Klausnitzer,2005)toapproximately35,500. Thus, while many new insect species are being continuously reported, the vast majority still remain unknown. Akeyconstrainttoclosingthecurrentinformationgaponinsects is the vast number of species and specimenrich samples that must be processed. Terrestrial and aquatic monitoring methods, like Malaise, canopy or light traps, as well as samples collected with nets from freshwater habitats, can collect thousands of specimens (e.g.Habeletal.,2023; Karlsson et al., 2020; Resh & Jackson, 1993), manyofwhicharesmallanddifficulttoidentifyevenforexperts, resultingintaxonomicneglect(Srivathsanetal.,2023).Nationaland international monitoring networks, such as the Global Malaise Trap programme(Geigeretal.,2016),BioScan(Hobern,2021),LifePlan (www. helsi nki. fi/ en/ proje cts/ lifeplan)ortheSwedishMalaiseTrap programme(Karlssonetal.,2020),havestartedcoordinatedlarge- scale insect sampling initiatives based on standardized traps and procedures (e.g. Hallmann et al., 2017). However, classical taxonomic analyses of these samples cannot keep pace with the rate of collection, particularly given the low and globally declining number oftaxonomicexperts(EuropeanCommission,2022).Whiletheimportanceoftaxonomicexpertiseremainsundisputed,thereisaclear need for alternative methods to assess insect species diversity both efficientlyandreliably(Chuaetal.,2023; van Klink et al., 2022). DNAmetabarcodingisonekeymethodforassessingspecimen- rich samples. Following more than a decade of research and trial applications, metabarcoding has now reached a high technology readiness level, presenting a promising solution for examining a fuller range of insect biodiversity in largescale monitoring programmes. A range of suitable protocols for specimen collection, processingand dataanalysis are now available (Buchner, Macher, et al., 2021; Montgomery et al., 2021).Thisincludessuitableprimer pairs(Braukmannetal.,2019;Elbrechtetal.,2019)andinsectbarcodereferencedataonBOLD(Ratnasingham&Hebert,2007).The keystrengthofDNAmetabarcodingisthatitrapidlydeliverstaxonomicallyhighlyresolvedtaxalists,whereasobtaiquantitativeinformationonspeciesabundanceorbiomassremainsachallenge(Sickel et al., 2023). Despite its potential, implementing DNA metabarcoding for largescale and longterm insect biodiversity monitoring programmes is still in its infacy. There are several reasons. In particular sample throughput is still constrained by the need for significant manual labour,expensiveDNAkitsandreagents,anddifficultiesinaccessing informationonlabandanalysisprocedures(McGeeetal.,2019)and thelackofformalstandardsfortheprocedures(Chuaetal.,2023). Alltheseaspectspresentroadblockstolarge-scaleimplementation. Furthermore, incomplete and partly inconsistent reference databases impact the accuracy and quantity of specieslevel assignments and thus the completeness and validity of the resulting species lists (Chuaetal.,2023).Thisproblemremainsparticularlypervasivefor highly diverse groups like Diptera and Hymenoptera that are underrepresented in reference libraries because of those species that are difficult to distinguish based on morphological criteria. The unknown (i.e. undescribed) species in these poorly explored groups areoftenreferredtoas‘darktaxa’(Hartop,2021).However,inthe absence of reference barcodes for species, or even when formal speciesdescriptionsarelacking,DNAmetabarcodingcanstillovercome theselimitationsusinggeneticdistancethresholdstoapproximate entities that roughly reflect species, such as molecular Operational TaxonomicUnits(OTUs)orBarcodeIndexNumbers(Ratnasingham & Hebert, 2013)(BINs).Evenintheabsenceofspeciesindatabases, such distance-based entities can approximate species numbers andhavebeenappliedinecologicalandecotoxicologicalresearch for many years (Beermann