ferrinetal_c&nmediatewarmingimpactonsoilinsects

This study was conducted at the ForHot research site in Iceland (Sigurdsson et al., 2016) between August 2017 and June 2018 (64°0′N, 21°11′W). Soil type was a Brown Andosol (Arnalds, 2015). Mean annual temperature at the site was 5.1 °C. The coldest and warmest temperatures in the neighboring village of Eyrarbakki in 2016 were -12.3 °C and 21.6 °C, respectively. Average annual precipitation for the same year was 1153 mm (Icelandic Meteorological Office, 2016). The vegetation was an unmanaged grassland dominated by <em>Agrostis capillaris</em> L., <em>Galium boreale </em>L. and<em> Anthoxantum odoratum</em> L.. Vascular plants cover 46% of the area over a moss mat which covers up to 88% of the ground. This grassland has been geothermally warmed since 29 May 2008, when an earthquake transferred geothermal energy from hot groundwater to previously unheated soils (Sigurdsson et al., 2016). Belowground temperatures at 10 cm depth now display a permanent warming gradient reaching +10 °C, with a discreet increase in aboveground temperature of +0.2 °C. The warming has only been mildly disruptive, with seasonality remaining unchanged. Soil humidity was only marginally affected, with volumetric water content changing from 40% to 38%, and water pH increased from 5.6 in unheated soil to up to 6.3 after warming. Geothermal groundwater has remained in the bedrock and has not reached the root zone, thus avoiding direct eco-toxicological effects (Sigurdsson et al., 2016). The resulting stable conditions and lack of artifacts provide a realistic natural belowground experiment on climate change. A N addition experiment was conducted in the same grassland but in areas free from this geothermal warming. Natural N deposition in the area is 1.3 ± 0.1kg N ha-1 y-1 (Leblans et al., 2014). Five transects were established, each one consisting of three 2 x 2 m plots, and each plot at different temperature: an unheated control, a low warming level of ca. +3 °C and a higher warming level of ca. +6 °C above the ambient reference in the control. Soil cores were collected using an auger to a depth of ~10 cm, excluding the O horizon. Soil cores were sampled seasonally four times: August 2017, corresponding to late growing season; November 2017, at start of winter and initial soil freezing; April 2018, with the first soil thaw in un-warmed soils, and June 2018, in the early part of the growing season. We thus collected a total of 20 core samples for each warming treatment (five replicates in four seasons for three temperature levels = 60 samples). The experimental N addition followed a similar approach to the observational study of soil warming, with five replicate transects, with a control plot and an experimental plot where 150 kg N ha-1 had been added annually (in three doses) as NH4-NO3 since 2014. Likewise, a soil core from each replicate was collected for each of the four seasonal