et al., 2018; Hoppeler et al., 2016; Sturmbaueretal.,1999). DNAmetabarcodinganalysestodatehavebeenlimitedtoeither a specific region within a country (e.g. Geiger et al., 2016; Habel et al., 2023; Uhler et al., 2021),shorttimespans(e.g.Huang et al., 2022; Li et al., 2023)orspecifictaxonomicgroups(e.g.Huang et al., 2022).Nonehasquantifiedthefullextentofinsectdiversity— includingdarktaxa—atthewhole-countryscaleandthroughtime. Thus,thepresentstudyfillsthisgapbypresentingarobustDNA metabarcoding workflow for application to largescale insect monitoring programmes, combined with a new multilevel procedure to assess the validity of species records. We analysed 1815 Malaise trap samples collected in 2019 and 2020 from 75 individual traps KEYWORDS biodiversitymonitoring,DNAmetabarcoding,insectdiversity,Malaisetrap 17550998, 2025, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.14023 by Capes, Wiley Online Library on [10/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 3 of 15 BUCHNER et al. acrossGermany(Figure 1).Ourprincipalaimwasto(1)presentthe workflow as a resource for use by other researchers by providing detailed and publicly available laboratory procedures and bioinformaticprogrammes.Additionally,we(2)quantifiedthetimeandcost investment required per sample for this workflow to evaluate its affordability, which is a key limitation to using any metabarcoding workflow.Furthermore,we(3)evaluatedthereliabilityofthedata provided by our workflow in terms of the number of known species detected,numberofdarktaxaandtheirlikelyvalidity,andvariation in species detected between years. The results obtained with the new workflow show that DNA metabarcoding is feasible for largescale and longterm insect monitoring, and providing insight into insect diversity at scales that have been challenging to study so far. 2 | MATERIALS AND METHODS 2.1 | Sampling SamplingwasconductedaspartofthenationwideGermanMalaise trap monitoring programme (Welti et al., 2022) comprising forests, grassland, agricultural and urban areas (https:// www. ufz. de/ l t e r - d / i n d e x . p h p ? d e = 46285 ). Most of the 31 sites in which the MalaisetrapswereplacedbelongeithertotheGermanLTER-Dnetwork(Haaseetal.,2016; Mirtl et al., 2018)(Long-TermEcological Research)ortothenetworkofnationalnaturallandscapes(ht tps:// natio nalenatur lands chaft en. de). At each site, one to six Malaise traps have been operated to monitor biomass and species composition of flying insects in different habitats. In the present study, we used a total of 1815 Malaise trap samples from 56 locations across Germanyduring2019,with19locationsaddedin2020(Figure 1). Trapswereemptiedevery2 weeksfromthebeginningofApriluntil theendofOctoberinbothyears(approx.15samplesperyear,dependingonsite-specificclimaticconditions).Insectswerecaughtin 80% denatured ethanol and their wet biomass was measured followingWeltietal.(2022).Sampleswerekeptin96%denaturedethanol and protected from light until later genetic analysis. 2.2 | Sample processing Allsamplesweredividedintotwosizeclasses(small≤4 mm;large >4 mm) to increase taxon recovery rates of small taxa (Elbrecht et al., 2021).Samplesspreadonaperforatedplatesieve(4 mmhole diameter) werestirredusingamagneticstirrer(750 rpm)in ethanol(Figure S1),sothatsmallindividualspassedthroughtheholes, whereas the large ones were retained on top of the sieve. The size fractions were homogenized following the protocol described in Buchner,Haase,andLeese(2021),exceptthatthehomogenization timewasreducedto30 s.Thetwosizefractionsweresubsequently pooledataratioof1:4(large200 μL:small800 μL)asrecommended byElbrechtetal.(2021). 2.3 | DNA extraction Generally, the laboratory steps followed the workflow described in Buchneretal.(2021).Allproceduresareavailableasstep-by-step protocolsinaprotocols.iorepository(Buchner,2022b).Allsubsequent steps after sizesorting and homogenization were completed on a Biomek FXP liquid handling workstation (Beckman Coulter, Brea,CA,USA).Aftersamplelysis(Buchner,2022e),sampleswere processed in duplicate during the entire library preparation to controlforpossiblecross-contamination.Additionally,ineach96-well plate12negativecontrolswereincluded.DNAwasextractedusing amagneticbeadprotocol(Buchner,2022a).Extractionsuccesswas verified on a 1% agarose gel. For all samples that did not amplify, the extractionwasrepeatedwithsilicaspincolumns(Buchner,2022a), which were always successful. FIGURE 1 (a)Locationofthein total 75 Malaise traps across Germany, (b)LateralviewofaMalaisetrapwith collection bottle protected from sunlight attheupperleftendand(c)Topviewofa preserved Malaise trap sample spread in a white tray. 17550998, 2025, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.14023 by Capes, Wiley Online Library on [10/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 4 of 15 | BUCHNER et al. 2.4 | Sequencing library preparation The PCR for the metabarcoding library followed a twostep PCR protocol(Zizkaetal.,2019)targetinga205-bpfragment(Vamos et al., 2017) of the cytochrome oxidase c subunit I (COI) gene. DNA was amplified in a first PCR using the Qiagen Multiplex PlusKit(Qiagen,Hilden,Germany)withafinalconcentrationof 1xMultiplexMastermix,200 nMofeachprimer(fwh2F,fwhR2n (Vamosetal.,2017))and1 μLofDNA,filledupwithPCR-grade watertoafinalvolumeof10 μL. The amplification protocol was 5 minofinitialdenaturationat95°C,20cyclesof30 sdenaturationat95°C,90 sofannealingat58°Cand30 sofextensionat 72°C,followedbyafinalelongationstepof10 minat68°C.Each of the PCR plates used in the first step was tagged with a unique combination of inline tags. Additionally, the primers contained a universal binding site for the primer used in the second PCR steptoanneal(Table S1).ThePCRproductswerepurifiedusing abead-basedprotocolandaratioof0.8xandanelutionvolume of40 μL to remove remaining primers and potential primer dimers (Buchner,2022c). InthesecondPCR,DNAwasamplifiedatafinalconcentration of 1×MultiplexMastermix,100 nMofeachprimer(Table S1),1× CorralloadLoadingDye,2 μL of the cleanedup product of the first PCRinafinalvolumeof10 μL.Theamplificationprotocolwas5 min of initial denaturation at 95°C, 25 cycles of 30 s denaturation at 95°C,90 sofannealingat61°Cand30 sofextensionat72°C,followedbyafinalelongationstepof10 minat68°C.PCRsuccesswas visualized on a 1% agarose gel. To achieve a similar sequencing depth, the PCR products were normalized to equal concentrations. Normalization was achieved with a beadbased protocol and a ratio of 0.7×(Buchner,2022d) andanelutionvolumeof40 μL. The whole volume of the normalized samples was then pooled into the final libraries. Libraries were then concentratedusingasilicaspin-columnprotocol(Buchner,2022a). Library concentrations were quantified on a Fragment Analyzer (HighSensitivityNGSFragmentAnalysisKit;AdvancedAnalytical, Ankeny,IA,USA).ThelibrariesweresequencedatMacrogenEurope usingtheHiSeqXplatformwithapaired-end(2 × 150bp,15lanes) kitoratCeGaT(Tübingen,Germany)usingtheMiSeqV2platform (2 × 150bp,1lane). 2.5 | Bioinformatics Rawdataofthesequencingrunsweredelivereddemultiplexedby indexreads.Sincenodifferencesweredetectedbetweensequencingruns,theywereallpooledbeforesubsequentanalyses.Additional demultiplexingoftheinlinetagswasachievedwiththePythonpackage ‘demultiplexer’ (v1.1.0, https:// github. com/ Domin ikBuc hner/ demultiplexer). Reads were further processed with the APSCALE pipeline (Buchner et al., 2022) (v1.4.0, https:// github. com/ Domin ikBuc hner/ apscale)usingdefaultsettings.Briefly,paired-endreads werefirstmergedusingvsearch(Rognesetal.,2016)(v2.21.1)before