samplings, for a total of 20 samples per treatment plot. All samples were immediately sieved to remove roots and stones larger than 2 mm. Fifteen grams of each sample were then frozen in plastic bags in liquid N in the field to immediately stop all biological processes. All frozen samples were freeze-dried in the laboratory. eDNA was extracted from 15 g soil samples as previously described (Taberlet et al., 2012; Zinger et al., 2016). The soil hexapod communities were genetically characterized based on Molecular Operational Taxonomic Units (MOTUs) using the retrieved eDNA and applying a metabarcoding approach. We amplified the 16S mitochondrial rDNA region using the Ins16S_l primer pair (Ins16S_1-F: 5′-TRRGACGAGAAGACCCTATA-3′; Ins16_1-R: 5′-TCTTAATCCAACATCGAGGTC-3′; Clarke et al. 2014). This primer pair, specifically designed for insect metabarcoding, introduces a very limited taxonomic bias and performs very well for identifications at the species level throughout the Hexapoda subphylum (e.g. Kocher et al., 2017; Talaga et al., 2017). PCR amplification was performed in triplicate in 20-μL mixtures consisting of 10 μL of AmpliTaq Gold Master Mix (Life Technologies, Carlsbad, USA), 5.84 μL of nuclease-free Ambion water (Thermo Fisher Scientific, Waltham, USA), 0.25 μM each primer, 3.2 μg of bovine serum albumin (Roche Diagnostic, Basel, Switzerland) and 2 μl of DNA template that was diluted 10-fold to reduce PCR inhibition by humic substances. The thermal profile of the PCR amplification was 40 cycles of denaturation at 95 °C (30 s), annealing at 49 °C (30 s) and elongation at 72 °C (60 s), with a final elongation step at 72 °C for 7 min. Tags had at least five differences between them to minimize ambiguities (Coissac et al., 2012). The sequenced multiplexes comprised extractions/PCR blank controls, unused tag combinations and positive controls (Kocher et al., 2017). The PCR products were then sequenced using the MiSeq platform (Illumina Inc., San Diego, USA), with the expected sequencing depth set at 400 000 reads per sample. The sequences were processed using OBITOOLS software (Boyer et al., 2016). Low-quality sequences (containing Ns, alignment scores &lt;50, lengths &lt;140 bp or &gt;320 bp and singletons) were excluded. The remaining sequences were clustered into MOTUs using SUMACLUST (Mercier et al., 2013) at a threshold of sequence similarity of 97%. The final number of MOTUs after curation was 11785. The hexapod MOTUs were taxonomically assigned using Blast. MOTUs showing &lt;80% similarity with either the local or the EMBL reference databases were removed, leading to 590 MOTUs. These retained MOTUs included taxa from classes Insecta and Entognatha, which both belong to the subphylum Hexapoda. Hereafter we may refer to both classes as insects <em>sensu latto</em> for simplicity. We then applied a post-processing pipeline (Zinger et al., 2021) to minimize PCR and sequencing errors, contaminations and