theprimersequencesweretrimmedusingcutadapt(Martin,2011) (v3.5).Onlyreadswithalengthof205 bp(±10)andwithamaximum expected error of 1 were retained. Identical reads less abundant than 4 were discarded prior to OTU clustering and globally dereplicated before OTUs were clustered based on a similarity threshold of 97%. Reads were then mapped to OTUs. This included singletons. The resulting OTU table was filtered for erroneous OTUs with the LULUalgorithm(Frøslevetal.,2017)asimplementedinAPSCALE. Taxonomic assignment was performed using BOLDigger (Buchner & Leese, 2020)(v1.5.4,https:// github. com/ Domin ikBuc hner/ BOLDi gger). The best hit was determined with the BOLDigger method andtheAPIverificationmethod.ThisresultedinarawOTUtable (Table S2)thatwasusedinsubsequentanalysis. 2.6 | Data filtering To control for possible contamination during the laboratory workflow, the technical replicates of each sample, as well as the negative controls, were merged by summing up the reads, provided that the readswerepresentinbothreplicates.Subsequently,themaximum number of reads per OTU present in all of the negative controls was subtractedfromtherespectiveOTU(Table S3).AllOTUswereanalysed for stopcodons, and any OTU containing stopcodons were removed.Weanalysedtwodatasets:(1)OTUsassignedatthespeciesleveland(2)OTUsassignedtoinsects.Tofurthercleandataset 1, all OTUs sharing species assignments were merged by summing up their reads. Retrieved species names containing numbers or punctuationmarkswerealsoremoved(e.g.incompletedatabaserecords).Theresultingfinalspecieslist(Table S4)forallsampleswas used for the validation procedure. 2.7 | Validation of taxonomic assignment To validate the resulting specieslevel list, three different approaches were used. First, the occurrences in conjunction with their specific locationsofallnamedspecieswerecheckedforplausibilitybytaxonomicexpertsattheEntomologicalSocietyKrefeld,Germany.As abasis,theexpertsusedthedigitallyavailableinsectspeciescataloguefromtheEntomofaunaGermanica(Klausnitzer,2005),which is best resembled through the list from the German Barcode of Life portal(https:// gbol. bolge rmany. de).Thisplausibilitycheckincluded fixingproblemsrelatedtosynonymyandincorporatedtheprimary scientific literature available to the experts. Second, all detected named species were checked for GBIF records within a 200km radius around the given trap. This value was selected to be sufficiently high to accommodate the often patchy and rare species records in GBIF, the large size of the study area and the fact that many flying insect species are highly mobile. To do so, a polygon was drawn around all trap locations with occurrences of the respective species,andtheborderofthispolygonwasthenextendedby200 km (Figure S2).RecordswereextractedfromtheGBIFdatabaseusing 17550998, 2025, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.14023 by Capes, Wiley Online Library on [10/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 5 of 15 BUCHNER et al. the‘rgbif’package(Chamberlainetal.,2022).Lastly,allnamedspecies were checked against the German Barcode of Life database, whichwasdownloadedwithacustomPythonscript(ScriptS1).If a species name occurred in the German Barcode of Life database, it was accepted as valid. Additionally, this step potentially identified species that are unlikely to occur in Germany, supplementing theinformationobtainedfromGBIF.Arecordwasacceptedasvalid when two of the three validation criteria were met, which also helps tocontrolforfalsepositives(Table S5).Suchmulti-criteravalidation approacheshavebeenusedelsewhere(Pereiraetal.,2021). 2.8 | Statistical analysis To assess if sequencing depth was sufficient across all samples, we conducted a readbased rarefaction using a custom python script. This approach involved randomly sampling reads without replacementfromeachsampleinincrementsof0.1%(with50iterationsper step)ofthetotalreadcount.Subsequently,wefittedaMichaelis– Mententype equation to the resulting rarefaction curve. Using this function, we computed OTU richness when doubling the sequencing depth. If OTU richness increased by less than 5% when doubling the sequencing depth, the sequencing depth was considered sufficient (Figure S3; Table S6). Inadditiontoexaminingsequencingrarefactioncurves,wealso examinedrarefactioncurvesforbothvalidatedandplausiblespecies using the iChao2(Chiuetal.,2014)estimator.Wedidsotodetermine the potential influence of collecting more samples from each site, such as by sampling for a longer time period or more frequently, on the number of detected insect species. This was done by randomly drawing subsets of samples (50 iterations each) from the dataset without replacement in increments of 5 from 5 to 1815. 2.9 | Plausible species and dark taxa estimation To address the potential overestimation of species diversity based on OTU numbers we computed the mean number of OTUs per validated species for each of the 20 insect orders within the dataset. This mean was then used to normalize the OTU count for each order. Specifically,wedividedthenumberofOTUsperorderbythecalculated mean, providing an estimate of additional species present in the dataset that have not yet been assigned a species name. We refer to these additional species as plausible species. Furthermore, weobtaineddatafromtheBarcodeofLifedatasystems(accession date:14April2023),includingthenumberofspeciesbarcodedwith a voucher specimen collected in Germany, along with the total number insectspecies known from Germany(Klausnitzer, 2005). This additional information allowed us to distinguish plausible species lacking a reference sequence from potential dark data within the dataset. The described correction was performed on order as well as on family level, but for simplicity we focused the subsequent analysis on the order level. 2.10 | Time and cost estimations To estimate the time and costs needed for our nationwide insect diversity assessment via metabarcoding, we used vendor list prices for materials, as well as runtimes on the liquid handler plus estimates for laboratory setup before and after each step of the workflow. These costs include all labware and chemicals needed to complete the respective step of the protocol. Incubation times are not included in the time estimates because the time can be used to process other samples. The calculated price per sample is based on 1815 samples including replicates and negative controls, resulting in a total number of 4149 individual reactions spread across 44 96well PCR plates. Atotalof10blenderswereusedforhomogenization,whichareincluded in the cost estimate. For sequencing, cost estimates are based on the current costs of 1100 € for 110 Gb output at a commercial service provider (Macrogen Europe) using the NovaSeq 6000 S4, whichhasreplacedtheHiSeqplatform,andasequencingdepthof atleast1.5millionreadspersample(~817Gb).Labourcostswere estimatedat60€perhour,correspondingtoanexperiencedscientist in Germany. The costs for the use of the liquid handler were estimatedbasedonalineardepreciationoveratotalexpectedlifetimeof 10 years.Annualgrossmaintenancecostsof40,000€wereassumed for instruments, resulting in total depreciation and maintenance costsofapproximately4000€forthepresentstudy.Norentaland auxiliarycostsneededtobeappliedinthisstudy. 