false-positive sequences, and by detailed curation of ecologically incongruent assignments (e.g. taxa with distributions clearly outside the study region). This highly conservative approach retained only 40 species. We then used checklists of Icelandic insect and springtail species and information from previous studies at the same study site (Fjellberg, 2007; Holmstrup et al., 2018) to assess the performance of our eDNA metabarcoding protocol to properly describe the insect communities in the soil.

本研究于2017年8月至2018年6月间，在冰岛ForHot研究站点（Sigurdsson等，2016）开展，站点坐标为64°0′N，21°11′W。供试土壤为棕色暗沃土（Brown Andosol）（Arnalds，2015），站点年平均气温为5.1℃。2016年邻近Eyrarbakki村的极端气温分别为-12.3℃（最低）和21.6℃（最高），同年该区域年平均降水量为1153mm（冰岛气象厅，2016）。
该区域植被为未管理草原，优势物种为*Agrostis capillaris* L.（细弱翦股颖）、*Galium boreale* L.（北拉拉藤）以及*Anthoxantum odoratum* L.（香草茅）。苔藓层覆盖了88%的地面，其上维管植物覆盖率达46%。
该草原自2008年5月29日起开始接受地热增温——当年一场地震将热地下水中的地热能量转移至此前未被加热的土壤中（Sigurdsson等，2016）。目前10cm地下土层存在永久性增温梯度，最高增温幅度达+10℃，地上气温仅小幅升高+0.2℃。增温对季节动态无显著影响，仅对土壤湿度造成极轻微改变：体积含水量从40%降至38%，土壤pH值从未增温土壤的5.6升高至增温后的最高6.3。地热地下水始终留存于基岩中，未抵达植物根区，因此未产生直接的生态毒性效应（Sigurdsson等，2016）。这种稳定的实验条件与无人工干扰的特性，为气候变化研究提供了贴近自然的地下原位实验体系。
本研究在同一草原的非地热增温区域设置了氮添加实验。该区域自然氮沉降量为1.3±0.1kg N ha⁻¹ y⁻¹（Leblans等，2014）。实验共设置5条样带，每条样带包含3个2×2m的样方，分别对应3种温度处理：未加热对照、约+3℃的低增温水平，以及相对于对照环境温度约+6℃的高增温水平。
采用土钻采集10cm深的土壤芯样，排除O层。分别于4个季节开展采样：2017年8月（生长季末期）、2017年11月（冬季初始及土壤初次冻结）、2018年4月（未增温土壤首次解冻）、2018年6月（生长季早期）。每个温度处理组共采集20个土壤芯样，3个温度水平×4个季节×5个重复样总计60个样品。
氮添加实验的设置与地热增温实验类似，共设置5条重复样带，包含对照样方与实验样方。实验样方自2014年起，每年以NH₄-NO₃的形式分3次施加150kg N ha⁻¹的氮肥。每个处理样方同样于4个季节各采集土壤芯样，每个样方总计采集20个样品。
所有采集的样品立即过2mm筛，去除根系与直径大于2mm的石块。随后称取15g样品置于塑料袋中，于野外用液氮快速冷冻以终止所有生物过程。所有冷冻样品均在实验室进行冷冻干燥。
采用15g土壤样品提取环境DNA（eDNA），方法参考已发表方案（Taberlet等，2012；Zinger等，2016）。基于获取的eDNA，采用元条形码（metabarcoding）技术对土壤六足类（Hexapoda）群落进行分子表征，通过分子操作分类单元（Molecular Operational Taxonomic Units, MOTUs）进行聚类分析。
使用Ins16S_l引物对扩增线粒体16S rDNA区域，引物序列为：Ins16S_1-F: 5′-TRRGACGAGAAGACCCTATA-3′；Ins16_1-R: 5′-TCTTAATCCAACATCGAGGTC-3′（Clarke等，2014）。该引物为专门针对昆虫元条形码设计的引物，仅带来极轻微的分类偏差，在六足亚门的物种水平鉴定中表现优异（如Kocher等，2017；Talaga等，2017）。
PCR扩增设置3次技术重复，反应体系总体积20μL，包含10μL AmpliTaq Gold Master Mix（Life Technologies，美国卡尔斯巴德）、5.84μL无核酸酶Ambion水（赛默飞世尔科技，美国沃尔瑟姆）、0.25μM上下游引物各1份、3.2μg牛血清白蛋白（罗氏诊断，瑞士巴塞尔），以及2μL稀释10倍的DNA模板以降低腐殖物质对PCR的抑制作用。PCR扩增程序为：95℃变性30s、49℃退火30s、72℃延伸60s，共40个循环，最后于72℃延伸7min。为减少测序歧义，不同标签之间至少存在5个碱基差异（Coissac等，2012）。测序多重文库包含提取/PCR空白对照、未使用的标签组合以及阳性对照（Kocher等，2017）。采用MiSeq平台（Illumina Inc.，美国圣迭戈）对PCR产物进行测序，设定每个样本的预期测序深度为40万条reads。
序列处理采用OBITOOLS软件（Boyer等，2016），过滤掉低质量序列（包含模糊碱基N、比对分数<50、长度<140bp或>320bp的序列以及单序列）。将剩余序列以97%的序列相似度阈值，通过SUMACLUST软件（Mercier等，2013）聚类为MOTUs，经整理后共得到11785个MOTUs。采用BLAST对MOTUs进行分类注释，移除与本地数据库或EMBL数据库相似度<80%的MOTUs，最终得到590个有效MOTUs，这些MOTUs分属于昆虫纲（Insecta）和内口纲（Entognatha），二者均隶属于六足亚门（Hexapoda）。为简化表述，下文将这两个类群统称为广义昆虫（*sensu lato*）。
随后采用后处理流程（Zinger等，2021）进一步减少PCR及测序错误、污染序列和假阳性序列，并对生态学上不合理的分类注释进行人工审核（例如分布明显超出研究区域的类群）。经过这种高度严格的筛选流程，最终仅保留40个物种。我们参考冰岛昆虫及跳虫物种名录，以及该研究站点此前的相关研究成果（Fjellberg，2007；Holmstrup等，2018），评估了本研究采用的eDNA元条形码方案对土壤昆虫群落的表征性能。