3 | RESULTS 3.1 | Sequencing results and species validation Weperformedhigh-throughputDNAmetabarcodingincludingreplicates and negative controls on 1815 Malaise trap insect samples (775for2019and1040for2020;Table 1)usinga205-bpfragment of the mitochondrial cytochrome coxidaseI(COI)geneasamarker. Allsamplesandsequencingrunscombinedyielded3,999,082,169 demultiplexedreadpairs.Theaveragereadnumberpersamplein thefinal readtablewas 1,401,469(±SD of618,631).Sequencing depth was sufficient for all samples, with a mean increase in richness of0.17%bydoublingthesequencingdepth(0%–1.46%;Table S6). Sequencing yielded a total of 52,981 raw OTUs, 50,087 of which were assigned to insects. We were able to assign 11,776 species names to a total of 15,042 OTUs. Two thirds of these species(10,803)werevalidatedviathreedifferentcriteriainvolving(i) expertjudgementfromentomologistswithparticularknowledge of longterm Malaise trap community data from German Malaise traps,(ii)acomparisonwithanonlinedatabasethatincludesthe knownGermanspecies(Klausnitzer,2005)(GBOL)and(iii)aGBIF recordcheckwithina200-kmradius.Specieswereregardedas ‘validated’ when two of the three validation criteria supported them. The three different validation criteria showed a pairwise agreementexceeding80%.Taxonomicexpertsvalidatedanadditional 355 species not yet not in the list of species reported from 17550998, 2025, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.14023 by Capes, Wiley Online Library on [10/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 6 of 15 | BUCHNER et al. Germany(Table 1),leadingtoanoverlapof97%withtheGBOL database.GBIFandtaxonomicexpertvalidationhadanoverlap of84%,andGBIFandGBOLof82%(Table S7).Consequently,a total of 35,045 insect OTUs remained unassigned, either because specieslevel reference data are lacking or the species are truly unknown(i.e.darktaxa;Table 1),resultinginalargediscrepancy between the number of named species and recorded insect OTUs. AlthoughsometimesseveralOTUswereassignedtothesamespecies,9061(83.9%)ofourvalidatedspecieswererepresentedbya singleOTU(Table S8). 3.2 | Species richness estimation Our sampling effort captured the majority of OTUs. Based on rarefaction curves, a greatly increased sampling effort in each site, for instance by sampling more frequently or for a longer period at all sites, mayhaveonlyresultedindetectinganadditional4725OTUs(+9.4%) and 934 validated species (+8.6%; Figure 2).Additionally,mostof theOTUswerefoundinbothsamplingyears(36,932 = 73.7%)with moreinsectOTUsoccurringexclusivelyin2020(10,720)thanin2019 (2435).Thesameistrueforthevalidatedinsectspecies.Mostwere found in both years (8456 = 78.3%) but more than three times as manyoccurredexclusivelyin2020(1820)comparedto2019(527). 3.3 | OTU distribution across insect groups and unknown taxa The four most diverse insect orders in Germany (Diptera, Hymenoptera,LepidopteraandColeoptera)werewellrepresented inthedataset(Figure 3,toprow).DipteraandHymenopterawere the most common orders both at the OTU (22,732 = 45.4% and 12,823 = 25.6%)andspecieslevel(3851 = 35.6%and2370 = 21.9%). The 10,803 validated insect species represent 33.5% of the 35,500 insect species known in Germany (Klausnitzer, 2005) and 82.6% of the 13,076 barcoded insect species recorded in the country (Table S9). The percentage of OTUs assigned to named species differed considerablyamonginsectorders,beinghighestforLepidoptera(75%), followedbyColeoptera(59%).LessthanafifthoftheOTUsidentified as Diptera and Hymenoptera could be assigned names at the species-level(17%and18%respectively).Thelowestpercentageof OTUsassignedtonamedspecieswasforOrthoptera(<1%)(Figure 3, bottomleftpanel).ThemeannumberofOTUspervalidatedinsect species varied slightly among insect orders and families, typically between1and1.5,butOrthoptera(specificallytheAcrididae)were represented by >4andZygentoma(specificallyLepismatidae)by~2 OTUsperspecies(Figure 3,bottomrightpanel). Based on the mean number of OTUs per validated insect species calculated at either the order or family level, we estimated that our data set respectively included either an additional 22,496 or 21,043plausibleinsectspecies(Table S10).AllOTUsbelongingto Orthoptera were removed for these estimates to avoid artificially inflating the total number of species (Information S1). The lower of the two numbers is a more conservative estimate, but since we identified 435 different families (see Table S11 for further family level information), our analysis focuses, for simplicity reasons, on the order level. The additional plausible species could be those that either(a)couldnotbeidentifiedtospecieslevelduetoalackofa referenceinthepresentreferencedatabase,(b)hadnotpreviously beenrecordedfromGermanyor(c)representnewspeciestoscience.ExamplesofspeciesnotreportedfromGermanyornewto science(‘potentialdarktaxa’inFigure 3; Table S10)wereparticularly evident in the Diptera and Hymenoptera, for which we respectively found ~6600 and ~1200 species, while the missing reference data aspectaffectedtheassignmentinallordersexcepttheMantodea (with1speciesonly). 3.4 | Time and cost estimation The total costs of the laboratory workflow for duplicate sample processingandnegativecontrols—includingallneededlabware,chemicals,sequencingandsalaries—wereestimatedat88,000€,equivalent toabout46€persample(Table 2; Table S12).Costsforlaboratory materialsaccountedfor12€(26%)ofthetotalcostspersample,and salariesfor34€(74%).Sequencingwasthemostexpensivestep(contributing4.85€persample),followedbyenzymaticstepssuchasPCR (1.65€)andsamplelysis(1.54€).Thetotalprocessingtimeforall 1815sampleswas1030workinghoursor27.5 weeks,equivalentto 141sampleswithin2 weeksforonepersonworkingfulltimeandsupported by one liquid handling robot. Most of the processing time was neededforsamplesize-sortingandhomogenization.Allsubsequent stepswerecompletedwithin6 weeks. TABLE 1 SummarystatisticsofOTUs,assignedinsectspecies and OTUs assigned to validated insect species according to the twooutofthree validation criterion. 2019 (n = 775) 2020 (n = 1040) Σ (n = 1815) Raw OTUs 41,189 50,166 52,981 Insect OTUs 39,367 47,652 50,087 Insect species 9728 11,183 11,776 Validatedspeciesaccordingto: Expert validation 9132 10,475 11,030 GBIF validation 7969 8867 9254 GBOL validation 8777 10,056 10,574 Validatedspecies 8983 10,276 10,803 Plausible species 21,043 Total species 31,846 Note: Plausible species calculated from the average number of OTUs per validated species and family. n = numberofMalaisetrapsamples. Abbreviation:OTUs,OperationalTaxonomicUnits. 17550998, 2025, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.14023 by Capes, Wiley Online Library on [10/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 7 of 15 BUCHNER et al. 4 | DISCUSSION 4.1 | A metabarcoding workflow for largescale biodiversity monitoring We here present a scalable and cost-efficient DNA metabarcoding workflow which allows analysing thousands of specimenand speciesrich samples within several weeks to months depending on theavailableworkforce.TheproposedDNAmetabarcodingworkflow differs from others (e.g. Braukmann et al., 2019; Hardulak et al., 2020; Hausmann et al., 2020)inseveralimportantaspectsthat are key to high sample throughput at reduced costs and processing time while ensuring highquality data. Key differences include: (i)homogenizationofsamplesinpreservativeliquidtoavoidatime- consumingdryingstep(Buchner,Haase,&Leese,2021);(ii)processingofallsamplesinduplicatebeforeDNAextractionasanessential quality assurance measure to reduce the probability of false positive signals;(iii)completionofalllaboratoryproceduresbyanautomated liquidhandlingrobottominimizeprocessingtime(Buchner,Macher, et al., 2021)(exceptforthegelelectrophoresis)andmaximizeconsistency; (iv) a mean sequencing depth increased to ~1.4 M reads persampletoboostspeciesdetectabilityand(v)publicationofall laboratoryproceduresasopen-sourceprotocols(Buchner,2022b) orprogrammes (Buchneret al.,2022; Buchner & Leese, 2020)to ensurefulltransparencyandreproducibilityfollowingtheFAIRprinciples(Wilkinsonetal.,2016). Labour and material costs of metabarcoding protocols are often considered prohibitively high for largescale monitoring programmes (Borrelletal.,2017; Montgomery et al., 2021),frequentlyexceeding 200€persampleandupto400€(Aylagasetal.,2018;Elbrecht et al., 2017; Ji et al., 2013).Theworkflowpresentedhereconsiderably reduces these costs, down to <50 €. This is primarily achieved by automating crucial laboratory steps and by preparing all required solutions instead of purchasing expensive commercial kits. Costs could be further cut in future largescale programmes by ordering chemicals and consumables in bulk and by further reducing reaction FIGURE 2 Toprow:RarefactioncurvesshowingrichnessofinsectOTUsandvalidatedspeciesrichnessthatwereeitherobserved(solid lines)orestimated(dottedlines).Bottomrow:SharedanduniqueabsolutenumbersandproportionsoftheinsectOTUs(leftcircles)and validatedinsectspecies(rightcircles)collectedin2019and2020. 17550998, 2025, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.14023 by Capes, Wiley Online Library on [10/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 8 of 15 | BUCHNER et al. volumes for all laboratory steps involving enzymes, wherever possible(Buchner,Beermann,etal.,2021).Animportantconsideration to lower labour costs is to reduce the processing time per sample, notably for sizesorting and homogenization, ideally by automating thesestepsaswell.Furthermore,expensesrelatedtosamplingin the field are not included in our analysis, although they can easily FIGURE 3 Toppanels:NumberoftotalinsectOTUs,plausibleandvalidatedinsectspeciesperorderidentifiedinthepresentstudy comparedtothetotalnumberofspeciesreportedfromGermanyorsampledinGermanyandincludedintheBOLDdatabase(dataaccessed on14April2023).Bottomleftpanel:PercentageofOTUsassignedtospecies.Bottomrightpanel:MeannumberofdistinctOTUsassigned to a given plausible species. missing referencepotential dark taxa 17550998, 2025, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.14023 by Capes, Wiley Online Library on [10/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 9 of 15 BUCHNER et al. TABLE 2 Costestimatesfortheproposedworkflow:Costestimatesarebasedonvendorlistpricesin2022whereavailable. Processing step Material costs per sample [€] Total material costs [€] Labour costs per sample [€] Time per sample [h:min:s] Total time [h:min:s] Total costs [€] Size-sorting — — 15.09 0:15:00 453:45:00 27,383.81 Homogenization 0.21 379.90 11.50 0:11:26 345:45:00 21,252.45 Samplelysis 1.54 2795.64 1.91 0:01:54 57:37:30 6264.26 DNAextraction 1.02 1845.55 1.09 0:01:05 33:00:00 3823.27 QADNAextraction 0.04 78.48 0.74 0:00:44 22:00:00 1417.24 1st PCR 1.65 2994.43 0.25 0:00:15 07:20:00 3450.83 PCR cleanup 0.50 898.46 0.74 0:00:44 22:00:00 2237.22 2nd PCR 1.65 2994.43 0.37 0:00:22 11:00:00 3663.81 Normalization 0.47 844.47 0.74 0:00:44 22:00:00 2183.23 Pooling <0.01 0.45 0.13 0:00:08 04:00:00 243.86 QClibraries <0.01 10.86 0.13 0:00:08 04:00:00 254.27 Sequencing 4.85 8800.00 — — — 8800.00 Bioinformatic analysis — — 1.59 0:01:35 48:00:00 2890.51 Liquid handler depreciation and maintenance — — — 4000.00 Total 11.92 21,642.67 34.28 34:04 1030:27:00 87,864.76 Note: Time estimates are either handson time in the laboratory or mean runtimes of the robotic protocols. Costs and time required per sample are based on 1815 samples. Total time and total costs are based on the respective number of replicates processed in each step. The bioinformatic analysis was performed on a 128 CPU server with 430 Gb of memory. 17550998, 2025, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.14023 by Capes, Wiley Online Library on [10/11/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License